Efficient Tomographic Inversion Based on Refined Scatterer Pre-Estimation

Abstract

During the last decade, synthetic aperture radar (SAR) tomography (TomoSAR) technique in 3D reconstruction is significantly improved. In particular, compressed sensing (CS) has become the best suit for urban tomographic inversion. Among CS algorithms, the orthogonal matching pursuit (OMP) combined with Bayesian information criterion (BIC) has relatively low computational complexity, however, its scatterer estimation accuracy needs to be improved. To address the problem, a prior scatterer estimation with an innovative triplethreshold detection algorithm is proposed which not only greatly reduces number of iterations but also helps suppress artifacts. Experiments on real data show that the priori overlay times maps provided by the proposed algorithm are closer to the reference maps than other two pre-estimation algorithms. Artifacts in 3D reconstruction results are well suppressed at less than half the time cost of original OMP-BIC.