| Assembly is an important part of production.According to statistics,it accounts for about 53% of total production time.And Grasping is the key technology for assembly.In order to increase production efficiency and meet the growing need for flexible production,3D machine vision technology is applied to Robot Grasping will become very important.The research of this topic is to use 3D vision technology to realize the robot's flexible Grasping,use 3D machine vision to obtain the workpiece information,and then the robot realizes the Grasping work according to the information feedbacked by the 3D vision technology.The main research content includes:Firstly,the Eye-to-Hand camera installation adopting the triangulation principle and the robot 3D vision control system based on position feedback were designed.Secondly,the whole system of robot-vision-conveyor was analyzed,and a visual-based non-contact TCP calibration method was proposed.A hand-eye calibration based on a standard sphere center was proposed,and the gradient descent method was used to accurately determine the hand-eye calibration parameters.Calibration accuracy is higher than traditional methods.Thirdly,it realized the recognition and positioning of target parts based on CAD model.Firstly,the point cloud was preprocessed and the point cloud segmentation based on regional growing points was performed;then RANSAC-based pose coarse positioning was performed,and finally the target was obtained by ICP algorithm.The precise positioning of the point cloud,compared with the coarse positioning,The average score of correct matching points increased from 13.79% to 14.87%.Fourth,the overall process planning based on engineering applications was proposed,and robot 3D vision capture software was developed.Finally,based on the prototype platform of the 3D visual grasping principle,experiments and analysis were carried out.the average time for each grasping was 9.18 seconds,and the capture success rate was 98%. |
PDF Full Text Request