Feature extraction and classification of PolSAR images based on sparse representation

In this paper, a supervised PolSAR images classification method based on sparse representation is proposed. Firstly, Polarimetric decomposition based on Multiple-component Scattering Model, Cloude-Pottier decomposition and Gray-level Co-occurrence Matrix are implemented to obtain features which can describe PolSAR images at multiple aspects. Then, the training samples are represented in an overcomplete dictionary in which the basic elements are the feature vectors of the training samples and this can be computed by minimize l¹-norm. Through comparing the residuals of the reconstructed training samples corresponding to different classes with the original training sample, the optimization problem can be solved and obtain the classification result. The proposed method is validated by the Danish EMISAR L-band fully polarimetric SAR data of Foulum Area (DK) and the preliminary experimental results confirm the performance and potential of the proposed method in PolSAR image interpretation.