Modelbased Polarimetric Decomposition using Pol In SAR Coherence
Model-based Polarimetric Decomposition using Pol. In. SAR Coherence Si-Wei Chen, Motoyuki Sato Tohoku University, Japan chensw@cneas. tohoku. ac. jp sato@cneas. tohoku. ac. jp
Outline l Introduction – Current model-based decompositions – Limitations l Pol. In. SAR Coherence – Estimation and optimization l Proposed Decomposition – Adaptive volume scattering model l Comparative experiments l Conclusions 2
Introduction Ø Polarimetric SAR (Pol. SAR) – Full polarimetric information – Covariance matrix ALOS/PALSAR East Japan earthquake and tsunami After-event Pre-event Optical Image Ø Model-based decomposition Decomposition Pd – Better understanding the scattering mechanisms Ps = � Double Bounce Volume Scattering � Pv �. . . Color-code Single Bounce 3 (Pol. SARpro tutorials)
Model-based decomposition ØFreeman-Durden decomposition (1998) Double Bounce Volume Scattering Ø Limitations Single Bounce Ø Improvements Reflection symmetry assumption Helix component Negative power Nonnegative eigenvalues Scattering mechanism ambiguity Deorientation Inadaptive General volume model (A. Freeman, Y. Yamaguchi, W. M. Boerner, J. J. Van Zyl, J. S. Lee, Y. Q. Jin, M. Neumann, M. Arii and et al. ) 4
Scattering mechanism ambiguity Ø General representation of the volume scattering model Skew-oriented building Decomposed volume scattering power – For Freeman-Durden – For Yamaguchi Pauli Image Table I Averaged Backscattered Power (In d. B) After Deorientation 19. 48 13. 52 16. 75 12. 79 18. 81 18. 53 17. 56 5
Possible reasons and countermeasures Ø Possible reasons Double Bounce Volume Scattering Only Volume scattering Terrain slopes, oblique buildings Single Bounce Cross. Polarization Ø Countermeasures ü Adaptive volume scattering model ü Indirect modification of double- and single-bounce models ü Balance the inputs and outputs Utilization of both Polarimetric and Interferometric information! 6
Outline l Introduction – Current model-based decompositions – Limitations l Pol. In. SAR Coherence – Estimation and optimization l Proposed Decomposition – Adaptive volume scattering model l Comparative experiments l Conclusions 7
Pol. In. SAR coherence Ø Polarimetric SAR interferometry (Pol. In. SAR) – Combination of Pol. SAR and In. SAR – Covariance matrix Ø Coherence magnitude Pol. In. SAR (T. Xiong) Polarimetric Dependence Ø Optimization 8
Pol. In. SAR coherence Ø Decorrelation sources – Signal-to-noise decorrelation – Baseline decorrelation – Processing decorrelation – Temporal decorrelation – Volume decorrelation – … NOTE: For manmade target For forest (K. Papathanassiou et al. ) Pol. In. SAR coherence: ü Sensitive to diverse terrains ü Close relationship to forest structures Potentially, the volume scattering can be modeled from it! 9
Pol. In. SAR Coherence Optical image Optimal 2 Optimal 1 Optimal 3 10
Pol. In. SAR Coherence Optimal 1 Optimal 2 Optimal 3 11
Outline l Introduction – Current model-based decompositions – Limitations l Pol. In. SAR Coherence – Estimation and optimization l Proposed Decomposition – Adaptive volume scattering model l Comparative experiments l Conclusions 12
Proposed decomposition Ø Adaptive volume scattering model (A. Freeman, 2007) Where, Modeled with Pol. In. SAR coherence is adjust to the spatial and temporal baseline parameters. Ø Model compatibility Use Freeman-Durden model If 13
Model Parameters Two unknowns: Principles for the choice of are: – More uniform distribution – More sensitive for diverse terrains 14
Model Parameters NOTE: For Indirect modification of double and single bounce scattering models! 15
