Method for Improving Positioning Accuracy of Rotating Scanning Satellite Images via Multi-Source Satellite Data Fusion

Rotating scanning systems are capable of acquiring ultra-wide swath satellite imagery, but they suffer from significant positioning accuracy degradation due to complex geometric distortions and the difficulty of obtaining ground control points (GCPs) over vast areas. To address these issues, this paper proposes a precise positioning method based on multi-source satellite data fusion. By comprehensively utilizing high-resolution images from ZY-3 and GF-2 satellites alongside DEM data, we establish a framework that integrates grid-based feature point extraction, high-precision matching, and multi-image joint adjustment. Specifically, we introduce a matching strategy combining geometric constraints with Least Squares Minimization (LSM) and a robust joint adjustment model to suppress geometric distortions. Experimental validation was conducted using a dataset covering the Beijing area. The results demonstrate that after joint adjustment, the planar accuracy of the imagery reached 4.01 m, and the edge matching Root Mean Square Error (RMSE) between adjacent images was 2.52 m. Furthermore, the cooperative positioning accuracy for segmented simulation data achieved 4.68 m in mountainous areas and 5.22 m in plain areas, meeting the requirements for meter-level positioning. These results verify the effectiveness of multi-source cooperative adjustment in correcting geometric distortions and significantly improving the positioning accuracy of rotating scanning imagery.