IGARSS’201124-29 July, Vancouver, Canada

Subsidence Monitoring Using Polarimetric Persistent Scatterers Interferometry

Victor D. Navarro-Sanchez and Juan M. Lopez-Sanchez

Signals, Systems and Telecommunications Group University of Alicante, Spain

IGARSS’201124-29 July, Vancouver, Canada

• Introduction

• PSI optimization

• Polarimetric behaviour

• Deformation results

• Conclusions and future work

Overview

IGARSS’201124-29 July, Vancouver, Canada

Introduction

Persistent Scatterers Interferometry (PSI)

• Used extensively to measure surface deformation evolution

• Exploits phase information of a stack of SAR differential interferograms

• Selects for processing only good quality points (pixel candidates)S

• Different quality criteria:

– Average coherence

– Amplitude dispersion index

– Others

• Traditionally formulated and applied to single-pol data

IGARSS’201124-29 July, Vancouver, Canada

Introduction

Concerning Polarimetry:

• Sensitive to physical features of the scene (orientation, shape...)

• Widely used for physical parameter retrieval (forest height, biomass, soil moisture, etc.)

• Polarimetric interferometry (PolInSAR) has a solid theoretical background

Motivation:

To find efficient ways to exploit polarimetric information in order to improve PSI performance:

- Increase of the quality and number of pixel candidates

- Extract additional information to classify scatterers

IGARSS’201124-29 July, Vancouver, Canada

PSI optimizationGeneral framework and formulation for vector interferometry

[Cloude and Papathanassiou, 1998]:

Scattering complex matrix (Sinclair Matrix):

Projection:

Target vector:

Scattering coefficient(scalar complex)(

Projection vector(normalized complex vector)(

IGARSS’201124-29 July, Vancouver, Canada

Formulation for vector interferometry adapted to the dual-pol case:

Projection vector parameterization:

Target vector:

Search

PSI optimization

Navarro-Sanchez et al., IEEE GRSL, 2010

IGARSS’201124-29 July, Vancouver, Canada

Average coherence optimization:

Average coherence:

Scalar interferometry:

Vector interferometry:

Constraint: ω1 = ω2 = ω ∀ k

PSI optimization

IGARSS’201124-29 July, Vancouver, Canada

Amplitude dispersion optimization:

Amplitude dispersion:

Scalar interferometry:

Vector interferometry:

Constraint: same ω ∀ i

Amplitude standard deviation

Average amplitude

PSI optimization

IGARSS’201124-29 July, Vancouver, Canada

Data set 40 images of Murcia (Spain) Feb-09 to May-10

• TerraSAR-X SLC data, stripmap mode

• Dual-pol: HH and VV

• Mean incidence angle: 37.8 degrees

• Resolution: 6.6m Az, 1.17m Rg

• Pixel spacing: 2.44m Az, 0.91m Rg

• Oversampling factors: 2.7 Az, 1.28 Rg

All images have been provided by DLR under the framework of project GEO0389

PSI optimization

IGARSS’201124-29 July, Vancouver, Canada

Histograms of γ and DA for conventional channels and computed optimum

PSI optimization

IGARSS’201124-29 July, Vancouver, Canada

Average coherence criterion, ML 7x7, threshold 0.7

VVPixels selected = 24.07%

OPTPixels selected = 38.57%

PSI optimization

IGARSS’201124-29 July, Vancouver, Canada

DA criterion, full-resolution, threshold 0.3

VVPixels selected = 6.01%

OPTPixels selected = 16.32%

PSI optimization

IGARSS’201124-29 July, Vancouver, Canada

Polarimetric Behaviour

Rural area

Urban area

IGARSS’201124-29 July, Vancouver, Canada

Polarimetric BehaviourAverage coherence criterion, ML 7x7, threshold 0.7

Interpretation of figures for the Pauli basisWhole scene Urban Rural

IGARSS’201124-29 July, Vancouver, Canada

Polarimetric BehaviourDA criterion, full-resolution, threshold 0.3

Interpretation of figures for the Pauli basisWhole scene Urban Rural

IGARSS’201124-29 July, Vancouver, Canada

Deformation results

• ML 7x7

•167 interferograms

• 3 selection layers:

γ > 0.9 - 0.8 - 0.7

IGARSS’201124-29 July, Vancouver, Canada

Deformation results

VV OPT

• New details are revealed

• Results are consistent with ground truth data and with those presented in [Herrera et al. 2010] (single pol, ML 3x3)

IGARSS’201124-29 July, Vancouver, Canada

Summary• PSI can be enhanced by exploiting polarimetric information:

– Average coherence: Improvement ≈ 60% more PCs selected

– Amplitude dispersion: Improvement ≈ 170% more PCs selected

• Polarimetric study of selected PSC:

– Multi-look data: Dominance of dihedral-like scatterers

– Single-look data: Dominance of anisotropic dihedrals and vertical structures

• Retrieved deformation maps are denser and consistent with ground truth, proving the suitability of the approach

• Future work: Study of selection criteria that make a more complete use of polarimetric information (i.e. polarimetric stationarity indicator)

• Future work: Quad-pol extension

IGARSS’201124-29 July, Vancouver, Canada

Subsidence Monitoring Using Polarimetric Persistent Scatterers Interferometry

Victor D. Navarro-Sanchez and Juan M. Lopez-Sanchez

Signals, Systems and Telecommunications Group University of Alicante, Spain