1 / 31

University of British Columbia September, 2007

Applying Target Decomposition Algorithms on the Detection of Man Made Targets Using Polarimetric SAR. Flavio Wasniewski*, Ian Cumming. University of British Columbia September, 2007. Objectives. Review the Detection of Crashed Airplanes (DCA) methodology applied by Lukowski et. al .

libra
Download Presentation

University of British Columbia September, 2007

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Applying Target Decomposition Algorithms on the Detection of Man Made Targets Using Polarimetric SAR Flavio Wasniewski*, Ian Cumming University of British Columbia September, 2007

  2. Objectives • Review the Detection of Crashed Airplanes (DCA) methodology applied by Lukowski et. al. • Test this methodology with a more diverse set of target clutters and types; • Compare its performance with available target detection algorithms; • Develop improvements to the methodology in order to give good detection performance to a range of target and clutter types. 2

  3. Detection of Man Made Targets with Radar Polarimetry High target-to-clutter ratio (not necessarily higher than in natural targets) Dihedral scattering expected (phase information can be explored) Polarimetric decompositions are among the most promising algorithms Most civilian operational applications focus in ship detection 3

  4. Detection of Crashed Airplanes (DCA) Promising in-land application Tested on airplanes and low vegetation clutter Tail and wings usually remain intact and provide dihedrals Can it be applied to all discrete man made targets? (will dihedrals always be present?) Source: Lukowski et. al., CJRS, 2004 4

  5. Methodology 1 (DCA) The cross symbol is a logical “and” combining the 3 results. 5

  6. Methodology 2 6

  7. Methodology 3 7

  8. Methodology 4 8

  9. Algorithms (1/5) – Polarimetric Whitening Filter Bright pixels represent strong radar returns, but targets are obscured; PWF reduces speckle (σ/µ) without affecting the resolution; Target-to-clutter ratio is improved 9

  10. Algorithms (2/5) – Even Bounce Analysis Explores the 180° phase shift between HH and VV 10

  11. Algorithms (3/5) – Cameron Decomposition Classifies the target according to the maximum symmetric component in one of six elemental scatterers. Source: Cameron, 1996 11

  12. Algorithms (4/5) – Freeman-Durden Decomposition Decomposition of backscatter into three basic scattering mechanisms: • Volume scattering: canopy scatter from a cloud of randomly oriented dipoles • Double-bounce: scattering from a dihedral • Surface scattering: Single bounce from a moderately rough surface Source: Freeman et. al. 12

  13. Algorithms (5/5) – Coherence Test Detects coherent targets based on the degree of coherence and target-to-clutter ratio. Degree of coherence a and b are the Pauli components 13

  14. Morphological processing • Closing (dilation + erosion) • Clustering • Erasing 1 and 2-pixel detections 14

  15. Experiments: data sets used (1) Gagetown dataset 15

  16. Experiments: data sets used (2) Westham Island dataset 16

  17. Results – Target 21 (House Among Trees) CV-580 data Target and clutter (Ikonos image) 17

  18. Results – Target 21 – Methodology 1 PWF and Even Bounce 18

  19. Results – Target 21 – Methodology 1 Cameron combined to PWF and Even Bounce 19

  20. Results – Target 21 – Methodology 2 20

  21. Results – Target 21 – Methodology 3 Detection map after morphology 21

  22. Results – Target 21 – Methodology 4 Cameron + PWF + Even Bounce + Coherence Test Detection map 22

  23. Results – Target 2 (Plow) 23

  24. Results – Target 2 – Methodology 1 - Same detection results were achieved by Methodologies 2 and 4 24

  25. Results – Target 5 (Horizontal cylinders) Man made target with no dihedral behaviour No detections 25

  26. Results – Target 7 (House) 26

  27. Results – Target 20 (Crashed Plane in Grass) Target Corner reflectors 27

  28. Results – Target 20 -Methodology 1 - Same detection results were achieved by Methodologies 2 and 4 28

  29. Results Total Per Vegetation type 29

  30. Summary • Methodology 1 (DCA) detected the targets with no false alarms when clutter is low vegetation. It did present false alarms in high vegetation; • Methodology 2 (Coherence Test) typically detects the target with few false alarms in both situations; • Methodology 3 (Freeman-Durden decomposition) generally presented high false alarm rates in this study; • Methodology 4 (DCA + Coherence Test) performs better than DCA methodology on high vegetation clutter. 30

  31. Thank you

More Related