Research And Implementation Of Key Technologies In Binocular Stereo Vision System

In recent years,the research of computer vision has developed rapidly.Binocular stereo vision technology is one of the most active development directions in the field of computer vision.It has shown superior application value in the fields of robot vision,virtual reality,driverless,and three-dimensional reconstruction.Binocular camera calibration and stereo matching technology are the most important and challenging parts of binocular stereo vision technology,and their research restricts the development of binocular stereo vision technology.In this thesis,the in-depth research on the problems of these two key technologies will be studied,improved and optimized the basic algorithms,and finally designed a stable and efficient experimental platform for binocular stereo vision system to achieve the depth information acquisition of the target object.The main work of this paper is as follows:(1)The thesis studies the basic theory of binocular stereo vision in detail,including camera linear and nonlinear imaging models,imaging principle of binocular stereo vision and geometric constraints of stereo vision.The key technologies of binocular stereo vision and the technical route of this paper are further elaborated.(2)For the problems of classic Zhang Zhengyou checkerboard calibration algorithm,such as the calibration accuracy is low,the calibration cannot be performed automatically,and a large number of artificial errors are introduced,this thesis improves the classic checkerboard template.The Harris corner detection and judgment algorithm is used to accurately extract the pixel corner points,and further obtain sub-pixel level corner points.An automatic selection algorithm for peripheral corner points is proposed to determine the range of corner points and achieve automatic calibration.The experimental results show that compared with the traditional calibration method,the method eliminates the artificial error and improves the parameter accuracy,which is convenient and effective.(3)For the calculation of matching cost and the difficult selection problem of matching window in local stereo matching algorithm,this "paper proposes a matching cost calculation method for multi-feature fusion by analyzing the defects of traditional algorithm.The dynamic window is selected and the weights are assigned to each pixel in the window for cost aggregation,and the disparity value acquisition and the disparity map are optimized.The proposed algorithm is simulated by using the standard image provided by Middlebury database.The experimental results show that the proposed algorithm improves the parallax accuracy of the occlusion area and the parallax discontinuous area,reduces the false matching rate,improves the robustness of binocular vision while maintaining good real-time performance.Using the camera calibration and stereo matching algorithm proposed in this thesis,the target depth information is acquired.The experimental results show that the binocular stereo vision system can obtain accurate target depth information within a certain range,which verifies the effectiveness of the key technology research.