Research On Stereo Matching Algorithm Based On Triangulation Method

Stereo matching is one of the key technologies in stereo vision. Its core problem is to establish correspondence between images under different views of the same scene structure which are projected. Stereo matching is the basis of the three-dimensional reconstruction,and stereo matched results directly related to the effect of three-dimensional reconstruction. However, due to the complexity of space scene, the distortion of the imaging system and other reasons, the resulting of disparity map is not ideal. The papers proposes stereo matching algorithm based on how to achieve rapid, real-time, accurate stereo matching, and discusses the stereo matching algorithm based on the triangulation method. The main work can be summarized as follows:Firstly, starting with the research of special scattered points, an improved region growing triangulation method was proposed to guarantee the accuracy of triangulation in the object surface with sharp boundaries. The mesh growth direction is determined by introducing and calculating the weight of boundary edge. The mesh growing is expanded from small weight edge to large weight edge step by step. As a result, the triangulation of object surface can be achieved from "flat" to "unflat", and the corresponding mesh topological operations and queue updating mechanism can guarantee correctness of the boundary edge queue.Secondly, this paper do studies on stereo images matching algorithm based on triangulation. At first, it calculates the matching credibility of the triangles, and then classifies the triangles by matching credibility. Different match strategies were used to obtain dense disparity map for the classified triangles next. At last, we can combine the obtained dense disparity map of all sorts to form the dense disparity map of the whole image. It makes up the shortage of using global matching algorithm or local matching algorithm alone. Finally, it uses Matlab as experimental platform of stereo matching based on triangulation method, the triangulation method and the stereo images matching algorithm have good performance. The effectiveness and the feasibility of the proposed triangulation method and the stereo matching algorithms are verified through experiments.