Design Of Industrial Robot Sorting System Based On Machine Vision

With the advent of the industrial era 4.0, the robot is playing an increasingly important role in the industrial intelligence. There will be a heavy task in work-piece sorting production line. If manually sorting operations be applied, it will be not only low work efficiency but also higher costs. It is a new trend in the industrial pipeline that the machine vision technology be applied to robotics in order to execute intelligent sorting task. So,using machine vision algorithm to identify and track the target and to achieve fast sorting operations are of great theoretical significance and practical value.In this paper, the theoretical research and experimental verification based on machine vision sorting robot system will be carried out. The sorting system involves two core issues: one is image recognition and object tracking subsystem based on machine vision, the other is sorting strategy based on SCARA robot. Targets to be identified are the common mechanical work-piece, including bolts and nuts. Overall program of intelligent sorting operations be designed, elaborating the basic operating principle and verifying the feasibility of the sorting system through experiments. The main contents are shown as follow:Firstly, there is a brief introduction about the components of the visual sorting system including that describing the indicators of SCARA and the controller, analyzing principles of selecting the industrial cameras and optical lens, introducing the specific steps of camera calibration and eventually showing the experimental results of the camera calibration.Secondly, the tracking problem of the mechanical parts on the conveyor belt is analyzed. Identifying and tracking targets is the premise of the sorting operation. There are many ways to target tracking at present. However, the work-piece tracking algorithm based on Cam Shift is applied to visual sorting system. The basic principle and main implementation steps of the algorithm are introduced in detail. The basic process of continuous tracking target work-piece is analyzed. Then, the experiment proves that the tracking is effective.Thirdly, the identification method for three different parts is studied. Because the recognition of the target images is based on some of the features of work-pieces, so the moment feature of image combined with the label of the work-piece is used as an important sample to train BP neural network. Adaboost algorithm is introduced to improve the performance and the recognition accuracy of neural network. The results show that the improved classifier is effective in target recognition based on the experimental data analysis.Finally, the control strategy of the manipulator is studied. The SCARA robot is modeled by the D-H parameter method and the inverse solution of robot kinematics is analyzed. After that, the relationship between the picking positions and the rotation angles of the manipulator is analyzed based on the optimal control. The basic principles of selecting the optimal grasping points are obtained. Finally, the sorting operation experiment is finished by using the visual sorting platform.