Experimental reaserch of unsupervised Cameron/ML Classification method for fully polarimetric SAR Data

Fully PolSAR data provided by the NASA/JPL laboratory are widely used to classify PolSAR image. In this paper, an unsupervised Cameron/ML approach is proposed to classify airborne fully polarimetric data collected by a research institute in China. Cameron’s method is used to initially classify the PolSAR image firstly. Secondly the initial classification map defines training sets for the maximum likelihood (ML) classifier. The classified results are then used to define training sets for the next iteration. The advantages of this method are the automated classification, and the interpretation of each class based on scattering mechanism. Formula of Cameron classification for the very measured data is also obtained here. The experiment demonstrates the proposed approach dramatically improves the classification result compared with the Cameron method.