PolSAR Cell Information Representation by a Pair of Elementary Scatterers

Abstract

This study exploits Cameron’s decomposition for polarimetric data analysis and presents an information extraction process so that each PolSAR cell (pixel) is interpreted by two dominating elementary scattering mechanisms each one contributing to the scattering behavior of the SAR pixel with its own weight. The co-distance matrix is introduced to depict the metric distances between these two nearest scattering mechanisms. For most of the scattering mechanisms in each resolution cell, the strength between the first and the second nearest elementary scatterer usually differs slightly. This indicates that the interpretation of the available information in a PolSAR pixel by a single dominant scatterer, as most methods employ, is not adequate. The proposed method presents an alternative to Cameron’s spherical topology by taking advantage of the elementary scattering mechanisms complementary nature and inspired by the philosophy of principal component analysis. According to the introduced topology four elementary scatterers, which are in pairs complementary to each other, are adequate to characterize each PolSAR pixel. The aim of this research is to present a new feature-toοl with a more stochastic nature that could fit a variety of techniques that utilize fully polarimetric data. To prove the strength of the proposed method, the double-scatterer model is applied for interpreting each pixel on a variety of land cover types presenting a richer feature extraction capability, effective in detection and classification procedures.