Research On Workpiece Recognition,Positioning And Grabbing System Based On Binocular Vision

In the modern production line,the identification and grasping of workpieces are important links,and are the basis of subsequent sorting,assembly and other operations.The binocular vision technology can identify the type of workpiece and determine the position and posture of the workpiece.The combination of robot and binocular vision,which makes industrial robots have perception ability,improves the level of automation and production efficiency of the production line.This paper proposes a workpiece recognition positioning and grasping system based on binocular vision,which can improve the automation of workpiece grasping in the production line.In this paper,the overall scheme of the workpiece recognition positioning and grasping system based on binocular vision is proposed,including the vision module and the robot module.The hardware is selected and the software environment is explained.Analyze and verify the commonly used image filtering methods,choose bilateral filtering to process the workpiece image,and use Gamma correction to improve the image contrast.According to the real-time requirements of workpiece recognition and higher rotation invariance,an improved ORB algorithm is proposed.In the improved ORB algorithm,the gradient direction histogram method is used instead of the gray centroid method to calculate the main direction of the feature point,and the FREAK descriptor is used instead of the rBRIRF descriptor to describe the feature point information,so that the algorithm has better rotation invariance and robustness.Experiments have verified the superiority of this algorithm in the extraction of workpiece feature points compared to common algorithms.The template matching of the feature points extracted by this algorithm is used to identify the type of workpiece,and it is verified under the conditions of illumination changes,rotation changes,and increased noise.Aiming at the positioning of the workpiece in the production line,a workpiece positioning method with local fitting of orthogonal function is proposed.Before the workpiece positioning,an image of a fitting plate was acquisited and processed,and the position information was obtained as the sample points.The image of the workpiece is collected and preprocessed,and the pixel centroids of the workpiece is obtained by feature extraction.The plane coordinates is obtained by local fitting of orthogonal function.The space coordinates of the workpiece is obtained by binocular reconstruction.The experimental results show that the method has better accuracy,avoids the calibration link,and has good application value through comparing with the traditional positioning method.In order to accurately determine the position state of the workpiece,the minimum posture rectangle is used to calculate the posture of the workpiece.The kinematics analysis is carried out for the robot to grasp the workpiece,and the joint angle and displacement of the robot are calculated according to the workpiece position.The kinematics analysis is verified by simulation.The software system is designed,including image acquisition,image pre-processing,image recognition,image positioning and inverse kinematics modules.The host computer transmits the signal to the motion controller to control the robot to grab the workpiece.