A Stereo Matching Algorithm for Non-metric Digital Images and the Application

Compared to aerial images, there are more complicated image distortion and occlusion in non-metric digital images in the aspect of close-range photogrammetry, which increase the difficulty of the image matching dramatically. Due to the particularity of non-metric digital images taken in alpine and gorge regions, this paper proposes a probability relaxation matching algorithm with an improved searching strategy. The algorithm integrates the gird points with feature points to determine the initial point matching process and conducts multiple constraints in the respects of epipolar lines and parallax to ensure the continuity and correctness of matching. The experiment shows the algorithm is applicable for the matching of non-metric digital images in alpine and gorge regions, whose stereo matching correctness can reach up to 98% in alpine and gorge regions.