Research On Reconstruction Method Of Point Cloud On Brake Disc Surface Based On Binocular Stereo

Stereovision is widely used in autonomous driving,virtual reality,three-dimensional reconstruction and other fields by simulating human eyes to obtain depth information of the environment.In order to improve image quality,reduce the influence of environmental background on model reconstruction accuracy and speed,and solve the problem of mismatching in weak texture region.In this paper,binocular stereo vision method is used to restore the spatial structure and surface point cloud model of brake disc through camera calibration,image segmentation,stereo matching and point cloud reconstruction.At the same time,put forward the corresponding improvement plan for the problems encountered in the actual work flow.The main work and innovations of this paper are as follows:(1)A camera calibration algorithm is proposed to solve the problem of camera parameters and distortion correction.The mathematical model of binocular stereo system was established,and the transformation relationship between imaging coordinate systems was studied.The internal and external parameters of binocular camera were solved by Camera calibration method,and the maximum likelihood estimation method was further used to improve the calibration accuracy.Finally,the distortion correction and stereo correction of camera were completed.Experimental results show that the average error of the calibration algorithm is 0.03 pixels,which can effectively remove lens distortion and enhance image quality.(2)An improved Grab Cut image segmentation algorithm was proposed to solve the problem of incorrect segmentation caused by uneven illumination in the workshop during the process of brake disc sorting.Firstly,adaptive median filter is used to remove image noise.The image brightness can be improved and contrast can be enhanced by limiting contrast adaptive histogram equalization.Laplacian operator is used to sharpen the image.Finally,the brake disc region was extracted by Grab Cut algorithm.Experimental results show that compared with the original algorithm,the proposed improved Grab Cut algorithm can effectively reduce the influence of illumination imbalance on the background removal of brake disc image,and improve the segmentation accuracy by 12.5%.Compared with the actual brake disc model,the segmentation result has a small error,which provides a new approach for the complex background removal of brake disc image.(3)Aiming at the low accuracy of traditional stereo matching algorithm and the problem of mismatching in weak texture region,an anisotropic stereo matching algorithm(ASMSI)integrating multi-scale information was proposed.Firstly,the anisotropic matching cost calculation function is constructed,and the gradient information is introduced into the cost calculation process to improve the robustness of the algorithm to illumination.Then,the matching cost in each support domain is calculated by cross-cost aggregation of multi-scale information.On this basis,left-right consistency detection and post-processing are carried out to obtain the refined parallax graph.Finally,the algorithm model is evaluated by comparing the false matching rate and running time of non-occluding and depth discontinuous regions.Experimental results show that the proposed algorithm can effectively solve the problem of mismatching in weak texture regions,and the matching accuracy is improved by 5.02%,which can meet the requirements of high efficiency and high precision in stereo matching process.