Experimental research of unsupervised Cameron/maximum-likelihood classification method for fully polarimetric synthetic aperture radar data

摘要

In this study, experimental research on classification is applied to fully polarimetric data in X-band from China. Considering the amplitude and phase error between H and V channels in the system, the authors firstly correct the error in original data. The authors also deduce the formula of Cameron¿s classification method for the real data in our study. Then Cameron¿s method is used to initially classify the site image. Finally, the initial classification map defines training sets for the maximum-likelihood (ML) classifier. The advantages of this method are the automated classification and interpretation of each class based on the scattering mechanism. The experiment demonstrates the feasibility of the proposed approach, which dramatically improves the X-band data classification result compared with the Cameron method and H/;1;/ML method.