Ocean Clutter Characterization Based on PolSAR Data and Second-Order Statistics of Elementary Scatterers

Abstract

New features are proposed for sea clutter characterization when PolSAR data are employed. Cameron coherent target decomposition is applied to characterize each SAR pixel by means of the 8 basic elementary scatterers obtained by this decomposition. Since the examined SAR pixel does not match exactly to a specific ideal elementary scatterer, the closest scatterer is assigned to this pixel along with its distance (closeness) from the ideal one. The features proposed are (a) the percentage of each elementary scatterer in the investigated region; (b) the distribution of the closeness of each declared elementary scattering mechanism to its ideal counterpart; (c) the co-occurrence of the same scattering mechanism, taking into consideration its distance from the ideal one; and (d) The co-occurrence of the different scattering mechanisms in pairs, taking again into consideration their distance from the ideal ones. Simultaneously, the skewness and the kurtosis and their 2D versions of the previously mentioned probabilities are employed to further improve ocean clutter characterization. The above features correspond to the sea state condition in each separate region, i.e., the wave height and the wind speed. A clear correspondence between the proposed features and the sea state conditions is established. Data were available from RADARSAT-2 and ALOS-1 PALSAR systems for four different ocean regions on Earth.