Research On The Matching Algorithm Of Linear Feature In Binocular Image Sequence

Using the certain characteristics of two images to link the binocular images is thebasic task of the binocular sequence image matching. Linear characteristics is animportant clue for human perception and is also the main features of human recognitionand matching objects. So the straight-line characteristics of the image sequence areoften used in the binocular sequence image matching as a research medium. Using thebinocular stereo vision system and the binocular sequence image matching method isimportant for3D reconstruction, motion estimation, and the matching of buildings,navigation and other work. This paper studies the linear characteristics of the binocularsequence image matching algorithm. The main research contributions are as follows:(1) Line feature matching technology is made in-depth of analysis and research.First we get the experimental images from binocular system platform; Next study thealgorithm of the image filtering denoising and a suitable denoising method is selectedby comparing the experimental text for Chapters III and IV experiments; And thendescribes several common edge detection operators and obtain the advantages anddisadvantages of the various operators by compared experiments to select theappropriate edge detector as the paper of the detection operator; Finally, study themethod of edge fitted straight line, similarity measurement criteria and constraints,according to the need for the line feature matching of binocular sequence image.(2) Propose a method of matching algorithm based on straight lines and featurepoints. The straight line matching algorithm is based on spatial proximity andgeometrical relationship, the extracted linear are clustered into a line feature set; Thenusing the feature vector as the similarity measure to get a coarse match of the straightline; Finally, using the feature points as constraints, the mismatch in the coarsematching are removed to achieve an exact straight line matching.(3) Propose a matching algorithm based on intersecting lines features. Thealgorithm solves the problem of a single straight line of low texture matching and theambiguity of the matching, using intersecting lines to replace the traditional single-linematching. First using the normalized cross-NCC matching algorithm for imagepre-match to get the initial matching of linear. In order to solve the problem ofmismatch in the cross matching, the paper take advantage of the epipolar geometryconstraints on the initial match of intersecting lines to remove the mismatch. The experimental results show that the algorithm is feasible and effective.