Unsupervised classification of multi-frequency and fully polarimetric SAR images based on the H/A/Alpha-Wishart classifier

Abstract

The authors introduce a new classification scheme for dual frequency polarimetric SAR data sets, based on the analysis of a (6/spl times/6) polarimetric coherency matrix defined from the polarimetric information of both images. From the Wishart probability density function of this polarimetric representation, is defined a maximum likelihood decision rule to perform an iterative adaptive classification. Another classification scheme, based on the derivation of the conditional probability of the cross-correlation between both data sets, is presented.