Supervised classification of PolSAR images using adaptive sample censoring strategy

In this paper, a supervised classification algorithm of polarimetric synthetic aperture radar (PolSAR) images based on SVM with training sample optimization is proposed. For the supervised strategy, three main steps are there including feature extraction, design of classifier and training of classifier. Firstly, some features of the PolSAR images are extracted by employing the target decomposition theories. Secondly, a strategy for optimizing training sample is applied on each of preliminary selected sample sets, of which an adaptive threshold is determined according to the statistics of the characteristics similarity. Finally, classification of a real PolSAR image is performed based on SVM with the refined training sample sets. The experimental results have proven that the proposed algorithm is able to improve the quality of the samples and also the classification accuracy and the robustness of the classification algorithm for PolSAR images.