Unsupervised Classification of Fully Polarimetric SAR Image Based on Deorientation Theory

The randomly distributed target orientation causes confusion in classification of the polarimetric scattering target.In this paper,an unsupervised classification method of the fully polarimetric SAR image is proposed to solve this problem.Firstly,a certain angle rotation of the part along the sight line is made to minimize the cross-polarized scattering;then an initial classification using the scattering mechanism is implanted,which is described by the parameters u/v/H.Finally,using the initial classification result,the polarimetric SAR image is classified by the modified C-mean algorithm.Experiment is performed on the real measured data collected by NASA/JPL laboratory.Result shows the proposed method is efficient.