Image-based change detection to reduce false alarms in the - - PowerPoint PPT Presentation

image based change detection to reduce false alarms in
SMART_READER_LITE
LIVE PREVIEW

Image-based change detection to reduce false alarms in the - - PowerPoint PPT Presentation

Image-based change detection to reduce false alarms in the Vision1200 synthetic aperture sonar Dr. C. Erdmann and Dr. J. Groen a sound decision a sound decsion The ATLAS ELEKTRONIK Group/ 1 Image-based Change Detection Content


slide-1
SLIDE 1

The ATLAS ELEKTRONIK Group/ 1

… a sound decsion … a sound decision

Image-based change detection to reduce false alarms in the Vision1200 synthetic aperture sonar

  • Dr. C. Erdmann and Dr. J. Groen
slide-2
SLIDE 2

The ATLAS ELEKTRONIK Group/ 2

  • Introduction
  • Data
  • Data preprocessing
  • SAS processing
  • Normalization and filtering
  • Registration
  • Coarse registration
  • Fine registration (coherent, incoherent)
  • Performance analysis
  • Detectors
  • Results
  • Receiver operating characteristics (ROC)
  • Robustness analysis
  • Summary

Image-based Change Detection Content

slide-3
SLIDE 3

The ATLAS ELEKTRONIK Group/ 3

Image-based Change Detection Basic processing chain

Preprocessing Coarse Registration Preprocessing Subtraction Detection Fine Registration

t1 t2

slide-4
SLIDE 4

The ATLAS ELEKTRONIK Group/ 4

  • ITMINEX NATO Trial 2014
  • Study commissioned by WTD 71
  • Provision of RV „Alliance“ and

trial organization by CMRE

  • 3 identical missions
  • 2 different sets of 7 objects
  • 34 usable legs with total of 116 MLO images
  • Sea Otter AUV
  • ATLAS ELEKTRONIK UK „Vision MK1 1200“ SAS System

Image-based Change Detection Survey

slide-5
SLIDE 5

The ATLAS ELEKTRONIK Group/ 5

Image-based Change Detection Data: typical example

slide-6
SLIDE 6

The ATLAS ELEKTRONIK Group/ 6

  • ATLAS ELEKTRONIK SAS processing chain
  • Artificial defocusing by sway data distortion

Data processing SAS processing

slide-7
SLIDE 7

The ATLAS ELEKTRONIK Group/ 7

Data Processing Normalization and Filtering

  • Normalization
  • Based on along-track mean
  • Based on roll data (eliminate roll effect)
  • Based on combined along-track and range median
  • Filtering
  • No filtering
  • Lee-filter: speckle-reducing
  • Anisotropic diffusion filter: edge-preserving
slide-8
SLIDE 8

The ATLAS ELEKTRONIK Group/ 8

Image Registration Coarse registration

  • Rigid registration
  • Maximize correlation coefficient of whole image
  • Rotation correction
  • Δx, Δy: 2cm, Δϕ: 0.1°
slide-9
SLIDE 9

The ATLAS ELEKTRONIK Group/ 9

Image Registration Fine Registration

slide-10
SLIDE 10

The ATLAS ELEKTRONIK Group/ 10

Image Registration Coherent Fine Registration

Δt = 3 h

slide-11
SLIDE 11

The ATLAS ELEKTRONIK Group/ 11

Image Registration Coherent Fine Registration

slide-12
SLIDE 12

The ATLAS ELEKTRONIK Group/ 12

Image Registration Coherent vs. Incoherent Fine Registration

slide-13
SLIDE 13

The ATLAS ELEKTRONIK Group/ 13

Image Preparation Subtraction

slide-14
SLIDE 14

The ATLAS ELEKTRONIK Group/ 14

Image Preparation Subtraction

slide-15
SLIDE 15

The ATLAS ELEKTRONIK Group/ 15

Image Preparation Subtraction

slide-16
SLIDE 16

The ATLAS ELEKTRONIK Group/ 16

Example 2

slide-17
SLIDE 17

The ATLAS ELEKTRONIK Group/ 17

Example 2

slide-18
SLIDE 18

The ATLAS ELEKTRONIK Group/ 18

Example 2

slide-19
SLIDE 19

The ATLAS ELEKTRONIK Group/ 19

Performance Analysis Performance Analysis: Overall Image Contrast

Coherent, 32x32 px µ σ σ Blue: Δt = 26 h Red: Δt = 56 h

slide-20
SLIDE 20

The ATLAS ELEKTRONIK Group/ 20

Performance Analysis Performance Analysis: Overall Image Contrast

Incoherent, 64x64 px µ σ σ Blue: Δt = 26 h Red: Δt = 56 h

slide-21
SLIDE 21

The ATLAS ELEKTRONIK Group/ 21

Performance Analysis Performance Analysis: Overall Image Contrast

µ σ σ Incoherent, 512x512 px Blue: Δt = 26 h Red: Δt = 56 h

slide-22
SLIDE 22

The ATLAS ELEKTRONIK Group/ 22

Detectors ROC curves

Two simple detectors (single score for comparability)

