Research On Machine Vision And Muti-joint Manipulator Based Dynamic Target Recognition And Grasping

As one of the most important technologies in the field of intelligent manufacturing,machine vision has been widely introduced into the system of robot control in the process of industrial production,and improve the level of automation and intelligence.In recent years,dynamic targets' capture has gradually became a hot research topics in this field.Aiming at the issues of intelligent recognition and grasping of dynamic targets in industrial production,machine vision and muti-joint manipulator based dynamic target recognition and grasping system was constracted by means of technology of robot,machine vision and image processing technology in this dissertation.Firstly,the principle and method of machine vision are expounded,the model of pinhole camera is established,the binocular camera is calibrated by the perspective transformation,then the parameters of camera is derived and the accuracy calibration is verified by experiments,the results show that it has small calibrated error and can meet the demands of positioning.On this basis,the image processing algorithms are discussed.Hereinto,the Gussian filter method is used for image smoothing,the threshold transform is used for image enhancement,and the Canny operator is used for image edge extraction.Secondly,the kinematics model of GRB-3044 manipulator is established,based on the screw theory,the forward and inverse kinematics of manipulator are deduced,the equation between the end effector and each joints,and the workspace of the proposed manipulator is obtained,respectively.In addition,the chord mid-point based hough transform is used for targets' recognition and positioning,the experimental results show that the proposed algorithms are effective for multi-targets' recognition and positioning in complex situations.According to the 3D information of targets,the point for grasping is determined,and the targets are captured with the PTP control of manipulator and the operation of each axis.Thirdly,research on the dynamic grasp technology of the conveyor belt,the infinitely variable speed is adopted to control the conveyor belt.Then,the conveyor belt system is modeled and calibrated to determine the grasping position of the dynamic targets.Based on the dynamic trajectory of the targets,the control strategy establishing and trajectory planning of the manipulator are realized.The PID algorithm is used to monitor and grasp the dynamic targets,and the parameters are adjusted to enhance the dynamic response and grasping efficiency of targets.Finally,the experiment platform of dynamic targets' recognition and capture with machine vision is built,the software of system is developed based on the VC++,then,the messy targets are effectively recognized and captured,and the grasping efficiency of targets in different speeds is calculated.The results show that the designed system has some merits,e.g.high recognition accuracy,reliability and efficiency.