Design And Implementation Of Part Detection Experiment System Based On Machine Vision

With the rapid development of computer technology,machine vision has been widely used in industrial fields.The application of machine vision has greatly improved the efficiency of industrial inspection,and has played a great role in promoting the automation and intellectualization of industry.This thesis establishes an experimental system to study the industrial application of visual inspection algorithms.The experimental system can provide a experimental platform for verification of visual inspection algorithm,and can also study the coordinated control of optoelectronics and electronics in industrial production.The main contents of this thesis are as follows:Firstly,the overall structure and hardware construction of the part inspection experiment system based on machine vision are introduced.The hardware part of the experimental system consists of three parts: the detection module,the execution control module and the mechanical module.According to the functional requirements and design requirements of the experimental system,the hardware is selected and the installation of hardware equipment is designed.Secondly,to solve the problem of category detection caused by part orientation change,this thesis applies the sparse representation algorithm to the category detection of the part with arbitrary orientation.By establishing the Gabor feature dictionary of the part and solving the sparse reconstruction solution according to the feature vector of the detected part image,the category label of the part can be then obtained.Compared with other similar algorithms,the proposed algorithm has better accuracy and efficiency.Thirdly,aiming at the reconstruction of three-dimensional(3D)point clouds of the part,this thesis adopts the binocular vision system paradigm to obtain the part's images in two different orientations,and uses the method of feature point matching to reconstruct 3D point clouds.According to the two part's images,the feature points are extracted and matched.According to the matched point pairs,the coordinatetransformation matrix of part can be solved.Then the space position of part feature points is restored according to the coordinate transformation matrix,and the 3D point cloud can be obtained.The validity of the proposed method is verified by reconstructing the 3D point cloud of the typical parts.Finally,the part detection method studied is implemented on the experimental system and the performance of the proposed algorithms is also verified.The visual control interface of the system is established,and the visual inspection algorithms are combined with the functional modules of the designed system,including the application of sparse representation algorithm in the part category detection and sorting module and the part assembly detection module.The traditional binocular vision system realizes the 3D point cloud reconstruction of part systematically.Based on the realization of the system,several experiments were carried out on the experimental system,and the practical application of the algorithm was further conducted.