A novel solution for stereo three-dimensional localization combined with geometric semantic constraints based on spaceborne SAR data

Abstract

Improved resolution and geolocation accuracy of spaceborne SAR images make refined digital surface model extraction and delicate three-dimensional reconstruction promising. Currently, extracting a refined digital surface model by stereo SAR remains challenging due to the significant matching errors. In this paper, we quantitatively analyze the effect of matching errors on stereo SAR based on a simulation method. Subsequently, a novel stereo SAR approach combined with geometric semantic constraints is proposed to reduce the effect of matching errors on localization. In this approach, a stereo localization model combined with geometric semantic constraints is utilized to jointly locate a group of points with planar or linear structures. The least squares solution of the system of equations will be solved by an iterative method based on correcting characteristic values. Finally, the simulation experiments show that the planar and linear semantic constraints improve the relative positioning accuracy by about 26% and 70%, respectively. Experiments based on real data also demonstrate that the proposed method is significantly better than traditional stereo SAR.