Transcript Use of Radiography for Maize Seed Evaluation
Slide 1
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 2
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 3
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 4
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 5
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 6
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 7
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 8
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 9
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 10
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 11
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 12
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 13
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 14
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 15
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 16
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 17
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 18
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 19
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 20
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 21
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 22
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 23
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 24
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 25
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 26
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 27
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 28
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 29
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 30
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 31
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 32
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 33
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 34
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 35
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 36
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 37
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 38
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 39
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 40
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 41
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 42
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 43
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 44
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 45
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 46
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 47
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 2
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 3
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 4
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 5
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 6
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 7
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 8
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 9
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 10
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 11
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 12
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 13
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 14
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 15
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 16
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 17
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 18
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 19
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 20
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 21
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 22
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 23
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 24
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 25
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 26
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 27
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 28
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 29
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 30
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 31
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 32
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 33
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 34
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 35
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 36
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 37
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 38
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 39
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 40
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 41
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 42
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 43
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 44
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 45
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 46
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010
Slide 47
X-Ray for Agricultural
Applications
Dave Utrata
Iowa State University
Center for NDE
IPRT Company Assistance – NDE Group
Alexandra Black, Brett Barnes
Iowa State University
Three Rivers Technical Conference
Elk River, MN
August 3-4, 2010
Presentation Outline
Background of Company Assistance
program
◦ Current efforts, interests
Review of progress in three areas
◦ Detection of frost damage in maize seeds
◦ Distinguishing between haploid/hybrid
condition in corn seeds
◦ Imaging of root development
Summary
◦ Applications, potential, challenges
Three Rivers Tech Conf - 3-4 Aug 2010
Iowa State University –
Institute for Physical Research and Testing
IPRT Company Assistance
Technology Commercialization - accelerate
commercialization of ISU and company-developed
technology through research and development
assistance.
Technical Assistance - provide short-term, no-cost
technical assistance to Iowa companies in materials and
nondestructive evaluation.
Three Rivers Tech Conf - 3-4 Aug 2010
Operational Guide – NDE Group
Explain & demonstrate the principles of various
inspection methods
Perform feasibility studies to determine which NDE
method will work best for your application
Develop inspection protocol for new designs or
evaluate current inspection procedures
Provide information on sources of equipment, testing
laboratories
Help manufacturers become educated
consumers of NDE
Three Rivers Tech Conf - 3-4 Aug 2010
Beyond the usual
suspects of
fabricated metal,
joining operations,
current NDT
practitioners…
“Round up the usual suspects”
Three Rivers Tech Conf - 3-4 Aug 2010
Nontraditional applications of NDE
Detection of frost damage in maize (corn)
seed – Seed Science Institute, ISU
Distinguishing between haploid v. hybrid
conditions in corn – Agronomy Dept. ISU
Visualizing root growth – North Central
Regional Plant Introduction Station
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of
frost damage
in seed
Three Rivers Tech Conf - 3-4 Aug 2010
Brief background of some work in field
X-ray used to look into potential for
characterizing seeds
◦ Often adapted from medical systems, very often
restricted to film
◦ “Soft” x-rays
Feasibility of digital high speed evaluation
demonstrated using industrial system with
pattern recognition for sorting
Three Rivers Tech Conf - 3-4 Aug 2010
Detection of frost damage in corn
seed
Can x-ray methodologies provide useful
assessment of seed structure, viability?
◦
Use modern industrial equipment: microfocus
X-ray tube, all digital imaging
Differentiate between control group (no frost
damage) and damaged maize seeds (artificial
frost cycle)
◦
◦
B37 inbred line, male parent IRF311, hybrid 3
Moisture content at harvest 50%
Provide more rapid alternative to current
sectioning & staining (tetrazolium) test
Three Rivers Tech Conf - 3-4 Aug 2010
Results
Using Flat Panel
◦ Seven samples from both control and
frost damaged groups were examined
Three Rivers Tech Conf - 3-4 Aug 2010
Control seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Frost damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Qualitative difference apparent
between groups – loss of internal
structure
Next: quantitative analysis of images
to remove subjective evaluation
Three Rivers Tech Conf - 3-4 Aug 2010
X-ray Luminance Profiles of Maize
Seeds Method
Line profile used to
see if there is a
numerical difference
in the luminance
profiles between a
control corn seed and
a frost damaged seed
Three Rivers Tech Conf - 3-4 Aug 2010
Trends
In general, the
luminance profile of a
good seed with no
frost damage is more
erratic then the
smoother bad seed’s
profile
This trend presumably
reveals frost damage to
internal seed structure
2500
2400
2300
2200
Luminance
2100
2000
Control
Good
1900
Frost
Bad
Damage
1800
1700
1600
300
400
500
600
Horizontal Position
700
800
Three Rivers Tech Conf - 3-4 Aug 2010
Further analyses of
maize seed X-rays
The luminance profiles were exported
to spreadsheets
Two methods were used to
numerically separate the control from
the frost damaged seeds
◦ Large Difference
◦ Large Slope
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Difference
2500
2400
Valid Region
2300
Luminance
2200
Total Range
2100
2000
4
1 2
1900
In the profile shown
here, there are 4
differences in the
“valid” region
(internal seed
features) that are
larger then the
threshold.
