DEM Generation Using Circular SAR Data Based on Low-Rank and Sparse Matrix Decomposition

This letter presents a new approach of digital elevation model generation using circular synthetic aperture radar. The approach is composed of two steps, target separation and height estimation. In step 1, subaperture images are reshaped into vectors and stacked to build a composite matrix. The composite matrix is decomposed into a low-rank matrix and a sparse matrix based on Semi-Soft Go Decomposition algorithm. The targets, whose height is equal to the reference imaging height plane, are contained in the low-rank matrix, whereas the targets with other heights are contained in the sparse matrix. In step 2, the sparse matrix is further decomposed into several shifted low-rank matrices, each of which corresponds to the targets sharing one height, based on Shifted Subspaces Tracking algorithm. The height of the targets contained in each low-rank matrix is estimated from the shifts of the matrix's subspaces. The effectiveness of the proposed approach is investigated by the Gotcha public release data set.