Adaptive Re-weighted Block Adjustment for Multi-Coverage Satellite Stereo Images without Ground Control Points

Georeferencing satellite imagery is the basis for mapping and other subsequent processing in photogrammetry and remote sensing. Traditional georeferencing methods, which require adequate and well-distributed ground control points (GCPs), are time-consuming, costly, constantly cumbersome and inefficient. This study proposes a method of adaptive re-weighted block adjustment without GCPs for satellite stereo images. Furthermore, an approach of using multi-coverage images to improve the direct georeferencing accuracy is also investigated. The bias-compensated rational function model with affine transformation parameters is used as the adjustment mathematical model. The “weighted virtual control points” are introduced as constraints in the adjustment. Space intersection and resection are executed iteratively to solve the accurate affine transformation parameters. A total of 722 three-line array stereo scenes of Chinese Mapping Satellite-1 images captured from December 2010 to March 2016, which cover the entire Shandong Province of China (about 158,000 km2) nearly 10 times in space, are used as experimental data to evaluate the performance of the proposed method. Experiment results demonstrate that the proposed method can effectively eliminate the systematic errors of images. Moreover, the additional comparison with experiments of three groups of mono-coverage images indicates that the redundant observations from multi-coverage images can significantly improve the direct georeferencing accuracy, more specifically, from 13.69–16.02 m to 11.09 m in the horizontal direction, and from 16.39–29.27 m to 8.47 m in the vertical direction.