Unsupervised Classification of Polarimetric SAR Images Based on Scattering Mechanism

In this paper, an unsupervised classification method is proposed for polarimetric synthetic aperture radar (PolSAR) images. The proposed method combines the Freeman decomposition and Cloude decomposition to improve the initial classification take advantage of the scattering mechanism. Decomposition coefficients of three scattering mechanism components are extracted through Freeman decomposition first; then the scattering entropy and angle are calculated by Cloude decomposition. Employing the aforementioned five parameters, the initial classification plane is redefined and divided based on the dominant scattering mechanism. Finally, a clustering algorithm is adopted to the iteration of the initial classification, which optimizes the boundaries between different terrains. The experiment results demonstrate the superiority and effectiveness of the proposed method.