Research Of Monocular Stereo Vision Matching Algorithm Based On Prism

At present,stereo vision is one of the hottest research topics in computer vision.Stereo vision is a discipline that studies how to use machine to imitate human vision.and then to understand and recognize the three-dimensional space.The theory of visual computing was proposed by Marr in the 80’s in twentieth Century.This basic theory makes stereo vision become a hot research in computer vision.Binoocular stereo vision is a very important branch of computer vision.However in the practical application,traditional binocular stereo vision’s large size and high cost lead to monocular stereo vision based on prism.Compared with the traditional binocular stereo vision system the single camera based on the prism has the advantages of smaller size annd lower cost.Stereo matching has always been the most important in stereo vision and the most difficult problem in stereo matching.and directly determines the quality of the 3D reconstruction.After decades of development.stereo matching has been studied by many researchers.The stereo matching is the depth information of the corresponding points in two or more images taken in the same space scene.Because the traditional stereo matching is mainly for the binocular stereo vision,compared with the traditional stereo matching,prism introduced results that stereo matching has changed a little in the prism Monocular Stereo Vision System.Aiming at the problem of stereo matching in the single camera stereo vision system based on prism,the stereo matching is improved.The main、work of this paper is:Firstly,the geometry of the single camera stereo vision is proposed,and the relationship between the left and right images in Monocular Stereo vision system is presented,the relationship model is the fundamental matrix.Then,the fundamental matrix is calculated from the image coordinates of the feature points which are few and accurate.Secondly,the feature points of the two parts of the image are extracted by SIFT algorithm,and the feature points are matched w ith the Euclidean distance.In order to eliminate the error matching,the SIFT features of the initial extraction are selected by RANSAC.and the matching points are found as far as possible.Thirdly,using the second part of the accuracy of the matching points of the image coordinates,derived from the first part of the fundamental matrix.Then,the basic fundamental matrix and the image coordinates of the feature points are used to search every feature points in the image of the corresponding polar line.Then,the two images are corrected for the polar line.The experimental results show that the prism Monocular Stereo Vision Measurement Method Baseed on.compared with the traditional binocular stereo vision,the improved stereo matching algorithm not only saves the camera calibration of the time,but also reduce the matching process in the search matching point range,reducing the algorithm’s time complexity.