3
1800
Large Difference Threshold
1700
1600
300
400
500
600
700
800
Horizontal Position
Large Difference Threshold = (Scale Factor)*(Total Range)
Three Rivers Tech Conf - 3-4 Aug 2010
Analysis : Large Slope
2500
2400
Valid Region
For example, the
profile of this seed
has 5 Large Slopes
in the valid region.
2300
Luminance
2200
2100
2000
2 3
1900
1
1800
4
6
5
1700
1600
300
400
500
600
700
800
Horizontal Position
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Difference
12
10
# of Large Differences
Large Difference Data
Frost
Sample
Control
Damage
1
10
1
2
12
3
3
12
0
4
7
0
5
4
1
6
0
4
7
11
2
Avg
8
1.5
Scale Factor = 0.055
8
Control
Good
6
Frost
Bad
Damage
4
2
0
There are typically more “large differences”
within radiographs of control seeds than for
frost damaged seeds.
Three Rivers Tech Conf - 3-4 Aug 2010
Results : Large Slope
Avg
29.2
Slope Threshold = 10
45
40
35
# of Large Slopes
Large Slope Data
Frost
Sample
Control Damage
1
40
20
2
27
18
3
43
10
4
21
18
5
13
9
6
19
40
7
42
27
30
Control
Good
25
Bad
Frost
20
Damage
15
10
5
0
20.3
There are more “large slopes” on average within
radiographs of control seeds than for frost
damaged seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
The internal structure of maize seeds may
be revealed using standard industrial
radiographic equipment
The adverse effects of frost damage were
observable in this small study,
independent of seed knowledge or
significant manipulation of test conditions
Three Rivers Tech Conf - 3-4 Aug 2010
Problems with analysis
The threshold multiplier has to be adjusted
to maximize the divergence in counts/core
counts between averaged results for good and
bad learning sets
◦ .055 used initially, .038 probably better
Similarly, the large derivative threshold is
chosen to maximize this difference
◦ 10 used initially, 12 probably better
The “valid” region of interest was subjectively
chosen
Three Rivers Tech Conf - 3-4 Aug 2010
Positive modification (attempts)
Change source to detector distance
◦ Magnified image
Structural variation seen in more detail over greater
pixel count
◦ Work at lower voltage (40 kV 25 kV)
Strive for contrast variation afforded by lower energies
More automated line profile placement
◦ Remove subjective analysis
Three Rivers Tech Conf - 3-4 Aug 2010
New Seed Radiographs
Control
Frost
Three Rivers Tech Conf - 3-4 Aug 2010
Frost – Saturday
Radiograph with Equator, Automatic &
Manual Profiles Highlighted
Line Profiles
Frost - Saturday
250
Auto 3
Manual
Auto 2
Luminance
200
150
Auto 1
Auto 2
100
Auto 3
Manual
Auto 1
Equator
50
0
0
200
400
600
800
1000
Horizontal Position
Auto 1 Auto 2 Auto 3 Man.
Large Difference
Counts
11
226
241
261
Large Slope Counts
1
5
1
1
Three Rivers Tech Conf - 3-4 Aug 2010
Large Difference Results Summary
Large Difference Count Table
Auto 1
Ctrl-Sun
216
Ctrl-Mon
12
Ctrl-Tues
7
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
Auto 2
228
14
15
Auto 3
250
9
28
Manual
233
214
25
15
41
240
226
176
48
276
241
129
35
99
261
16
36
80
11
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Difference Counts for Full Batch Seed X-Ray
300
250
200
Auto 1
150
Auto 2
Auto 3
100
Manual
50
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Large Slope Results Summary
Large Slope Count Table
Auto 1
Ctrl-Sun
1
Ctrl-Mon
4
Ctrl-Tues
1
Frost-Thur - 13.5
Frost-Thur - 11
Frost-Fri
Frost-Sat
1
1
1
1
Auto 2
1
1
1
Auto 3
1
1
1
Manual
1
1
2
1
1
1
5
1
1
1
1
1
2
1
1
Difference is not as clear as was
earlier believed;
faulty logic due to magnified image?
