| Binocular stereo vision ,which acquires perceptive pictures in different perspectives from two eye points,is a process that obtain the three-dimensional structure of things through calculating positional deviation between image pixels based on Imaging Geometry Principle. That is similar to human's visual perception process. Binocular stereo vision is always one of the key research of computer vision. With the development of technology,binocular stereo vision will be widely applied in topographic survey, robot navigation, and so on. Therefore, the research of this subject has significant practical significance and application value.Firstly, the basic principle and steps(camera calibration , stereo matching and 3D reconstruction) of binocular stereo vision system are introduced detailedly in this paper. Then, it focuses on the research of triangulation algorithm of binocular stereo vision.All programs in the paper are completed based on the platform of Visual C++.In the stage of camera calibration, considering the camera and experimental conditions, we use the method of Zhang Zhengyou Demarcate to obtain the relevant parameters of camera. In the stage of stereo matching, by selecting angular point as features and using matching algorithm found on the combine of template matching and feature matching, we could calculate the feature points of image. Then, through calculating parallax of selected points after camera calibration and stereo matching, so as to obtain the deep information of data point. In the stage of 3D Reconstruction, methods of spatial points'restruction are used to generate Delaunay triangulation by Bowyer-Watson algorithm, and improve searching strategies of point location. Meanwhile, it advances a new data structure,which improves the executed efficiency of triangulation programs. Delaunay triangulation, which has been disposed, achieves three-dimensional visualization By programming with OpenGL and Visual C++ finally. |
PDF Full Text Request