Research On Three-Dimensional Reconstruction Based On Binocular Stereo Vision

With the rapid development of computer technology, three-dimensionalreconstruction of image has become a research focus in computer stereovision. Three-dimensional reconstruction of the image is based on the usefulinformation which is obtained from the images, then finally getthree-dimensional information of objects in the scene. It’s widely used invisual navigation, medical diagnosis, aerial mapping, industrial inspectionand other fields. Therefore, the research has important scientific value andpractical significance.Binocular stereo vision method is an important way ofthree-dimensional reconstruction which based on Marr’s computer visiontheory. In the method, the same scene was shot by using two cameras indifferent viewpoints. Based on geometric principles, the recovery ofthree-dimensional scene could be achieved from the images of object. Thispaper researched mainly on three-dimensional reconstruction based onBinocular stereo vision and focused on the research of camera calibration,image pretreatment and edge detection, stereo matching, three-dimensionalreconstruction.In the part of camera calibration, the paper first introduces the linearcamera model and distortion model, detailed studies the classification of thecamera calibration method and conventional classic calibration algorithm:Tsai method and Zhang zhengyou’s plane template method. According to useZhang zhengyou’s plane template method, we can acquire camera’s insideand outside parameters. In the part of image preprocessing and edgedetection methods, the classic algorithms of image smoothing are: meanfilter、improved selective masking、median filter. The classic algorithms ofimage sharpening are: gradient sharpening and Laplacian sharpening. Weintroduce classic operators of edge detection such as Roborts operator, Prewitt operator, Sobel operator and Laplacian operator, proposing a noveltemplate based on Laplacian operator and verifying its available accordingto test. In the part of stereo matching, we introduce the classify of method ofstereo matching and research marcato feature point extraction based onHarris operator and feature points matching based on Constraint of EpipolarLine, acquiring test result. In the part of three-dimensional reconstruction,we describe the theory based on Parallel Two-eyes-visual and practicalTwo-eyes-visual model and achieve three-dimensional reconstruction toobject.