Research And Application Of 3D Reconstruction On Moving Vehicle Based On Binocular Stereo Vision

As a new field based on various interdisciplinary, stereo vision is an important attribute of intelligent robot. Binocular stereo vision is stereo vision an important branch, has been applied in many fields, such as medical image processing, multimedia technology, the analysis of sensing image, image databases, etc.The integrated binocular stereo vision system generally includes image acquisition, camera calibration, features extraction, stereo matching, 3D reconstruction, etc. It is through further research of binocular stereo vision that we can use mobile robots instead of human examiners in Automatic Stake Test System to provide a theoretical basis and reference.The following are the major contents in the thesis:1. Camera calibration based on one-dimensional calibration object is proposed. The 1D calibration object is moving around a fixed point. By taking at least six pictures of the rolling object at a fixed camera view position and direction. We can set up a set of linear equations using the property of perpendicular lines. Thus, a camera intrinsic matrix can be obtained by solving these equations. The simulation results show that the method is feasible and effective.2. Stereo matching algorithm based on image segmentation is proposed. Based on Bleyer algorithm framework the calculation of planes and the selection of the global energy function has been improved. The plane parameters of each segment are computed only using reliable pixels, and in accordance with the characteristics of unreliable points, construct a new segment matching cost for group neighboring segments. The scope of the initial disparity is used as a constraint of the global energy function, the energy function is globally minimized using graph cuts.The experimental results show that the algorithm has a better performance in low texture regions and occluded portions, and has high accuracy and real time.3. Three-dimensional reconstruction based on binocular vision is proposed. Making use of mid-point of crossover line to get 3D points coordinate, as the algorithm to consider linear equations of geometric significance, thus enhancing the3D points coordinate precision. The simulation results show that mid-point of crossover line method has higher accuracy than the least squares.In the end, the results of 3-D reconstruction is given. The simulation are performed by Matlab language. Experimental results show that the results of three-dimensional reconstruction can basically reflect the geometric features of the moving vehicle.