Research On Welding Robot Seam Correction Method Based On Active Vision

With the development of modern industrial technology,welding automation and intelligence have become the necessary support of modern industrial welding.Among them,the precision of welding automation is particularly important,but the workpiece placement deviation and thermal deformation during welding have a serious impact on the precision of welding positioning,thus affecting the quality of welding.This paper focuses on the key technologies involved in the welding robot weld correction method based on active vision,such as weld positioning and weld correction.The main works are as follows: complete weld positioning through image processing and feature point extraction,and modify weld trajectory deviation through offline and online correction methods on the basis of the original teaching trajectory,so as to improve the quality of welding.In order to solve the problem that arc light or smoke in the weld image collected by the structured light sensor interferes with the weld location,the method of image processing and feature point extraction is used to locate the weld.First,the location of the region of interest in the weld image was determined,and the laser fringe was smooched by median filtering,Otsu method threshold segmentation and morphological open and close operation.Then,the center line of laser fringe is extracted by gray barycenter method.Finally,according to the groove types of butt,lap and V-shaped welds,the linear scanning method,slope method and fitting intersection method were separately designed to extract the information of weld characteristic points and complete the location of weld.Aiming at the problem that the welding parts are placed with a large deviation,resulting in the loss of the weld track detected or the positioning deviation of the welding fixture,an off-line weld correction method based on curvature similarity improved ICP matching was proposed.First,the corresponding points of the weld and the corresponding points of the two groups of tracks were extracted by means of curvature similarity,and the initial matching was performed by principal component analysis according to the corresponding points.Then,the ICP algorithm is improved by combining the curvature and Euclidean distance to carry out fine matching.Finally,the feasibility and effectiveness of the method are verified by simulation and comparison experiment,and the off-line correction effect is good.In order to solve the problem that the thermal deformation of welding workpiece leads to deviation of the obtained deviation value,an online weld deviation correction method based on the combination of median filter and Kalman filter was proposed.First,according to the position of the structured light sensor,a linear chain queue is designed to store the prospective data.Then,the prospective data is removed by median filtering,the observed value of Kalman filtering is corrected,and the tracking prediction of Kalman filtering is carried out to complete the on-line correction of weld deviation.Finally,the stability and effectiveness of this method are verified by simulation experiments.In order to verify the effectiveness and stability of the welding robot’s weld correction system based on active vision,this paper adopts the arc weld of the butt groove to conduct the weld correction experiment.First,the hardware and software platform of the weld correction system is constructed,and the hand-eye calibration between the structured light sensor and the welding gun is completed.Then,experiments of weld off-line correction and weld on-line correction are designed respectively.Finally,through the experimental verification,the results show that after offline matching,the teaching track basically coincides with the weld track,which meets the accuracy of offline correction.The online correction error is controlled within 0.5mm,and the average correction error is about0.2mm,which meets the actual welding requirements.