Research On Three-dimensional Reconstruction Of Shoe Mold Based On Binocular Vision

In traditional shoe mold automation processing,traditional industrial robots can only perform the technical operations like cutting,sanding and brushing on the bottom of the shoe mold according to a fixed template due to the lack of access to external information.However,such a unified template processing method ignore the problem of positional deviation of the shoe mold during the initial fixing and the possibility of the position change during the transmission process,which easily leads to defective products and affects the production efficiency and the economic benefits.Aiming at solving these problems,the binocular stereo vision was used to be the basis of the 3D reconstruction and the image acquisition,camera calibration image preprocessing,stereo matching and the 3D reconstruction in binocular stereo vision were studied.Furthermore,the 3D reconstruction of the bottom of the shoe mold is achieved by the binocular vision system built in this article.The main work of this article is as follow:(1)The camera imaging model and the camera calibration methods were studied to solve the mapping relationship between three-dimensional object points and two-dimensional planar image points.The cameras were calibrated with the method of the Zhang Zhengyou calibration method and the results of the calibration were evaluated by the recalibration error,witch verified the accuracy of the calibration.(2)The fundamental theory of the binocular stereo vision was analyzed and the feasibility of the 3D reconstruction was verified theoretically.Then the binocular camera calibration,polar constraints and stereo correction in the binocular vision system were studied.Besides,the calibration and correction of the binocular vision system were completed.(3)The fundamental theory of the stereo matching algorithm was analyzed and the key steps involved in the stereo matching algorithm such as cost calculation,cost aggregation,disparity selection and optimization were studied.After analyzing the application background and the limitations of the SAD-based(Sum of Absolute Differences-based)fixed window stereo matching algorithm,the BT-based(Birchfield and Tomasi Dissimilarity-based)adaptive window stereo matching algorithm was used in this article and the effectiveness was also verified using the standard test images provided by the Middlebury University website and the PBM evaluation criteria.(4)The binocular vision system experiment platform was built through the Daheng CCD cameras and the computer.With this experiment platform,the image acquisition,binocular camera calibration,image preprocessing,stereo matching and 3D reconstruction were completed and the 3D reconstruction of the shoe mold based on binocular vision was realized.