Geometric constraints based 3D reconstruction method of tomographic SAR for buildings

The mainstream methods of tomographic synthetic aperture radar (tomoSAR) 3D reconstruction are usually realized by processing the registered 2D SAR images pixel by pixel without considering the geometric structure of the observed targets. As a class of targets with regular facades, tomoSAR 3D reconstruction of buildings is of important application value nowadays. In this study, to optimize the tomoSAR 3D reconstruction results of buildings, we first introduce and analyze the tomoSAR 3D reconstruction theory based on a back projection imaging algorithm and then present a tomoSAR 3D reconstruction method based on geometric constraints. We can accurately extract the building's facade as the geometric constraints by processing and fusing the point density map and the height map extracted from the spatial distribution of the tomoSAR 3D point cloud. Taking the extracted geometric structure of the building's facade and signal to noise ratio (SNR) of registered 2D SAR images as prior information, we then impose geometric constraints on tomoSAR sparse 3D reconstruction with an adaptive iterative shrinkage-thresholding algorithm and achieve an optimized tomoSAR 3D reconstruction result. Finally, measured data of a P-band airborne tomoSAR system is used. 3D reconstruction results show that the proposed method outperforms the traditional methods without geometric constraints regarding 3D reconstruction performance, which proves the validity and practicality of our proposed method.