Role of polarimetric indices based on statistical measures to identify various land cover classes in ALOS PALSAR data

The present paper deals with the potential application of fully polarimetric data in identification of various land cover types based on polarimetric indices, namely, backscattering coefficients and their ratios of various polarizations (linear, circular, linear 45°), entropy, weighted polarimetric sum, correlation coefficient, normalized difference polarization index, and ratio vegetation index. In order to make decision boundaries for separating land cover classes two statistical measures, namely, median and standard deviation were critically analyzed for each polarimetric parameter. In order to select the appropriate value of all polarimetric indices for each class, a relationship was obtained between that particular polarimetric index, and their median and standard deviation along with some integer, the value of which was chosen to optimize the proposed method. The proposed method was successfully applied on ALOS PALSAR data for which satisfactory results were obtained.