Transcript SAR Contributions to Ship Detection and Characterization

SAR Contributions to Ship
Detection and Characterization
J.K.E. Tunaley
London Research and Development Corporation,
114 Margaret Anne Drive,
Ottawa, Ontario K0A 1L0
1-613-839-7943
http://www.london-research-and-development.com/
Outline
Ship Detection
– Robustness, Timeliness, Development Costs,
K-distribution
Detection Threshold
Wakes from Surface Ships and Internal
Wave Wakes
Space-Based AIS Performance
www.london-research-and-development.com
2
Ship Detection
Clutter Statistics
– Remove the ship from a clutter cell
Cut out the ship
Handle statistically
Estimate Clutter Parameters
– Choose a statistic
Moments (simple)
Logarithms
www.london-research-and-development.com
3
Threshold Calculation
Density by steepest descents
– J.K.E. Tunaley, “K-Distribution Algorithm”, Sept. 2010 (www.londonresearch-and-development.com/K-Distribution Algorithm.Version2.pdf )
– Avoid Bessel functions
Distribution/Threshold
– Suggest using expansion with polynomial
correction in shape parameter and number of
looks
www.london-research-and-development.com
4
Probability Density Comparison
THRESHOLDS OF
DETECTION
L=1
L=4
PFA

Accurate
10-9
0.5
214.7
214.8
91.59
91.62
10-9
5.0
47.49
47.50
18.796
18.800
10-9
50.0
24.24
24.24
8.841
8.842
10-6
0.5
95.43
95.55
46.40
46.43
10-6
5.0
25.69
25.70
11.263
11.267
10-6
50.0
15.337
15.338
6.128
6.128
Approx. Accurate Approx.
www.london-research-and-development.com
5
Parameter Estimation
Fisher information
J.K.E. Tunaley, “Ship Detection”, December 2010, (www.londonresearch-and-development.com/Ship Detection.Version3.pdf )
Mean can be estimated reasonably
accurately
In spiky clutter shape parameter tends to
require 10,000 resolution cells using
moments
www.london-research-and-development.com
6
Detection Threshold (PFA=1.0E-9)
Detection Thresholds
1000
100
10
1
0.1
1
10
100
Nu
Fig. 8. Detection thresholds for PFA = 10-9, L = 4 and N = 100 (), N = 1000 (),
N = 10,000 () and N =  ().
www.london-research-and-development.com
7
Ship Image Information
Position
Length (may be poor estimate)
– Heavy ship motion in high sea states
Velocity (from wake displacement)
– Ocean going ships usually create visible wake
D.M. Roy and J.K.E. Tunaley, “Visibility of the Turbulent Wake”,
March 2010 (www.london-research-and-development.com/Visibility
of Turbulent Wakes.Ver2a.pdf )
– Wake characteristics depend on propulsion
system; screw number and sense of rotation
www.london-research-and-development.com
8
Internal Wave Wake
0
100
Depth (m)
200
300
400
500
600
700
0
0.005
0.01
0.015
0.02
Brunt-Vaisala Frequency (rad/s)
Typical Brunt-Vaisala Vertical Profile.
2


N
2
 k  2  1Q  0
2
dz


2
d Q
www.london-research-and-development.com
9
0.08
0.04
0.07
0.03
0.06
0.02
Wave Amplitude
Wave Amplitude
Zeroth and First Modes
0.05
=0.005
0.04
0.03
0.02
0
-0.01
-0.02
=0.01
0.01
0.01
-0.03
0
-0.04
0
100
200
300
400
500
600
700
0
100
200
Depth (m)
300
400
500
600
700
Depth (m)
Sinuous Modes
Varicose Modes
www.london-research-and-development.com
10
Frequency-Wave Number
0.016
Wave Frequency (rad/s)
0.014
0.012
0.010
Modes 0, 1, 2
0.008
0.006
0.004
0.002
0.000
0.0
0.1
0.2
0.3
0.4
0.5
Wave Number (rad/m)
Determines phase and group velocities
www.london-research-and-development.com
11
Crest Pattern
6
5
y (km)
4
3
2
1
0
0
5
10
15
20
25
30
35
x (km)
Zeroth mode crest pattern for a source moving horizontally
at 5 m/s in the above profile.
www.london-research-and-development.com
12
Internal Wave Wake Conclusions
From Crest Pattern
– Ship velocity from angle of wake (if strength of
layer known)
– Maximum B-V frequency
From Amplitudes (Tentative)
– Layer thickness/Position of vessel in layer
– Vessel size
www.london-research-and-development.com
13
Space-Based AIS Performance
Problems
– Signal Collisions and Range Overlap
– Message 27 solution with AIS channels 3 & 4
(ITU-R M.2169)
FFI Theoretical Model
– Based on signal corruption with one or more
signal collisions
Extension to multiple collisions
– (www.london-research-and-development.com/Space-Based-AISPerformance.pdf )
www.london-research-and-development.com
14
Multiple Collision Model
1
q=0.8
Probability of Detection
0.9
0.8
0.7
0.6
q=0.5
0.5
0.4
0.3
q=0
0.2
Single Collision
0.1
0
0
10000
20000
30000
40000
50000
60000
70000
80000
Number Ships
q is the probability that a collision can be tolerated
www.london-research-and-development.com
15
ITU-R.M2169
Probability of
Detection
Detection Statistics with 3rd AIS Satellite Channel
(Assuming Uniform Ship Distribution)
100%
80%
60%
3-min
interval
3-minute interval
6-min interval
6-minute interval
3-min,
2-chan
40%
20%
0%
0
20000
40000
60000
80000
Number of Ships in Satellite
Footprint
www.london-research-and-development.com
16
Theoretical Performance
1
Probability of Detection
0.9
3-min, 2 chan
0.8
0.7
0.6
0.5
6-min, 1 chan
0.4
0.3
0.2
3-min, 1 chan
0.1
0
0
10000
20000
30000
40000
50000
60000
70000
80000
Number Ships
www.london-research-and-development.com
17
Space-Based AIS Conclusions
Model 1 is based on the receiving system
resolving a fixed number of collisions
Model 2 is based on the system resolving
an average number of collisions
Model 1 more or less consistent with
simulations in ITU-R M.2169
www.london-research-and-development.com
18
END
Thank You All!
www.london-research-and-development.com
19