Research On Algorithms For Images Matching Based On Two Views

Matching of images is a key subject in computer vision field and the base of theory and application of computer vision, which is widely applied in precise industrial measurement, object recognition, virtual reality, auto navigation, automatization of production, military affair and many other fields. Meanwhile, matching of images is a difficult subject in computer vision field, and many important theories and applications of computer vision are expanded under the precondition that matching problem has been solved. Consequently, matching of images is significant in theory and practice to study on this method.The systematic research on the algorithm of images matching based on two views is addressed in this dissertation, which includes images matching based on epipolar and homography constraints, images matching based on color and adjacency matrix, images matching based on Laplacian matrix. The main research works and achievements are outlined as follow:1. After studying the epipolar geometry, homography and solution of fundamental matrix in details, an algorithm for images matching is presented, which uses epipolar geometry and homography constraints. First, the cross-correlation matching is used to get the initial matching, then the fundamental matrix and homography matrix are estimated robustly with RANSAC algorithm, and in the same time most of wrong points are picked and deleted. At last, using optimized fundamental matrix and homography matrix, more precise matching points can be obtained through the guide matching.2. After studying the images matching algorithm based on graph spectra in details, an algorithm for images matching based on color grads and adjacency matrix is presented. First, the adjacency matrix is constructed using Gauss weighted function and color information of the neighborhood of feature points and decomposed by SVD. Then a relation matrix, which denotes the matching degree among feature points, is...