Research Of The Key Technology Of Stereo Matching And Point Cloud Reconstruction

Stereo matching and point cloud reconstruction are traditional research topics in thefield of computer vision, which use cameras to shoot the same object from different anglesin3D scene, and then obtain the images of different perspective and the3D information ofthe object through a series of algorithms.First, research of the topic background and the present situation of stereo vision andmake a detailed study of the basic theory of the stereo vision technology.Second, conduct a detailed analysis of the classification and the problems of stereomatching and the relationship of disparity and depth.Third, propose an image Interpolation zoom algorithm based on the SIFT operator.The innovation of this paper is enlarging the images through tnterpolation before usingSIFT matching. Experimental results show that the stereo matching algorithm this paperproposed not only improves the matching accuracy of the original algorithm, but preservsthe robustness and the immunity of the SIFT algorithm for image affine transformation,3D perspective transformation and image gray nonlinear transformation also. Therefore,the proposed algorithm can be applied to the stereo vision system which needs higherrobustness.Finally, propose a surface reconstruction method based on Delaunay triangulation.The method uses the the incremental insertion point to build space delaunay divided, makecircumcircle (circumradius) minimum as criteria and constitue right triangle surfacesextracted from the tetrahedral delaunay triangulation. The experiments show that thealgorithm is effective and it can avoid the generation of the narrow triangle to a certainextent. Whether the build surface is a closed surface or not the reconstruction algorithmcan get better effect.