Research On Visual Recognition And Location Technology For Robotic Arm Grasping Applications

With the booming development of robot vision technology,the algorithm performance of robot vision has been gradually improved and the cost of vision hardware has been significantly reduced,making robot vision technology widely used in various industry fields.The traditional offline programming and online teaching methods of robotic arms can no longer adapt to the requirements of modern automation industry.By applying machine vision technology to robotic arm operation,the degree of automation and intelligent application of robotic arm can be improved in the industrial production.This project focus on the location problem caused by the irregularity of workpiece shape,and research on the recognition and location algorithms for robot grasping.The research works are as follows:Firstly,this paper have consulted a large number of relevant literature,and systematically summarized the research status of machine vision technology.The machine vision grasping system is designed,and the experiment platform is also built.The paper introduce the mechanism and the process of visual system calibration.Based on the calibration algorithm of Zhengyou Zhang for the monocular camera,the internal parameters of camera are solved,and the accuracy of internal parameters is analyzed.The mechanism of two robotic hand-eye calibration methods are analyzed,and the method of eye-to-hand system model is preferred.The posture transformation between the camera and the robot arm base is derived,and the transformation matrix of the hand-eye calibration is obtained,and the accuracy of the calibration results is verified and analyzed through experiments.The robot arm linkage parameters and linkage coordinate system were established by the standard D-H method,and the forward and inverse kinematics analysis of the six-degree-of-freedom collaborative robot arm was carried out to facilitate the subsequent control of the robot arm motion by the program.Secondly,the image noise is filtered by the method of median filtering.And then to complete the binarization segmentation of the image,the binarization segmentation is used to calculate the threshold of the image automatically.The paper improves the Canny edge detection algorithm.The detected target image edges have fewer spurious points,some pseudo-edges can be removed,and the target image edges are more continuous.The improved edge detection algorithm improves the accuracy of edge detection.After the target contour is detected,the Hough projection method is used to find the main axis of the target contour,and the minimum enclosing rectangle of the target workpiece is established to realize the recognition of the target workpiece and determine the location of the target image.Finally,the image coordinates of the target workpiece are transformed into the position coordinates of the actual robot arm movement.The grasping information is transferred into ROS(Robot Operating System).The robot arm is programmed to reach the designated grasping position and complete the workpiece grasping operation under the Move It platform.From the error of the experimental data,it can be seen that the result of the grasping position error does not exceed 2mm,and the error of the deflection angle of the workpiece does not exceed 2°.Therefore,the error accuracy of the image processing algorithm of this topic is good,and it can complete the positioning of irregularly shaped workpieces,which has practical engineering application value.