The Research Of Three-dimensional Reconstruction Base On Binocular Vision

Three-dimensional (3D) reconstruction is a separate direction in branch of computer vision, which has intensively been developed thanks to the potentials of modem computers. In its promotion of a wide range of applications, resulting in a lot of algorithms, the use of three-dimensional reconstruction of binocular vision is one of them. Binocular vision is to use the camera gets the perceived image of multi-dimensional perspective of the scene objects through geometric principle of imaging, on access to image pixels between the location of deviation (parallax) is evaluated to recover three dimensional information of scene objects. Research on it, in both the Visual Physiology and its application in engineering has very important significance.This paper, starting from binocular vision imaging, for the binocular image acquisition, camera calibration, image preprocessing, stereo matching up to three-dimensional reconstruction of the whole process of research and analysis in detail, working mainly for the following several:1,Camera calibration, the calibration of the traditional methods were compared. Uses a range of traditional and self-calibration method between calibration-Zhang Zhengyou's the camera calibration template based on planar calibration approach to achieve calibration.2,Image feature extraction, and studied the classical algorithm of noise removal, in traditional Laplacian operator in the lack of edge detection,we proposed combining Sobel operator with Laplacian operator with the edge of the feature extraction.3,Stereo matching, lays emphasis on the research of matching method based on feature points, using the SIFT feature extraction and matching algorithms, and obtained the image coordinates of matching points. This method is fast, and to facilitate the matching of three-dimensional reconstruction rapidly.Finally, the paper experiments with Matlab simulation environment, realizes from camera calibration to image preprocessing and stereo matching and the whole process of three-dimensional reconstruction, obtained good effect of reconstruction.