3D Reconstruction Method Based On Binocular Stereo Vision Research

Vision system is the most important sensory systems of the human body system, human can obtain information from the outside world come mainly from the visual. At the same time, the visual technology is a big hot field of computers and robots. Compared to humans, the robot obtaining external information also rely mainly on visual. This paper mainly discusses and research several keys of binocular stereo vision technology. To the same object from two different angles to observe, Binocular stereo vision gains the awareness to different perspectives on the same object image, then through calculation of the principle of imaging geometry scene parallax between image pixels, can achieve the3D information of target scene.In this paper, we study the3D reconstruction based on binocular stereo vision, which consists of image acquisition system, camera calibration, the eigenvalue extraction and stereo matching, and several important component of3D reconstruction.Camera calibration is the premise of the link, this article selects a kind of methods which bases on planar calibration template, between the traditional calibration methods and self-calibration method, through experiment and computer calculation to get the parameters of the camera.Looking for the same target points in two images of the imaging position, stereo matching mainly contains three main steps, the selection of matching characteristics, the principle of matching rules, the realization of the matching algorithm. Stereo matching method based on the characteristics mainly discusses the SIFT feature point extraction and matching, to obtain digital image coordinates of the matching point in the image. Then, according to the basic principle of stereo vision,the method of3D coordinate, calculation, the mothod calculates the three-dimensional coordinates of the image and obtains the error by comparing the distance of feature (?)oints and laser ranging sensor information.Comprehensive analysis of the causes of eror, according to the measurement results,this paper generate the depth map of3D environment, The experiment is studied by using MATLAB, OpenCV algorithm (?)rogramming.