A study of high-precision stereo positioning with multi-scene and multi-mode GF-3 satellite images

Due to factors that include systematic sensor delay, it is difficult for synthetic aperture radar (SAR) satellite stereo positioning to meet the requirements of large-scale mapping. In traditional stereo positioning methods, ground control points (GCP) are required to eliminate errors. In this study, a method is proposed to improve the stereo positioning accuracy without GCP. First, the factors influencing the Gaofen-3 (GF-3) satellite stereo positioning accuracy were analysed. The errors were divided into systematic fixed errors, spatial and temporal variation errors, and random errors. Subsequently, a systematic geometric error compensation model and a stereo adjustment model were constructed, and a method to improve the multiscene and multimode stereo positioning accuracy without GCP was proposed. Finally, the effects of different intersection angles on the accuracy of stereo positioning were compared and analysed, and the optimal intersection angle of the SAR satellite was quantitatively analysed. The effectiveness and accuracy of this method were verified using GF-3 high-resolution image data. The experimental results indicate that the proposed method can improve the SAR satellite stereo positioning accuracy. The plane positioning accuracy improved by 95.1%, and the elevation positioning accuracy improved by 97.9%. In general, the plane and elevation positioning accuracies of the GF-3 satellite were better than 3 m and 2 m, respectively. The stereo-positioning accuracy of the SAR satellite was the best and most stable when the intersection angle was between 40° and 50°.