Tomographic Inversion of Urban Area via Tikhonov Regularization and Bayesian Information Criterion

Abstract

As an extension of synthetic aperture radar (SAR), SAR tomography (TomoSAR) technology can reduce the overlapping in 2-D SAR image and separate multiscatterer along the elevation direction, thereby achieving the high-precision 3-D reconstruction of the surveillance area. However, in practical spaceborne TomoSAR application, the quality of 3-D imaging is restricted by the limited number of baselines and their uneven distribution. Therefore, it is necessary to find advanced signal processing technology to achieve the target 3-D recovery when the amount of data is limited. In this letter, a novel Tikhonov regularization and Bayesian information criterion (BIC)-based nonparametric iterative adaptive approach (IAA), named RIAA-BIC, is proposed and introduced to the spaceborne data processing. Compared with conventional spectral estimation, compressed sensing-based, and IAA algorithms, the proposed method incorporates the Tikhonov regularization term to avoid the problem of solving nonlinear ill-posed equation in the elevation inversion. Furthermore, the BIC model selection tool can eliminate the false or weak scatterers, thereby improving the 3-D reconstruction accuracy of the surveillance area. Experimental results based on TerraSAR-X dataset verify the proposed method.