Polarimetric Interferometric SAR Data Analysis Based on ESPRIT/MUSIC Methods
University of Rennes 1,
(2) University of Rennes, Universitť de Rennes 1 - UFR SPM, 35042 Rennes Cedex, France
(3) Technische Universitšt Berlin, Stasse des 17. Juni 135, EB9, 10623 Berlin, Germany
Interferometric SAR provides a two-dimensional image of elevation angle related to the scatterer height. By construction, SAR imaging is a projection of a volume response onto a plane. Thus, scattering points are distributed over a two-dimensional surface. The retrieval of the scatterer height assumes that only one scattering mechanism is present in each resolution cell. Unfortunately, this assumption is invalid in the sense that multiple scatterers, with distinct elevation angles, arise in a single resolution cell. This effect introduces artefacts during the projection under the form of a phase centre bias. To improve the interferometric phase estimation, it is thus necessary to discriminate the different scattering mechanisms arising over man-made or volumetric targets, like urban area or forest, but also to filter artefacts related to the SAR device, like noise. Recently, many studies have been proposed to estimate the interferometric phase over forest areas. One of them is based on the ESPRIT (Estimation of Signal Parameters via Rotational Invariance Techniques) algorithm, often employed for Direction-Of-Arrival estimation using antenna arrays. For volume areas, this algorithm can directly retrieve the interferometric phases of the ground and the canopy. Man-made targets are constituted of different kinds of scatterer and the interferometric phase estimation becomes complex. This paper addresses a polarimetric approach based on ESPRIT/MUSIC algorithm, using multibaseline fully polarimetric SAR images. A polarimetric interferometric speckle using segmented data is presented. This allows to filter cells showing same behaviour. The different scattering mechanisms may be separated using ESPRIT/MUSIC algorithms, A polarimetric technique based on an eigenvalue spectral analysis is applied. This principle, used over forested area, is extended over man-made targets in order to separate different kinds of scatterers. It allows the estimation of the polarisation states of the dominant scatterers, in H and V polarisations, based on Jones vector polarisation ratio retrieval. Finally, this paper introduces a multibaseline analysis using a scattering model. The efficiency of this polarimetric interferometric SAR data analysis is demonstrated using fully polarimetric multibaseline SAR images, obtained from DLR-ESAR airborne sensor in L-band repeat-pass mode.