Research On Visual Tracking Weld And Control System Of Robot

At present,China has entered the golden era of high-speed railway construction.The surface of the rail vehicle body is usually welded by splicing separated parts,in the welding process,the welding area will produce a large welding stress.The stress will decrease the strength and the fatigue resistance of the welded parts.In order to eliminate the welding stress and provide high-quality welding surface for the following process,the welding seam generally needs to be grinded and polished to eliminate the internal stress after welding.Currently,the majority of large-scale structural parts' welded seam are grinded and polished manually by workers,which is low production efficiency,labor intensity and high technical level for the works,at the same time,the accuracy of consistency of the surface is difficult to guarantee and the basic surface is even to be worn and torn.In this paper,based on the mobile grinding and polishing robot mechanical body,develops a robot based on visual servo control system of position,which can achieve the weld tracking,three-dimensional measurement of the weld feature point and robot automatic operation control function of the robot.The main contents of this paper include:(1)The binocular vision measurement system platform of robot is built,and the binocular camera stereo calibration is carried out.Then a three dimension coordinate vision measurement model of feature points is established.To realize the robot tool system navigation,the feature point coordinates need to convert into the robot coordinate system.A calibration method of hand eye transformation relation based on three target points is designed,which can be used to obtain the robot hand eye transformation matrix easily and quickly.(2)To eliminate noise interference and obtain high quality welding seam image,the filter and filtering algorithm is adopted to do image processing.A method based on gray threshold is proposed,which can realize to extract the laser line region.To fill the defect and smooth boundary of the laser line,the extracted laser region is processed by morphological processing.To solve the problem of extracting the centerline of the laser line,the region skeleton algorithm is applied,which can obtain accurately and smoothly center line and avoid the problem of central line discontinuity.To improve the matching speed of the binocular feature points,the search method of weld feature points is improved by searching along with the center line.The three dimension coordinates of the searched matching feature points are reconstructed to realize the robotic function of recognition,tracking and measurement of the weld.(3)The structure of the robot control system is designed,and the two level distributed structure of the upper and lower computer is determined.As the upper computer,PC is responsible for the entire robot system management and motive trajectory calculation,image processing,and so on.The lower computer is composed of a motion control card and an image acquisition card,which can be used to control robot motion control system and vision system acquisition,respectively.This structure makes the control system work speed and control performance improved obviously.The multi-mode control strategy of the robot is designed,and the robot control software is programmed.The software has the function of parameter setting,manual mode and automatic mode control.Experiments showed that the vision servo control system of the robot had certain accuracy,stable,and reliable function,was able to provide a strong support for polishing technological parameters.