Persistent Scatterer Clustering for Structure Displacement Analysis based on Phase Correlation Kernel

Abstract:
This paper proposes a persistent scatterer clustering method using kernel K-means algorithm for structure displacement analysis. Persistent scatterer interferometry, which monitors displacement of structures such as buildings by analyzing pixels named persistent scatterers (PSs), has difficulty in identifying the moving structure since the PSs lose the geometric information of the structure. The proposed method collects PSs on the same structure into a cluster using the kernel K-means with a new kernel defined on the basis of phase correlation and distance between PSs. Experimental results show that the proposed method can extract clusters indicating the shape of structures.