Compact Polarimetric SAR Interferometry: PALSAR Observations and Associated Reconstruction Algorithms

Marco Lavalle, Eric Pottier, Domenico Solimini, and Yves-Louis Desnos

Abstract

Polarimetric SAR interferometry (PolInSAR) has opened the path to promising applications, especially in the field of forest remote sensing, such as tree height estimation and multilayer topographic information retrieval. However, full polarimetry (FP) implies some system constraints to cope with data downloading rate, size of the processed swath and power consuming. To overcome these limiting factors, a new approach for SAR polarimetry, called compact polarimetry (CP), has been proposed. In the CP architecture, the SAR transmits either a circular polarized wave (pi/2 mode) or a linear polarized one oriented at 45 degrees (pi/4 mode), and receives at horizontal and vertical polarization. A theoretical formulation of CP, a reconstruction algorithm of the pseudo full polarimetry information and the associated performance on PolSAR classification have been presented in [J.-C. Souyris et al, IEEE TGARS, 2005].

In this contribution, we first discuss the synthesis of a compact polarimetry dataset from full polarimetry data. In previous works, the synthesis of CP data has been performed using FP single-look complex (SLC) data, i.e. after the focusing of the raw data. On the basis of the basic principles of electromagnetic theory, we argue that the synthesis of compact polarimetry is more appropriate on raw data, i.e. before the focusing process. We show the ALOS-PALSAR observations of the CP modes following the two alternative approaches, using the ESA ALOS-PALSAR Prototype Processor. Moreover, we compare the performance of the pi/4 mode and pi/2 mode in terms of H/A/alpha decomposition after pseudo full-pol reconstruction.

Secondly, we consider the interferometric scenario. The compact PolInSAR (C-PolInSAR) formulation is addressed and a FP reconstruction algorithm of the PolInSAR covariance matrix is presented. The reconstruction of the full PolInSAR (F-PolInSAR) covariance matrix is based on the symmetry properties of some natural media. By assuming rotation symmetry or reflection symmetry plus an additional constraint, some PolInSAR characteristics of the vegetation result preserved. We show the performance of our reconstruction algorithm of the pseudo full PolInSAR information. Our analysis uses both a coherent PolInSAR simulator, that generates a fully polarimetric dataset of different forest scenarios, and real PALSAR data acquired over forested area. The comparison between C-PolInSAR and F-PolInSAR is based on the interferometric degree of coherence in the complex plane (coherence region) and the forest height retrieval algorithm. The comparison shows a general good agreement between the two configurations

Index Terms—PolInSAR. Interferometry. Compact polarimetry. Forest. ALOS. PalSAR

POLinSAR 2009, the 4th International Workshop on Science and Applications of SAR Polarimetry and Polarimetric Interferometry, ESA/ESRIN, Frascati, Italy, 26-30 January 2009.

© Authors, 2009