Research On The Technology Of Workpiece Positioning Based On Binocular Vision

With the development of science and technology,industrial robots have been widely used in manufacturing.Integration with vision systems is an effective method to improve the intelligence of industrial robots.Binocular vision systems,through which three-dimensional information can be obtained,have incomparable advantages over monocular vision systems,enabling industrial robots to perform complex tasks.In this thesis,research on the technology of workpiece positioning based on binocular vision is carried out.The main contents are as follows.Firstly,the background and the significance of the research are introduced.The current research status of both binocular vision robot systems and stereo matching algorithms is analysed,based on which the research content of this thesis is presented.After that,the requirements of the binocular vision based manipulator positioning system are analysed,the overall design of the system is given,and hardware of the system is selected accordingly.Secondly,the transformation of the coordinates systems of the binocular vision based manipulator positioning system is studied:the transformation from the world coordinates system to the camera coordinates system is realized by camera calibration,through which the intrinsic and extrinsic parameters of the camera can be obtained,and the transformation from the camera coordinates system to the end effector coordinates system is achieved by hand-eye calibration,through which the positional relationship between the camera and the end effector can be determined.Thirdly,datasets commonly used in the field of stereo vision are introduced.The fundamental principle of the traditional stereo matching algorithm SGM is analyzed,and the limitations of the algorithm are pointed out.Next,an end-to-end convolutional neural network for stereo matching is designed to achieve disparity estimation,which is able to overcome the disadvantages of traditional algorithms.The three-dimensionally reconstruction of the scene is achieved based on the result of disparity estimation.Finally,Hough circle detection is used to identify the workpiece,and then the target point clouds are segmented.The RANSAC algorithm is used to fit the upper plane of the workpiece to achieve pose estimation.Experimental results show that workpiece positioning algorithm of this thesis is able to meet the accuracy requirements.