| With the application and popularity of robots in industrial production,in order to improve production efficiency and production quality,this paper takes the Double SCARA Robot to grab the workpiece on the conveyor belt as the working environment,and designs the target tracking algorithm and grab strategy based on vision.In this paper,the mapping relationships between the coordinate systems in the hand-eye system are investigated and experiments and analyses of camera calibration,robot inverse kinematic analysis and hand-eye calibration are carried out;the transformation relationships between the camera coordinate system,the conveyor belt coordinate system and the robot coordinate system are obtained,which paves the way for the subsequent use of robot gripping.Then,the GMM background model is established,and the target on the conveyor belt is detected by the background difference method based on GMM.A multi-target discrimination algorithm in the image was also designed to reduce the computational burden of the overall system.After that,the Kalman filter model is designed according to the actual situation,and the coordinates of the workpiece position outside the robot workspace are predicted,and the parameters in the Kalman filter are optimized.Aiming at moving multi-target grasping,a grasping strategy based on particle swarm optimization(PSO)is designed.By adjusting the weight ratio,the robot can not only realize the path optimization,but also ensure the robot to grasp smoothly as far as possible.By controlling the camera node and planning the speed of the conveyor belt,the optimization strategy of the system operation efficiency is formulated.In this paper,the circular gasket is taken as the test object,and the Double SCARA Robot is used to grab the gasket,and several methods are verified and analyzed.Through experiments,it can be obtained that the calibration accuracy of monocular camera is less than 0.013 mm,and the joint calibration of the whole hand-eye system,the overall grasping position error is less than 1.3mm.After target detection based on GMM background difference method,target prediction and accuracy verification are carried out through Kalman filter,and the average distance error is less than 0.98 mm,which meets the requirements of grasping.The comparison between the PSO algorithm and the traditional grasping strategy proves that the grasping strategy has faster efficiency through different group experiments. |
