Research On Calibration Method Of Binocular Vision System

Machine vision detection mainly collects the image of the object to be tested,in order to realize the detection and measurement of the object to be measured,the feature extraction and analysis calculation are carried out.Camera calibration is the basis of machine vision detection,and the accuracy of calibration has an important effect on the accuracy of the whole measuring system.At present,the research group carries on the quality inspection research for the commodity barcode on the curved surface packaging.Because the object to be measured is a space curved surface structure,there is a visual distortion.At present,there is no effective equipment and technology to complete the detection.Therefore,the research group uses binocular vision system for detection.According to the spatial relationship between the binocular images,the image of the barcode region is reconstructed,and then the feature extraction and analysis are carried out to complete the detection.This paper is an important part of the research.The task is to study the calibration method of binocular vision system.The first purpose is to obtain the internal and external parameters of the system,determine the transformation relationship between the image and the real world,and realize the measurement of the actual data from the image.The second purpose is to reduce the influence of lens distortion and improve the detection accuracy.The main work of this paper includes the following four parts:First,construct binocular vision system.Combined with the actual situation of the target to be tested,the design scheme of binocular vision system is made.Complete the space structure design of the left and right cameras,and complete the selection of cameras,lenses and other hardware equipment according to the detection requirements,and build a binocular acquisition system.Second,single camera calibration.According to the size of the acquisition window,the appropriate size of the checkerboard target image is designed.Firstly,the left and right target images are obtained by the acquisition device.Then feature extraction was carried out,and a method combining image morphology and minimum convex hull was designed to extract the corner points of checkerboard,and subpixel processing is performed on corner points.Finally,the calibration method based on the planar template method is used to determine the left and right cameras respectively to obtain the internal and external parameters.Third,stereo calibration.The stereo calibration is completed according to the relative position of the left and right camera coordinate systems.The spatial pose of the right camera relative to the left camera is calculated,that is,the rotation matrix and the translation vector between the two cameras are obtained.It provides a basis for the subsequent 3D reconstruction of the cambered barcode.Fourth,error analysis and compensation.The lens distortion model is used to correct the image distortion.The calibration parameters of the left and right cameras are optimized by using the corner reprojection error to construct the optimization function.According to the image coordinates of the corner on the left and right images,the three-dimensional coordinates of the corner are reconstructed,and the error function is established based on the offset between the actual physical coordinates and the reconstructed coordinates to correct the calibration parameters.Based on the fact that the row and column of chessboard keep linear consistency,the nonlinear offset of reconstruction corner is counted,the optimization model is established to further optimize the parameters.In this paper,the design and development is carried out under the environment of Visual Studio.The calibration of binocular camera is completed,and the calibration parameters are combined with the bar code detection algorithm.A total of 600 target images are collected by the binocular acquisition device for testing,and the method in this paper is compared with the traditional method.The accuracy of the method is evaluated by the reprojection error and the reconstruction point offset error.The results are better than the traditional method.It is proved that the corner detection algorithm and error compensation can effectively improve the calibration accuracy of binocular camera.The quality test was carried out by selecting 40 barcodes on the outer packaging of curved surface commodities.The results show that the improvement of binocular camera calibration accuracy can effectively improve the detection accuracy of the system.