Research On The Vision Inspection System For The Dimension Of Automobile Gearbox Shift Fork

With the rapid development of the automobile industry,the research and application of vehicle manufacturing technology are becoming more and more extensive,and the demand for vehicle gearboxes is increasing.The shift fork is a key part of the gearbox,and its quality determines the quality of the gearbox.However,the traditional quality inspection of gearbox fork products still relies on manual and special inspection tools,which have disadvantages such as low detection accuracy,inability to quantify errors,large human error factors,and high labor costs,which cannot meet the production needs of enterprises.In response to the above problems,this paper combines machine vision algorithms to design a visual inspection system for gearbox shift fork size.The main contents are as follows:(1)In accordance with the structural characteristics of the shift fork and the detection requirements of the enterprise,a high-precision shift fork visual inspection platform was designed and built,the lighting scheme design and hardware selection were completed,and the high-definition shift fork contour image acquisition was realized.Aiming at the problem of shift fork coordinate system conversion and image distortion,this paper studies the camera imaging model and lens distortion principle,analyzes the camera calibration parameters and errors of Halcon and Matlab,the Halcon calibration assistant is used as the camera calibration tool to realize the distortion correction of gearbox shift fork.(2)Aiming at the problem of image preprocessing of the shift fork,the types of image noise,contrast enhancement principle and image segmentation algorithm are analyzed.The image bilateral filtering algorithm,linear contrast enhancement algorithm and OTSU image segmentation algorithm are selected to achieve the purpose of extracting the edge contour information of the fork image.(3)Aiming at the problem of edge contour extraction and fitting,the principles of Canny,Sobel and other edge detection operators are studied,the advantages of Hough transform and least square fitting algorithm are analyzed,and an improved least square fitting algorithm is proposed to realize and fit the extraction of fork edge contours.(4)Aiming at the problem that the fork image cannot directly detect the symmetry through the two stations of the system.The existing SURF,SIFT,ORB and other image stitching algorithms are studied,and the shortcomings of traditional image stitching algorithms and the characteristics of the fork image are analyzed.Based on the SURF algorithm,a local LoG-FAST image stitching algorithm is proposed.Experimental analysis shows that this method can effectively reduce invalid feature points,reduce the computation complexity of feature point description and matching calculation,and improve the calculation speed of the algorithm and the accuracy of matching pairs.The experimental results show that the precision,accuracy and operating efficiency of the image processing algorithm detection system can meet the design requirements by using the user interaction interface written by Visual Studio platform and the open source vision library of OpenCV,which verifies the feasibility of this system and has practical applications.