Research On Following Technology Of Planar Motion Robot Based On Monocular Vision

In the palletizing industry,a planar 2-DOF redundant-driven parallel robot is taken as the research object.The vision system is fused with it.Through the research of image detection,image recognition and vision positioning,the moving target tracking task is completed under the guidance of vision,so that the terminal of the parallel robot can track the goods waiting for palletizing on the assembly line,so as to facilitate the following actions such as grabbing and palletizing.Based on this background,this paper studies the kinematics,error,calibration,control strategy,camera calibration,image processing and moving object tracking experiments of the robot.The main research contents are as follows:Firstly,the kinematics inverse solution,velocity and acceleration of planar 2-DOF redundant parallel robot are analyzed.On this basis,the error model of the robot is established based on the loop increment method,and the error sensitivity of the robot is analyzed.Secondly,the industrial camera model is established based on pinhole model theory,and then the camera model is calibrated by Zhang's method.Aiming at the characteristics that the robot can only move on the plane,the camera model is simplified and the simplified camera model is established.On the basis of establishing the camera model,on the one hand,the static image is preprocessed based on Gauss filter,and then the maximum class is used separately.The variance method and sub-pixel edge detection technology are used to double segment the image.Finally,the H-channel of HSV color space is used to recognize the image,which lays the foundation for the experiment of robot calibration using vision.On the other hand,the Camshift algorithm based on SURF is used to recognize and capture moving objects,which lays the foundation for the following experiment.Thirdly,the servo control model of AC synchronous motor is established;the parameters of the three-loop controller in the servo control model of AC synchronous motor are tuned by using RBF neural network,and the advantages of the control strategy are verified by comparing the parameters of the three-loop controller with the parameters of PID manually adjusted.A hybrid force/position control strategy based on calculating moment is designed,and the dynamic model of the robot is established by using Newton Euler method.The minimum two-norm is used to optimize the driving force.Simulink is used to simulate the driving moment curve of each joint after optimization.Finally,a control experiment platform is established based on Beckoff controller and prototype;the error calibration model of the prototype is established;the kinematics calibration experiment of the prototype is carried out to compensate the geometric error of the robot,and the change of the output accuracy of the end of the mechanism before and after calibration is analyzed.The experimental results show that the output accuracy of the end of the robot is increased three times after calibration;on the basis of the completion of the calibration of the robot,the visual-based In the sensory tracking experiment,the target trajectory simulates the complex cargo trajectory on the pipeline,so that the end of the robot moves with the target.By analyzing the tracking curve,it can be seen that the tracking error does not exceed 2 mm.