Landslide Identification Based on Hierarchical Fuzzy Contour Model Clustering Algorithm Using Polsar Images

In order to accurately extract landslide disaster information from SAR images, we propose a clustering algorithm for landslide detection using multi-temporal fully polarimetric SAR image. By combining Freeman-Durden decomposition with H /α / A decomposition method, this algorithm makes full use of scattering power and scattering entropy information. Furthermore, a hierarchical method is proposed to overcome the drawback that fuzzy membership function family of multi-region classification cannot always satisfy the constraint conditions to guarantee functional convexity. Such method ensures that the constraint conditions can always be satisfied and improves the accuracy of classification. Finally, the landslide disaster area is accurately extracted from the two classification maps by change detection. The results of the C-band RadarSat-2 data are provided to demonstrate its competitive general performance of image classification.