| With the rapid development of intelligent manufacturing,machine vision technology represented by binocular stereo vision has gradually become an important research direction of intelligent upgrading.At present,the technology of manufacturing and parts assembly based on off-line control of robots is not flexible enough,and the working efficiency is low,which makes the existing resources cannot be fully utilized.Therefore,combining vision technology with robot technology can better solve this problem.In vision technology,binocular vision can not only make up for the shortcomings of monocular vision,but also more vision,binocular vision integration is simple and low cost.Binocular vision has a high degree of fit in intelligent assembly positioning,which can further improve the precision of intelligent assembly and the degree of automation of industrial production.On this basis,this topic intends to study the key technologies of binocular vision based intelligent assembly recognition and positioning.The research contents are as follows:According to the theory of imaging and positioning,the binocular vision imaging model and structure model are constructed.The binocular parallel structure model is selected as the structure model of this paper,and the algorithm module of the positioning system is analyzed.After that,the principle of Zhang Zhengyou’s plane calibration method was studied,and the binocular vision calibration platform was built to calibrate the camera,and the stereo correction of the binocular camera image was carried out according to the parameters inside and outside the camera obtained from the calibration experiment.Image binarization is used for image processing to obtain the target information in the image.Then,according to the advantages and disadvantages of different categories of features,point feature extraction algorithms are selected for research.ORB feature detection algorithm and SURF feature detection algorithm are mainly studied,and the feature detection effects of these two algorithms are compared and analyzed in each case.It is determined that SURF algorithm has a high degree of fit and applicability to the research object in this paper.Based on the research and analysis of feature stereo matching algorithms,a parallel binocular vision model based on SURF algorithm was proposed in this paper.The mismatching was optimized for the first time through sequential consistency constraints and similarity constraints,and then the mismatching was optimized again by RANSAC.Experimental results showed that the improved matching algorithm improved the accuracy of feature point matching.Make the target location more accurate.The parallel binocular vision experiment platform was built,the positioning system of intelligent assembly was experimentally demonstrated,the composition of each module of the binocular vision positioning system was analyzed,the hardware integration and program design of each module were carried out,the preliminary three-dimensional coordinate information of the assembly parts was calculated according to the principle of preliminary acquisition of threedimensional coordinates,and the assembly positioning experiment of parts was designed.The error between the initial three-dimensional coordinate of the component and the actual coordinate of the object is within 0.5mm.Finally,the error of binocular vision system is analyzed and the reprojection error is defined as the error formula model.Aiming at the model,a nonlinear optimization method is proposed to reduce the reprojection error to the greatest extent,so as to achieve more accurate target positioning.The experimental results show that the positioning accuracy of the three-dimensional coordinates of the preliminary positioning is improved to within 0.3mm by nonlinear optimization method. |
PDF Full Text Request