Research On 3D Reconstruction Technology Based On Binocular Stereo Vision

With the development and advancement of computer vision technology and image processing technology,three-dimensional reconstruction technology has been widely applied in many fields such as automatic driving,video surveillance and industrial inspection.As an implementation method of three-dimensional reconstruction,binocular stereo vision has the advantages of simple principle,excellent effect and low cost so that it has received extensive attention.Binocular stereo vision system uses two cameras to imitate the human binocular vision system and take images for the target scene under different observation positions.The three-dimensional information of the target scene is calculated according to the position deviation of the matching points in the left and right images based on the triangulation principle.This thesis is devoted to study some problems in the three-dimensional reconstruction technology based on binocular stereo vision,which has certain application scenario and clear research significance.This thesis mainly studies the main steps in the binocular stereo vision system,including camera calibration,stereo matching and three-dimensional coordinate calculation,and improves the result of the three-dimensional reconstruction.The main works are as follows:Firstly,this thesis analyzes the principle and model of binocular vision,selects the reasonable binocular structure model according to the actual application scenario,and selects the cameras,optical lens and light source according to the shape and size of the research target,shooting distance and shooting range.According to the actual application,a binocular stereo vision hardware experiment platform is built.Secondly,this thesis analyzes the camera calibration theory and uses the checkerboard calibration board to perform camera calibration.The method for detecting the checkerboard corner based on the circular edge is proposed,which improves the effectiveness of checkerboard corner detection and reduces the workload in the calibration process.Thirdly,in the research of stereo matching algorithm,this thesis focuses on the adaptive support weight algorithm in order to improve it.Considering the shortages in the adaptive support weight algorithm,a stereo matching algorithm based on sparse adaptive cost aggregation is proposed.Specifically,the cost function is improved and the initial cost aggregation is performed according to the improved adaptive support window,and then the disparity range of each pixel is selected to form sparse disparity subset.The adaptive support weights in the adaptive support window are improved based on the coarse disparity information.The cost aggregation is performed on the disparity subset so that final disparity is calculated.The experimental results show that compared with the traditional adaptive support weight algorithm,the improved algorithm has improved the evaluation index in different regions,especially in the disparity discontinuous region and the repeated texture region.Finally,the whole experimental flow framework for the three-dimensional reconstruction system is designed.The improved stereo matching algorithm is used to solve the disparity image of the target scene,and the three-dimensional information of the target scene image is calculated based on the principle of triangulation.