

A General Modelbased Polarimetric Decomposition Scheme for Vegetated Areas
Maxim Neumann^{(1)}, Laurent FerroFamil^{(1)} and Eric Pottier^{(1)}^{(1)} University of Rennes 1, 263 Avenue General Leclerc, 35042 Rennes, France
Abstract
A simple vegetation model for polarimetric covariance and coherency matrix elements is presented. The model aims to represent vegetation characteristics which are observable by radar polarimetry, including the average particle anisotropy, the main orientation of the volume, the degree of orientation randomness in the volume, and the terrain slopes. The decomposition consists, in analogy to the FreemanDurden model, of volume, surface, and doublebounce scattering components considering all vegetation characteristics. The goal of this approach is to quantify these parameters and to enable their estimation in a remote sensing parameter inversion framework.
The vegetation particles are characterized by the average shape, size and dielectric constant, represented by the effective anisotropy. It is usually assumed that vegetated areas exhibit reflection symmetry about the incidence plane. This is often not the case in the presence of oriented volume elements and terrain slopes. Often, given fully polarimetric data, one can rotate the incidence plane around the line of sight (LOS) to achieve reflection symmetry. However, the presence of an azimuth slope of the terrain can introduce helixlike scattering components which undermine reflection symmetry.
We derive the general expressions to quantify the degree of orientation randomness and the degree of helicity, characterized by the probability density function of orientation angles under the central limit theorem for the vegetation and the surface. After examination of these effects, we present all zeroth and firstorder coherency and covariance matrix elements of a single vegetation layer over the ground. Most agricultural and forestry vegetation types can be realistically modeled by a single layer taking into account the orientation characteristics, but an extension to multiple layers to model complex vegetation structures is possible.
Using additional data sources, such as interferometry (PolInSAR), external DEM, multifrequency, or multiple incidence angles makes the inversion of the mentioned parameters possible. Under some assumptions, one can invert these parameters based on SAR polarimetry only, as will be presented. Experimental validation is presented on simulated data as well as on real SAR data acquired by the ESAR system of the German Aerospace Center (DLR).
This work is strongly related to the physical model and eigenvector based decomposition approaches by Freeman and Durden (1998), Freeman (2007), Yamaguchi et al. (2005, 2006), Cloude and Pottier (1996, 1997), Cloude et al. (1999), Lee et al. (2000, 2002), and Schuler et al. (2002).
Workshop presentation
Full paper

