3D Reconstruction Based On Binocular Vision

Binocular stereo vision is one of important research directions in the field of computervision. It’s a method that simulate directly the human eyes and use the information of imageparallax to recover3D scene, and it has the advantages of low cost, wide range of applicationsand so on. And in the field of computer vision, three-dimensional reconstruction is always ahot issue. A lot of researchers at home and abroad are engaged in the study of this issue. Withthe rapid development of computer technology, the technology of binocular stereo vision hasbeen widely used in the field of robot navigation, industrial automation, medical diagnostics,military and so on.First, this paper introduced the status of the research about three-dimensionalreconstruction based on binocular stereo vision at home and abroad, which elicited the goal ofthis paper, that was to achieve three-dimensional scene reconstruction via the binocular stereovision system. Because real-time capability and accuracy of three-dimensional reconstructionare both important, this paper focuses on five key technologies, namely Image pre-processing,feature extraction, camera calibration, stereo matching and the solving of three-dimensionalcoordinates. At the same time, many corresponding experiments have been done for each keytechnology.In the part of Image pre-processing, We use the technology of Cross-Median filtering toeliminate the noise signal, and use the technology of histogram equalization to extrude imagedetails. In the part of feature extraction, An improved algorithm of corner detection ispresented to locate corner position to sub-pixel accuracy and laying a good foundation forestimating fundamental matrix. Canny edge operator is used to detect the edges of images. Inthe part of camera calibration, the calibration based on Zhang method is used to get theinternal and external parameters of camera. In the part of stereo matching, after consideringthe demands about accuracy, speed and dense disparity map, an improved stereo matchingmethod which based on region stereo matching, is proposed to obtain scene parallax map. Theimproved method which combine image segmentation method, two-way stereo matchingstrategy and pseudo-epipolar constraint can achieved efficient stereo matching between twoimages. Finally, the paper will use the disparity distance measurement method to get distanceinformation, and then the disparity map which gets from stereo matching process will beturned to a3D rendering, and the3D view can be displayed by MATLAB and OpenGL,respectively. The paper used MATLAB and Visual C++6.0to compile the corresponding algorithm.The results of experiments showed that the methods used in this article are correct andfeasible. The work of the paper provides a basis for the research and application of robot.