Research On Position-based Visual Control System For Industrial Robot

At present,the manufacturing industry is developing rapidly in the direction of intelligentization,putting higher demands on the interaction between the robot and the environment.The visual system has drawn much attention due to its rich visual information and high reliability.Robot vision servo system has become one of the most popular research fields in robotics in recent 20 years.In this paper,an industrial IRB1200 six-degree-of-freedom robot is selected as the research object.The Eye-in-hand model is used to build a visual servo robot system.A hand-eye calibration method based on geometric mapping is proposed to calibrate the hand-eye parameters.The method of target recognition based on Holistic-Nested Edge Detection(HED)combined with Canny operator is proposed to fulfill the task of robot accurately grasping and placing the target workpiece.The details are as follows:1?the development of robot experimental platform.Set up the slave computer,server,client architecture mode,easy to use the platform-independent high-level language for rapid secondary development or remote operation of the robot.2?the calibration of robot vision system.The internal parameters of camera is calibrated through Matlab calibration toolbox.The calibration of rotation matrix of camera external parameters can be realized by controlling the manipulator to do orthogonal translation motion.For the calibration of translation vector of camera external parameters,a calibration method based on geometric drawing is proposed.The geometrical drawing method is used to extract the world coordinate corresponding to the projection point of the end center of the robot on the working platform to determine the feature point and find the camera external parameter translation vector according to the determined feature point.Finally,the accuracy of the visual system calibration parameters is verified by measuring the distance of known line segments in the scene.3?Robot grasping based on deep learning.A novel method based on Holistic-Nested Edge Detection(HED)combined with Canny operator is proposed.HED extracts the gray contour of the target object,Canny operator extracts the binarized contour of the target object,finds the target workpiece contour after the template matching,estimates the pose of the target workpiece by the moment information and the minimum enclosing rectangle,workpiece velocity were measured by frame difference method,finally the workpiece position information by hand-eye relationship transform to the base coordinate system to guide the robot to grab.4?Implementation of position based visual servoing control system for industrial robot.This system is based on the OpenCV library,Caffe library and Visual Studio 2013 provides the MFC class library and Windows Forms to achieve client and server of human-machine interface design,client include robot control module,the speed detection module,workpiece positioning fetching placed module,image processing module,network communication module,etc.The service includes robot connection module,network communication module and so on.Finally,the verification system captures the success rate of the workpiece and the accuracy of the placement of the workpiece to the target position.