  • 1. Variance detector

– Threshold in difference image variance

  • 2. Template matching detector

– Template: mean of all MLO signatures

slide-23
SLIDE 23

The ATLAS ELEKTRONIK Group/ 23

Results Tested Combinations

Normalization Filter Detector RRn Range-Roll-normalization ADf Anisotropic Diffusion Filter VARd Variance detector SASn Median-based normalization LEEf Lee-Filter TMd Template matching detector Rn Range normalization NOf No Filter

slide-24
SLIDE 24

The ATLAS ELEKTRONIK Group/ 24

ROC Curves No Change Detection

Template Matching Detector Variance Detector

slide-25
SLIDE 25

The ATLAS ELEKTRONIK Group/ 25

ROC Curves Incoherent Change Detection

slide-26
SLIDE 26

The ATLAS ELEKTRONIK Group/ 26

ROC Curves Coherent Change Detection

slide-27
SLIDE 27

The ATLAS ELEKTRONIK Group/ 27

ROC Curves Robustness: Best Change Detection (Incoherent)

slide-28
SLIDE 28

The ATLAS ELEKTRONIK Group/ 28

ROC Curves Robustness: 0.5λ

slide-29
SLIDE 29

The ATLAS ELEKTRONIK Group/ 29

ROC Curves Robustness: 0.75λ

slide-30
SLIDE 30

The ATLAS ELEKTRONIK Group/ 30

ROC Curves Robustness: 1.5λ

slide-31
SLIDE 31

The ATLAS ELEKTRONIK Group/ 31

Results Summary

No CD CCD ICD ICD-DPCA ½λ ICD-DPCA ¾λ ICD-DPCA 1½λ TM 90% 6200 720 650 1100 1800 8700 TM 95% 14000 1100 780 1300 2400 12000 Var 90% 47000 5800 1600 2700 4200 11000 Var 95% 76000 10000 2000 3500 5600 18000

slide-32
SLIDE 32

The ATLAS ELEKTRONIK Group/ 32

  • Change detection enhances detection performance by factor 10 to 40 as compared to „No CD“.

Results Summary

No CD CCD ICD ICD-DPCA ½λ ICD-DPCA ¾λ ICD-DPCA 1½λ TM 90% 6200 720 650 1100 1800 8700 TM 95% 14000 1100 780 1300 2400 12000 Var 90% 47000 5800 1600 2700 4200 11000 Var 95% 76000 10000 2000 3500 5600 18000

slide-33
SLIDE 33

The ATLAS ELEKTRONIK Group/ 33

  • Change detection enhances detection performance by factor 10 to 40 as compared to „No CD“.
  • Incoherent change detection slightly outperforms coherent change detection.

Results Summary

No CD CCD ICD ICD-DPCA ½λ ICD-DPCA ¾λ ICD-DPCA 1½λ TM 90% 6200 720 650 1100 1800 8700 TM 95% 14000 1100 780 1300 2400 12000 Var 90% 47000 5800 1600 2700 4200 11000 Var 95% 76000 10000 2000 3500 5600 18000

slide-34
SLIDE 34

The ATLAS ELEKTRONIK Group/ 34

  • Change detection enhances detection performance by factor 10 to 40 as compared to „No CD“.
  • Incoherent change detection slightly outperforms coherent change detection.
  • The different normalization schemes and filters have a noticeable impact on performance. The median-based

normalization method without filtering performs best on well focused imagery. Lee-filtering becomes beneficial when dealing with defocused SAS imagery.

Results Summary

No CD CCD ICD ICD-DPCA ½λ ICD-DPCA ¾λ ICD-DPCA 1½λ TM 90% 6200 720 650 1100 1800 8700 TM 95% 14000 1100 780 1300 2400 12000 Var 90% 47000 5800 1600 2700 4200 11000 Var 95% 76000 10000 2000 3500 5600 18000

slide-35
SLIDE 35

The ATLAS ELEKTRONIK Group/ 35

  • Change detection enhances detection performance by factor 10 to 40 as compared to „No CD“.
  • Incoherent change detection slightly outperforms coherent change detection.
  • The different normalization schemes and filters have a noticeable impact on performance. The median-based

normalization method without filtering performs best on well focused imagery. Lee-filtering becomes beneficial when dealing with defocused SAS imagery.

  • Future work aims at connecting change detection to the automatic target recognition (ATR) for which the target shadow

needs be treated such that its information is preserved.

Results Summary

No CD CCD ICD ICD-DPCA ½λ ICD-DPCA ¾λ ICD-DPCA 1½λ TM 90% 6200 720 650 1100 1800 8700 TM 95% 14000 1100 780 1300 2400 12000 Var 90% 47000 5800 1600 2700 4200 11000 Var 95% 76000 10000 2000 3500 5600 18000

slide-36
SLIDE 36

The ATLAS ELEKTRONIK Group/ 36

ATLAS ELEKTRONIK GmbH Sebaldsbruecker Heerstrasse 235 28309 Bremen | Germany Phone: +49 421 457-02 Telefax: +49 421 457-3699 www.atlas-elektronik.com

Contact … a sound decision