Color 3D Scene Reconstruction And Display Based On Binocular Stereo Vision

Binocular stereo vision is a category of computer vision, it has been more than 30 years since it was proposed. At the same time of gradually improvement, it is active in robot navigation, virtual reality, digital entertainment and 3D reconstruction of the more popular fields. Achieving three-dimensional reconstruction of the scene by taking advantage of the binocular stereo vision theory has very important application merits in human-computer interaction, virtual reality and other aspects.In view of the three-dimensional scene reconstruct and display based on binocular vision, the following aspects were studied:Firstly, pursuant to the parallel binocular vision camera system was established, taking the pinhole camera imaging model for example, the transformation between camera calibration involved coordinate systems was analyzed; the disparity map transformation depth map geometry measurement principle was discussed, the theory of camera calibration algorithms and formulas were derived. The corresponding simulation experiments are accomplished based on the Open CV and software platform.Secondly, how to selected the matching primitive and similarity measure function are analyzed and the algorithms theoretical and primary processes are summarized on the basic of the researches of stereo matching algorithm at home and abroad, the main factors should be considered in the presence of matching error variation are analyzed. The seed point growing local algorithm and the graph cuts global algorithm is accomplished on the VC++ platform and the algorithm time complexity are contrasted, the conclusion proved that choice the reasonable matching window is the key way to optimize local algorithms; global matching algorithms can achieve dense disparity maps, but the algorithms execution time is slower; disparity map can process better by image filtering. These conclusions will provide a reliable reference for the future researches in stereo matching algorithms.Thirdly, the filtering and edge thinning algorithm of image enhancement are analyzed in view of the color image processing methods involved in the reconstruction programmer; the suitable algorithms for different situations are discussed on the basic of comparing the simulation results. Using the pseudo-color processing in disparity maps strengthen the sensory awareness, the inverse relationship between the depth and disparity are proved directly.Fourth, obtaining the three-dimensional coordinates of the object and the depth of information and generating the data cube based on the principle of triangulation, increasing the Open GL texture mapping to display more realistic,using the red and blue glasses observed three-dimensional view. For the causes of the human visual fatigue and discomfort in the current 3D view, the corresponding solution are proposed, the researches will provide a reference for the subsequent optimization of the 3D display quality.