Research On Three Dimensional Reconstruction System Based On Binocular Vision

With the in-depth research of robot, machine vision research and exploration will continue to the depth and the breadth development, machine vision modeled human eyes to obtain the information, information, information storage, information processing,information transmission, try to achieve in some occasions instead of human visual working, in some of the scenes of human visual function complete the information acquisition and storage. There are two main directions in the research of machine vision,one is imitating biology, one is based on the principle of geometric imaging. In this paper,we choose the research direction based on the principle of geometric imaging, the three-dimensional reconstruction of binocular vision depends on the application of machine vision to a certain extent, this technology has a lot of places to improve and optimize. In this paper, after studying the results of the study, the paper summarizes some of the process and methods of using binocular camera to get the left and right images for3 D reconstruction.The framework of Marr vision theory lays the foundation for the theory and model of machine vision, and promotes the development and progress of computer vision. The theoretical framework is divided into the visual system from the level of computing theory,the expression and algorithm level and the hardware level. According to the Marr vision theory, two-dimensional image input in the process of 3D reconstruction, 3D object shape position and attitude of the output, the visual system to achieve to establish relations and constraints between input and output, the research and Realization of correlation function algorithm. Binocular vision based three-dimensional reconstruction technology relies on machine vision, using two cameras to obtain the same three-dimensional objects from different angles around the two-dimensional image to simulate the reconstruction of three-dimensional model.Three dimensional reconstruction based on binocular vision includes image acquisition, image preprocessing, camera calibration, feature point extraction, feature matching, 3D reconstruction and so on. Based on the geometrical theory, in around a two-dimensional image search feature pixel point, theory of matching based on, by pairwise matching points for the 3D coordinates of the feature points, and depth map said to calculate the effect of 3D point, around objects, up and down, before and after the relationship expression. The focus of camera calibration method, feature point extraction method, feature matching method of research and elaboration, process studied Harris corner extraction algorithm, based on window sparse point matching and matchingalgorithm based on window. The intensive, in view of the advantages and disadvantages of the two algorithms proposed a better matching algorithm based on window compromise sparse dense point matching method, the matching method can improve the accuracy rate of matching and improve the matching rate.Build a hardware platform, combined with the environment to get the picture to meet the research requirements, the picture format is BMP format, the left and right images are related to the experimental VC platform in combination with OpenCV and OpenGL. The combination of OpenCV and OpenGL is an important tool in the study of computer vision and 3D reconstruction.