Quality Grading of Tomatoes AMHPAC 2015 - Welcome! Optical Setup Resolution 0.4mm per pixel External Quality.
Download ReportTranscript Quality Grading of Tomatoes AMHPAC 2015 - Welcome! Optical Setup Resolution 0.4mm per pixel External Quality.
Slide 1
Quality Grading of Tomatoes
AMHPAC 2015 - Welcome!
Slide 2
Optical Setup
Resolution
0.4mm per pixel
External Quality
Slide 3
Optical Setup
•
•
•
•
•
•
•
The less camera’s the less adjustments
High Resolution digital fire wire camera’s
Infrared and color image taken at the same time
Mirrors above the lanes
2 views of the fruit in one image
(one camera)
1
Stereo vision setup
20 image of the fruit while rotating
External Quality
2
Slide 4
Optical Setup
• Illumination with LED’s
• Both infrared and visible light
• LED’s don’t age over time
• LED’s have a very long lifetime
External Quality
Slide 5
Color Image
Slide 6
Infrared Image
Slide 7
Measurable Parameters
Size
Shape
• Diameter
• Volume
• Weight
External Quality
Color
• Average
Color
• % Color
Blemish
Slide 8
Varieties
Mature
Red
Tomatoes
Mature
Green
Tomatoes
External Quality
Roma
Tomatoes
Cherry
Tomatoes
Grape
Tomatoes
Slide 9
Example Solutions
External Quality
Slide 10
External Color
GREEN
BREAKERS
TURNING
PINK
LIGHT RED
RED
Slide 11
External Color
Slide 12
External Color
Slide 13
External Color
Slide 14
External Quality
Slide 15
Neural Networks
External Quality
Slide 16
Neural Networks
External Quality
Slide 17
Neural Networks
External Quality
Slide 18
Neural Networks
External Quality
Slide 19
Neural Networks
External Quality
Slide 20
Neural Networks
External Quality
Slide 21
Neural Networks
External Quality
Slide 22
Shape Defects
External Quality
Slide 23
Grow Damage
External Quality
Slide 24
Nose Rotting
External Quality
Slide 25
Caterpillar Damage
External Quality
Slide 26
Crownless
External Quality
Slide 27
Corking
External Quality
Slide 28
Cut Damage
External Quality
Slide 29
Sunburn
External Quality
Slide 30
Catface
External Quality
Slide 31
Internal Quality
Slide 32
Near Infrared Spectroscopy
Near Infrared (NIR) spectroscopy studies the
interaction between light and the analysed material
OPTICAL SIGNATURE
measured and analyzed
Internal Quality
Slide 33
Near Infrared Spectroscopy
Reflection
Transmission
Light from source
Internal Quality
Interactance
Light to detector
Optrode
Slide 34
What is Near Infrared Spectroscopy
Absorption of radiation
related to
chemical content
[ OH CH NH ]
absorbencies due to different
Chemical bonds occur at different
wavelengths.
•
•
•
•
Starch and sugars
Pigments
Carotenoids / chlorophyll
Water status
Internal Quality
Slide 35
Fruit Modeling and Calibration
Inscan Calibration – prediction model
The Inscan system does not directly measure the quality parameters but measures the light
interaction with the product and relates this information to the requested quality parameter. To
establish the relation (prediction model) the Inscan system must be calibrated.
CHEMOMETRICS ( PLS MODELING)
1600
200
1400
Up to 8 parameters
150
1200
100
1000
50
800
0
600
-50
400
-100
200
0
-150
0
20
40
60
80
100
120
140
160
180
200
OPTICAL SIGNATURE
Internal Quality
Brix
Internal decay
Dry matter
Slide 36
Measurable Parameters
Internal Quality
Slide 37
Slide 38
Roma Tomatoes – Internal Color
• Camera system delivers weak
information about the internal color
of tomato products
• IQA (Internal Quality Analysis)
system was used in order to
estimate the internal color of Roma
tomatoes.
• 40 Roma tomatoes were measured
using Internal Quality Analysis
system and optical scans recorded.
• The tomatoes were sliced to
observe internal meat color and
internal maturity recorded.
Internal Quality
Slide 39
Roma Tomatoes – Internal Color
•
•
•
An internal color model was
developed
Ripening index model with maturity
classes 1 to 10.
The prediction model is strongly
affected by chlorophyll content of
the product
Internal Quality
Slide 40
Roma Tomatoes – Internal Color
Tomato #
Predicted score
b9
9
b10
9
b11
8
b12
6
b13
6
b14
7
b15
8
b16
7
Slide 41
Ripe Green Tomatoes
• Tomatoes can have a wide
range of maturity (time to
ripen) with no difference in
appearance.
• Very difficult to determine
immature green visually
Internal Quality
Slide 42
Ripe Green Tomatoes
• Develop a ripening maturity model
• Take internal scans daily of tomato
• Record observed maturity
• Plot of maturity vs time
• All tomatoes started green
• We can see that the ripening process is
very different for each tomato
Internal Quality
Slide 43
Ripe Green Tomatoes
• All tomatoes follow the same
sigmoid curve during ripening
• Shift scans in time so that
sigmoid curves line up
• We can now create a new model
that from the data that represents
Maturity vs Relative Harvest Day
Piece 1 is Relative Harvest Day +1
Piece 2 is Relative Harvest Day +2
Piece 3 is Relative Harvest Day -2
Internal Quality
Slide 44
Ripe Green Tomatoes
Day 1
Day 3
Low
Maturity
High
Maturity
Internal Quality
Day 6
Day 7
Slide 45
Summary
External
Internal
• Size
• Diameter
• Volume
• Weight
• Shape
• Color
• Average
Color
• % Color
• Blemish
• Internal Color
• Softness
• Maturity
• Brix
• Acid
Quality Grading of Tomatoes
AMHPAC 2015 - Welcome!
