A Reliable Matching Algorithm Based on the Rational Function Model Bundle Adjustment for Satellite Imagery

Aim to the practical problem of high-resolution imagery automatic matching, the paper proposes a reliable matching algorithm based on the rational function model bundle adjustment, the basic idea of which is the introduction of the rational function model bundle adjustment to the pyramid image matching, this will establish the geometric relationship between images, and realize the accurate prediction of matching points´ initial position, fast matching with epipolar line constrained and rough geometric correction, and the elimination of error matching points. Refine matching layer by layer, and finally using the least squares matching algorithm to obtain the matching result of the sub-pixel level. Experiment has been made on the World View satellite imagery of Qinghai province in China, and it has shown that the method proposed in this paper can effectively overcome the impact of image radiation distortion, image rotation and some geometric deformation and so on, achieving the automatic and reliable conjugate point matching of satellite remote sensing imagery.