Research On Industrial Robot Sorting Based On Machine Vision

With the development of industry 4.0,the production mode of enterprises is changing from traditional mode to automation and intelligence mode.Using industrial robot to automatically grab workpiece can save industrial cost and improve working efficiency.It is an important research field to combine industrial robot with industrial camera for object sorting.In this paper,an industrial robot recognition grab system is established based on deep learning and machine vision.The industrial robot control system can obtain the category,position and posture of the target object from the outside world.According to the geometric parameters of the solid robot,the robot motion simulation platform is built,the kinematics model of the industrial robot is established,and the solid robot is controlled by the file transmission module.YOLO-V3 target recognition algorithm is used to obtain the category and position of the target object.The VOC data set needed for algorithm training is made in the Labelimg calibration tool,and a stable and reliable target recognition model is obtained by training the deep learning network.input the images taken by the industrial camera into the network model to obtain the category of the target object and the position in the image.The binocular industrial camera is calibrated by using the Matlab camera calibration toolbox and the opencv camera calibration library function,the internal and external parameters of the camera are obtained,and the binocular camera model is established.By means of pole line constraint,the left and right images are stereo corrected,and the parallax map of the target object is obtained by BM(Boyer-Moore)stereo matching algorithm.According to the camera imaging principle,the 3D point cloud of the object in the camera coordinate system is obtained,and the obtained point cloud information is filtered and segmented.At the same time,an evaluation function of feature points is established to verify the feature points in each region.Obtain stable and reliable pose evaluation feature points.The pose of the workpiece in the camera coordinate system is established according to the acquired feature points.According to the structural parameters of solid industrial robot,the DH parameter table and forward and inverse kinematics function of robot are established.The trajectory planning of the end-effector motion of the industrial robot is carried out by five polynomials,and the kinematics of each joint angle of the industrial robot is simulated and controlled.The simulation platform is built based on the Matlab GUI module.The motion simulation platform is established according to the mechanism parameters of the solid robot to realize the motion simulation of the industrial robot model.A control platform is established by using the motion control API of the entity robot.Through the file transmission module,the pulse path file of the motion planning is passed into the control platform to realize the motion control of the entity robot.