Slide 2
Optical Setup
Resolution
0.4mm per pixel
External Quality
Slide 3
Optical Setup
•
•
•
•
•
•
•
The less camera’s the less adjustments
High Resolution digital fire wire camera’s
Infrared and color image taken at the same time
Mirrors above the lanes
2 views of the fruit in one image
(one camera)
1
Stereo vision setup
20 image of the fruit while rotating
External Quality
2
Slide 4
Optical Setup
• Illumination with LED’s
• Both infrared and visible light
• LED’s don’t age over time
• LED’s have a very long lifetime
External Quality
Slide 5
Color Image
Slide 6
Infrared Image
Slide 7
Measurable Parameters
Size
Shape
• Diameter
• Volume
• Weight
External Quality
Color
• Average
Color
• % Color
Blemish
Slide 8
Varieties
Mature
Red
Tomatoes
Mature
Green
Tomatoes
External Quality
Roma
Tomatoes
Cherry
Tomatoes
Grape
Tomatoes
Slide 9
Example Solutions
External Quality
Slide 10
External Color
GREEN
BREAKERS
TURNING
PINK
LIGHT RED
RED
Slide 11
External Color
Slide 12
External Color
Slide 13
External Color
Slide 14
External Quality
Slide 15
Neural Networks
External Quality
Slide 16
Neural Networks
External Quality
Slide 17
Neural Networks
External Quality
Slide 18
Neural Networks
External Quality
Slide 19
Neural Networks
External Quality
Slide 20
Neural Networks
External Quality
Slide 21
Neural Networks
External Quality
Slide 22
Shape Defects
External Quality
Slide 23
Grow Damage
External Quality
Slide 24
Nose Rotting
External Quality
Slide 25
Caterpillar Damage
External Quality
Slide 26
Crownless
External Quality
Slide 27
Corking
External Quality
Slide 28
Cut Damage
External Quality
Slide 29
Sunburn
External Quality
Slide 30
Catface
External Quality
Slide 31
Internal Quality
Slide 32
Near Infrared Spectroscopy
Near Infrared (NIR) spectroscopy studies the
interaction between light and the analysed material
OPTICAL SIGNATURE
measured and analyzed
Internal Quality
Slide 33
Near Infrared Spectroscopy
Reflection
Transmission
Light from source
Internal Quality
Interactance
Light to detector
Optrode
Slide 34
What is Near Infrared Spectroscopy
Absorption of radiation
related to
chemical content
[ OH CH NH ]
absorbencies due to different
Chemical bonds occur at different
wavelengths.
•
•
•
•
Starch and sugars
Pigments
Carotenoids / chlorophyll
Water status
Internal Quality
Slide 35
Fruit Modeling and Calibration
Inscan Calibration – prediction model
The Inscan system does not directly measure the quality parameters but measures the light
interaction with the product and relates this information to the requested quality parameter. To
establish the relation (prediction model) the Inscan system must be calibrated.
CHEMOMETRICS ( PLS MODELING)
1600
200
1400
Up to 8 parameters
150
1200
100
1000
50
800
0
600
-50
400
-100
200
0
-150
0
20
40
60
80
100
120
140
160
180
200
OPTICAL SIGNATURE
Internal Quality
Brix
Internal decay
Dry matter
Slide 36
Measurable Parameters
Internal Quality
Slide 37
Slide 38
Roma Tomatoes – Internal Color
• Camera system delivers weak
information about the internal color
of tomato products
• IQA (Internal Quality Analysis)
system was used in order to
estimate the internal color of Roma
tomatoes.
• 40 Roma tomatoes were measured
using Internal Quality Analysis
system and optical scans recorded.
• The tomatoes were sliced to
observe internal meat color and
internal maturity recorded.
Internal Quality
Slide 39
Roma Tomatoes – Internal Color
•
•
•
An internal color model was
developed
Ripening index model with maturity
classes 1 to 10.
The prediction model is strongly
affected by chlorophyll content of
the product
Internal Quality
Slide 40
Roma Tomatoes – Internal Color
Tomato #
Predicted score
b9
9
b10
9
b11
8
b12
6
b13
6
b14
7
b15
8
b16
7
Slide 41
Ripe Green Tomatoes
• Tomatoes can have a wide
range of maturity (time to
ripen) with no difference in
appearance.
• Very difficult to determine
immature green visually
Internal Quality
Slide 42
Ripe Green Tomatoes
• Develop a ripening maturity model
• Take internal scans daily of tomato
• Record observed maturity
• Plot of maturity vs time
• All tomatoes started green
• We can see that the ripening process is
very different for each tomato
Internal Quality
Slide 43
Ripe Green Tomatoes
• All tomatoes follow the same
sigmoid curve during ripening
• Shift scans in time so that
sigmoid curves line up
• We can now create a new model
that from the data that represents
Maturity vs Relative Harvest Day
Piece 1 is Relative Harvest Day +1
Piece 2 is Relative Harvest Day +2
Piece 3 is Relative Harvest Day -2
Internal Quality
Slide 44
Ripe Green Tomatoes
Day 1
Day 3
Low
Maturity
High
Maturity
Internal Quality
Day 6
Day 7
Slide 45
Summary
External
Internal
• Size
• Diameter
• Volume
• Weight
• Shape
• Color
• Average
Color
• % Color
• Blemish
• Internal Color
• Softness
• Maturity
• Brix
• Acid