A Novel Unsupervised Iteration Classification Method For Fully Polarimetric SAR Image

Based on the theory of eigen-decomposition of fully polarimetric synthetic aperture radar (SAR) and scatter criterion clustering method, an unsupervised iteration classification method is proposed. It is simple, flexible and with high precision. Several schemes of the iteration classification method are given in this paper. The characteristic of each scheme is carefully analyzed according to the properties of eigen-decomposition and scatter criterion, and each scheme holds its own merit. Experiments are done on SIR-C/X-SAR data near Tien Mountains and perfect classification results are obtained. More importantly, this method is robust and has high adaptability.