Decomposition Flowchart Ø Double & single bounce model – Indirect modification Ø Adaptive decomposition – Pixel by pixel –. Volume scattering Double bounce Single bounce 16
Outline l Introduction – Current model-based decompositions – Limitations l Pol. In. SAR Coherence – Estimation and optimization l Proposed Decomposition – Adaptive volume scattering model l Comparative experiments l Conclusions 17
Experiment-I Ø E-SAR Pol. In. SAR data – Test site: Oberpfaffenhofen, Germany – L-band – Data size : 1300× 1200 E-SAR (Pol. SARpro tutorials) Azimuth Range Optical image Master track Pauli image HH-VV, HH+VV Pol. In. SAR coherence RGB image HH, HV, VV 18
Pd Decomposition _ After deorientation Ps Ø Full scene Freeman-Durden Pv Color-code Yamaguchi Proposed Ø Forest region Freeman-Durden 19
Ø Skew-oriented built-up region Freeman-Durden Yamaguchi Proposed 20
Experiment - II ØE-SAR Pol. In. SAR data Bio. SAR-2008 campaign Spatial Baseline: 30 m Data size : 1496× 840 L band Repeat-pass dataset Temporal baseline: 110 min Azimuth Range Logged after the Bio. SAR 2008 Mar. 2008 Optical Image Jan. 2009 Oct. 2008 Coherence RGB Image Pauli Image HH-VV, HH+VV VV, HH 21
Pd Decomposition _ After deorientation Ps Pv Color-code Optical Image Freeman-Durden Yamaguchi More sensitive and better fit for diverse forest terrains! Proposed 22
Outline l Introduction – Current model-based decompositions – Limitations l Pol. In. SAR Coherence – Estimation and optimization l Proposed Decomposition – Adaptive volume scattering model l Comparative experiments l Conclusions 23
Conclusions Ø Adaptive volume scattering model – Using Pol. In. SAR coherence – Better fit for different terrains – Indirect modification of double- and single-bounce scattering models Ø Adaptive decomposition – Fully usage of the information – Successfully discriminate the skew-oriented buildings as manmade structures – Overcome the scattering mechanism ambiguity – Sensitive to diverse forest terrains 24
Limitation of current model Ø General representation of the volume scattering model Skew-oriented building Decomposed volume scattering power – For Freeman-Durden – For Yamaguchi Table I Averaged Backscattering Power (In d. B) Pauli Image Before Deorientation 19. 48 14. 10 15. 26 14. 69 20. 71 20. 43 19. 47 After Deorientation 19. 48 13. 52 16. 75 12. 79 18. 81 18. 53 17. 5626
Pol. In. SAR Coherence Optical image VV-VV HH-HH HV-HV 27
Pol. In. SAR Coherence HH-HH VV-VV HV-HV 28
Model Parameters Optical Image Optimal 3 Coherence 29
Ø Skew-oriented built-up region Freeman-Durden Yamaguchi Proposed Ø Scattering power Table II Scattering Power Contribution (%) Method Built-up area Freeman-Durden 20 Yamaguchi 22 Proposed 29 32 25 8 48 53 63 Forest area 9 7 13 88 82 81 3 11 6 30
Ø Skew-oriented built-up region Freeman-Durden Yamaguchi Proposed 31
Ø Skew-oriented built-up region Freeman-Durden Yamaguchi Proposed 32
Ø Skew-oriented built-up region Freeman-Durden Yamaguchi Proposed 33
Pol. In. SAR Coherence Optimal 1 Optimal 2 Optimal 3
Pol. In. SAR Coherence Optimal 1 Optimal 2 Optimal 3
Decomposition _ Volume scattering contribution Optical Image Freeman-Durden Yamaguchi Proposed
ALOS/PALSAR datasets ALOS/PALSAR Range Pauli Image Azimuth Optical Image Spatial baseline: 299 m Temporal baseline: 46 days 2007 -4 -02 2007 -05 -1837
Pol. In. SAR coherence _ H-V HH-HH VV-VV HV-HV 38
Pol. In. SAR coherence _ Optimal Opt 1 Opt 2 Opt 339
Pol. In. SAR coherence _ Histogram 40
Decomposition _ After deorientation Pd Ps Pv Color-code Freeman-Durden Yamaguchi Proposed 41
Pd Built-up region - I Ps Pv Color-code Optical image Yamaguchi Freeman-Durden Proposed 42
Pd Built-up region - II Ps Pv Color-code Optical image Yamaguchi Freeman-Durden Proposed 43
- Slides: 43