POLSAR IMAGE CLASSIFICATION BASED-ON SEMI-SUPERVISED POLARIMETRIC FEATURE SELECTION

Abundant polarimetirc features have been used for PolSAR image classification. However, it is almost impossible to utilize all polarimetric features for classification, which may result in an unsatisfactory classification performance. Moreover, labeling samples in PolSAR images is costly and laborious. Therefore, we propose a semi-supervised polarimetric feature selection method for PolSAR image classification, which can utilize a small number of labeled samples and plentiful unlabeled samples. Specifically, the discriminative information of labeled samples is maintained by maximum margin criterion (MMC) based on the within-class scatter matrix and between-class scatter matrix. The manifold regularization is used to preserve the local structure of all samples as well as the label information. Moreover, the $l_{2,1}$ norm sparsity regularization is added for feature selection. Experimental results show that the proposed method can improve the classification performance comparatively.