Object Tracking In Robot Fast Teaching Based On 3D Vision

At present,with the appearance and development of intelligent factory concept,manufacturers of industrial robot programming also had higher levels of demand,how to reduce the production cost is an issue they care most about,and use the offline programming or with the aid of teaching implement drag-and-drop teaching robot programming methods can't meet the personalized requirements programming of flexible and efficient.At the same time,the rapid development and low cost of computer vision technology provide a good opportunity for this problem.Using vision to guide robots to quickly complete programming tasks has become another hot topic in the research field.In this thesis,a method of rapid robot teaching programming is designed by tracking the target object.Firstly,the point cloud data with color information generated by a single Kinect camera is initialized at the first frame,and the segment target object is used as a reference template for subsequent tracking processing.Then search the area suspected to be the target object within the line of sight,use Octree algorithm to detect the motion changes of each object,use particle filter algorithm to track the moving object in real time,through multiple detection and update template to obtain the motion data of each target object;Finally,these data are taken as the constraint quantity of the robot's movement.After the coordinate system transformation and the path planning of Bi TRRT planning algorithm,the UR5 robot is driven to reproduce the previously demonstrated tasks,and the rapid teaching of the robot is finally completed.Through experimental examples,the tracking algorithm proposed in this thesis will not affect the real-time performance of the system due to the increase of the number of target objects,and can meet the real-time computing needs.And the detection results will not be affected when other dynamic disturbances occur during the detection of the target object motion.In the same type of application scenarios,even if the target object is changed or the number of target objects is increased or decreased,the robot can still be flexibly endowed with new skills through observation and learning.More importantly,the teaching efficiency of this method is about the traditional drag-and-drop teaching method,so it is more practical.