Study On 3D Reconstruction Method Based On Binocular Stereo Vision

At first, this paper summarizes the commonly used method of 3d data acquisition,on this basis, it points out the superiority of binocular stereo vision, witch is the basic theory of 3d data acquisition in this paper.Binocular vision as an important branch of computer vision, can directly imitate the human visual processing objective scenes, with low cost, simple equipment, flexible use and other features, has been widely used in various industries. However, due to certain factors and conditions, binocular vision is not as universal as human vision, and the reconstruction accuracy is directly affected by the accuracy of the matching. The research of binocular vision technology is still of great significance in application and research.In this paper, several key technologies of binocular stereo vision are studied.Binocular vision through two cameras captured images from different angles, then the corresponding points in the image are obtained by stereo matching technique., next using the correspondence point to calculate the pixel position offset(disparity), finally obtain three-dimensional point information by geometric principles. The main contents of this paper are camera calibration, stereo matching, and 3D reconstruction.In the Camera Calibration section, according to the requirements of the actual problem, this paper calculates the requirements of the camera and the camera lens.A binocular vision measurement system with adjustable baseline is set up. Write code to achieve the simultaneous acquisition of image data. The Zhang Zhengyou plane calibration method is adopted between the self calibration method and the traditional calibration method. And refer to the MATLAB toolbox calibration results, the correctness of the calibration parameters are calculated and verified.In stereo matching section,realization of feature points matching based on SIFT feature.And the random sample consensus algorithm is adopted to eliminate the false matching points in the perspective projection model.The experimental results basically meet the requirements of the camera forward parallel configuration model.A global stereo matching algorithm based on structure feature is proposed from the theoretical point of view.Inspired by multiple criteria, the data items in energy function of stereo matching algorithm based on dynamic programming are improved.Disparity space search restrict for the smoothing term.At this time,The computation time is not increased too much while the matching precision is raised.In the three-dimensional reconstruction part, according to the principle of stereo vision, the ideal configuration model of the camera is used to calculate the 3D coordinates.In the left and right view images, the position of the circular workpiece is extracted. Matching and calculating the corresponding point disparity.Calculate the 3D coordinates of the surface points of the workpiece,and then Calculate the diameter of the circular workpiece,finally the result is compared with the actual measured value.