Research And Implementation Of Stereo Measurement System Based On Binocular Vision

Compared with other stereo measurement methods,binocular vision stereo measurement technology has the advantages of simple equipment,low cost and high efficiency.It is usually used in 3D facial recognition,robot obstacle avoidance,virtual reality and other fields.Therefore,it has been a research hotspot in the field of stereo measurement in recent years.This paper realizes the development and research of stereo measurement system based on the relevant principles of binocular vision.The main work contents are as follows:First,this paper systematically studies the binocular vision imaging model,and uses OpenGL,OpenCV,MFC and other tools to complete the design and implementation of a stereo measurement system based on binocular vision.The system provides functions such as camera calibration,stereo ranging,and 3D point cloud visualization.Second,this paper uses the checkerboard as the calibration template,and the corner feature information of the target needs to be extracted during camera calibration.Aiming at the problem that the traditional Harris corner detection algorithm is inaccurate in extracting the corner information of the checkerboard,this paper introduces the basis of the original algorithm.The fast binarization image preprocessing and corner filtering layer are improved,which realizes the effective extraction of checkerboard corners and improves the accuracy and reliability of the Harris corner detection algorithm.Third,in the stereo matching step,this paper uses the stereo matching algorithm based on SGM(Semi-Global Matching)of Census transform.Aiming at the partially matched areas in the disparity map obtained by the stereo matching algorithm,this paper proposes a local feature filling method based on interpolation to fill in the blanks to repair the unmatched pixels in the disparity map to a certain extent,and uses median filtering to smooth and denoise the obtained disparity map,which is applied to the disparity optimization step of stereo matching to improve the quality of the disparity map and the visualization effect of point cloud data.Fourth,in order to effectively manage point cloud data and improve the efficiency of 3D visualization of point cloud data,this paper designs and implements a hierarchical data structure based on depth features.This data structure can effectively filter out noise in point cloud data,and can use depth feature information to segment and extract point cloud data,which provides an efficient data structure for subsequent expansion research such as 3D point cloud visualization and 3D reconstruction of the measured object.Through the binocular vision stereo measurement system designed in this paper,the realization of functions such as camera calibration,image stereo matching,and 3D point cloud visualization is completed.The stereo ranging results in this paper are within a certain measurement range,and the error can be controlled within 1%.The 3D point cloud visualization effect is good and supports the segmentation and extraction of point clouds.