PolSAR Data Classification via Combined Similarity Based Immune Clonal Spectral Clustering

Abstract

Traditional spectral clustering (SC) employed k-means to find the cluster centers, which leads to the problem of sensitive to initialization and easily falls into local optimum. To address this issue, a novel superpixel-based immune clonal spectral clustering (ICSC) method in the spatial-polarimetric domain is proposed for PolSAR data classification. Firstly, the proposed method divides PolSAR image into superpixels, which not only considers the region homogeneity but also reduces the computational complexity. After that, combined manifold distance measures in the spatial-polarimetric domain are used to construct the similarity matrix. Finally, immune clonal algorithm (ICA) is substituted for k-means to obtain global optimum solution with large probability. Experiments results show the feasibility and efficiency of the proposed method.