Research On 3D Reconstruction Of Gearbox Shell Based On Binocular Stereo Vision

Binocular stereo vision is one of the forms of machine vision,with advantages such as simple implementation and low cost.With the rapid development of machine vision technology,binocular stereo vision has been widely applied in fields such as unmanned aerial vehicles,autonomous driving,and 3D reconstruction.Especially the three-dimensional reconstruction technology based on binocular stereo vision has been widely applied in the field of industrial inspection,with high research significance and practical value.This article conducts research and analysis on the various steps of the threedimensional reconstruction method of the gearbox shell based on binocular stereo vision,including camera calibration,stereo matching,and three-dimensional reconstruction and measurement.Among them,stereo matching is the most crucial step in the 3D reconstruction calculation process,and the quality of the matching results directly affects the measurement results.How to effectively solve the impact of external factors such as lighting and noise interference in actual scenes,thereby improving measurement accuracy,is the focus and difficulty of research.This article focuses on the application of stereo matching in 3D reconstruction,with a focus on researching and improving stereo matching methods.The main research content of this article is as follows:(1)In terms of binocular camera calibration,this article selects the ZED mini binocular camera as the camera structure model based on the actual scene application requirements of gearbox shell 3D reconstruction.The imaging model and measurement model of the binocular stereo vision system are theoretically analyzed,and the Zhang Zhengyou calibration method is used for camera calibration.The camera parameters were obtained through experiments,and the results showed that the average error of the calibration experiment was 0.31 pixels,indicating a high calibration accuracy.(2)In terms of stereo matching algorithms,this article proposes improvement ideas based on the AD Census stereo matching algorithm.In the cost calculation stage,in response to the problem that the Census transform relies too much on the grayscale value of the central pixel,a fusion of AD transform,improved Census transform,and gradient transform is proposed as the matching cost to increase the robustness of the algorithm to noise interference and lighting effects.In the cost aggregation stage,to address the issue of large matching errors in image edge regions,gradient information is introduced as a constraint condition to improve the cross window construction method and improve the matching performance of the algorithm in edge and weak texture regions.Through stereo matching experiments,the disparity map obtained by this algorithm has smoother edges,clearer contours,significantly reduced error matching rates in edge regions,and improved matching performance in weak texture regions.(3)Conduct three-dimensional reconstruction experiments on the gearbox shell,design the overall process of the three-dimensional reconstruction system,and build a hardware platform for the three-dimensional reconstruction system.Perform three-dimensional correction processing on the collected gearbox shell images,use the improved stereo matching algorithm in this article to solve the disparity map of the gearbox shell,and obtain the three-dimensional coordinate information of the gearbox shell surface based on the principle of triangulation,completing the three-dimensional reconstruction and measurement of the gearbox shell.The experimental results show that the reconstruction effect is good and the measurement accuracy is high.