Large Slope Counts for Full Batch Seed X-Ray
6
5
4
Auto 1
3
Auto 2
Auto 3
2
Manual
1
0
Ctrl-Sun
Ctrl-Mon
Ctrl-Tues
Frost-Thur -Frost-Thur - Frost-Fri
13.5
11
Frost-Sat
Three Rivers Tech Conf - 3-4 Aug 2010
Radiographic detection of frost damage
Differentiation is certainly possible
◦ To what degree?
50% moisture at frost is no big challenge
Algorithm/logic needs to be refined
◦ Slope/difference thresholds need account for
morphology changes over higher pixel count
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid & hybrid seeds
The inducing of haploid corn from which to generate more
vigorous hybrid lines is desired
◦ Currently takes six to eight generations of corn, or 3-4 years, to
produce the inbred lines to cross for hybridization
Effective method for inducing haploids would reduce this time to
one or two generations
◦ Allow for the faster production of hybrid lines
◦ Greater freedom to experiment with other seed attributes
Distinguishing between stained and unstained seeds (current
method) takes a trained eye and much time
CT scans on seeds were performed
◦ Embryos of the haploid seeds were ~7% smaller by volume than the
hybrid seeds
Three Rivers Tech Conf - 3-4 Aug 2010
Comparison of Hybrid & Haploid Collections
Haploid
Hybrid
Three Rivers Tech Conf - 3-4 Aug 2010
Data Analysis Procedure
A section of the radiograph was
selected and a histogram of that
luminance data was generated
Two sections were analyzed for each
sample:
◦ 600 pixel diameter circle (not adequate)
◦ Oval with 900 & 600 pixel major and minor
radii
Three Rivers Tech Conf - 3-4 Aug 2010
Haploid – Saturday – 600 x 900 pixel oval
Pixel Frequency
Haploid Saturday
0
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Luminance profiles of Haploid and
Hybrid Seeds
90000
Haploid & Hybrid Sets
Do lower luminance values
reflect smaller embryos?
70000
80000
Pixel Frequency
60000
50000
40000
30000
20000
10000
0
0
1000
2000
3000
4000
5000
6000
Luminance
Three Rivers Tech Conf - 3-4 Aug 2010
Results not definitive
Scatter in data may prove analysis to be
hopelessly clumsy
Is data showing density/volumetric info of
interest?
Placement/size of measurement area still
problematic
Three Rivers Tech Conf - 3-4 Aug 2010
Interest in observing root structure
Commercial system prohibitively
expensive
What can be done?
Three Rivers Tech Conf - 3-4 Aug 2010
Styrofoam Holder
Configured for two plants
Three Rivers Tech Conf - 3-4 Aug 2010
30 kV
Energy: 30 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: ~24”
Three Rivers Tech Conf - 3-4 Aug 2010
40 kV
Energy: 40 kV
Current: 250 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
50 kV
Energy: 50 kV
Current: 200 µA
Spot Size: Small
FPS: 1
Window Average: 10
Source to Detector
Distance: 26”
Source to Sample
Distance: 20”
Three Rivers Tech Conf - 3-4 Aug 2010
Images of “Sprout & Grow”
with Bean & Pea plants
Energy: 25 kV
Current: 250 µA
Spot Size: Small
Window Average: 10
Source to Detector Distance: 24”
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
A pea and
bean seed
grown in water
and Styrofoam
beads in half of
the container.
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Front View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Sprout & Grow Holder – Rear View
X-Ray Image
Optical Image
Three Rivers Tech Conf - 3-4 Aug 2010
Conclusions
X-ray imaging of roots quite possible
◦ Use Styrofoam beads, not standard growing
media
Scale of imaging important
◦ Subtle initial growth or large developed
network
Three Rivers Tech Conf - 3-4 Aug 2010
Overall conclusions
Digital radiography is
powerful tool for
nontraditional
applications
Communication of
needs, tradeoffs
necessary
◦ Content revealed v.
inspection rate
Three Rivers Tech Conf - 3-4 Aug 2010
Thank you
Three Rivers Tech Conf - 3-4 Aug 2010