Research On Partial Occlusion And Feature Uncertainty In Line Stereo Matching

Line stereo matching is the key step of target pose estimating, image registration and 3D target reconstruction. But partial occlusion and feature uncertainty are the two difficult problems which make line stereo matching hard to be used practically. Our group have already researched these problems, a method for partial occlusion based on line feature grouping and a method for feature uncertainty based on Vague Set theoty are proposed. Basing on these achievements, an algorithm of line stereo matching for solving partial occlusion and feature uncertainty is proposed, and the problem for computing stereo imaging model parameters is researched in this thesis. As a whole, the main achievements are listed as following:(1)A method for evaluating stereo camera intrinsic parameters and relative orientation model parameters is devised and implemented. It computes camera intrinsic parameters through finding the corresponding relation of the identifying points in the image and on the demarcate board, and computes the spatial relation of stereo cameras simultaneously using the transform parameters between each camera and the demarcate board, and then gets all the unknow parameters in stereo imaging model. The proposed algorithm is proved to be simple, efficient, convenient and stable.(2) An algorithm of line stereo matching for solving the problems of partial occlusion and feature uncertainty is proposed. Firstly, it establishes the uncertainty model of all lines and computes their uncertainties. Secondly, it computes the similarity between uncertain features through the concept of Vague Set theoty, and establishes their initial matching relations. Thirdly, it uses line feature grouping method to solve partial occlusion, and computes the energe of line feature groupings using each line's uncertainty. Finally, it uses a optimization method from Gragh theoty to find out a global consistent line feature grouping set, and then gets the matching relations between all lines. This algorithm is proved by many experiments.