A Weld Recognition And Tracking System Of Welding Robot Based On Machine Vision

In recent years,machine vision technology has rapidly been integrated into industrial production.Its research and application in industry are increasingly valued,and remarkable results have been achieved in many industrial fields.Welding is one of the widely used technologies in the traditional industry.Traditionally,most of the welding robots use artificial teaching method to reproduce the work.This method has caused problems like large human force consumption,poor working environment,and the efficiency and quality of welding cannot be guaranteed.The growing maturity of machine vision technology provides sufficient possibilities to ensure the quality,efficiency,automation,and intelligence of the traditional welding robot welding.Therefore,the research of welding robot based on machine vision is of great significance.In this paper,the welding requirements of two typical welds,straight edge geometry welds and local deformation welding seams in industry,are taken as the direction of guidance and the method of recognizing workpiece welds using visual technology is studied.A welding robot identification and tracking system based on machine vision is designed.Firstly,the status of research on machine vision,weld seam tracking,and vision-based welding robots at home and abroad were introduced.Secondly,it introduces the framework of the system structure,including mechanical motion system,visual system,visual software and development environment.Later the article will introduce the vision system calibration theory,and use LabVIEW and Halcon to calibrate.And then it will explain the control of ABB robot and its Ethernet communication with IPC,and verify the accuracy of the calibrated system.Finally,the article gives two kinds of welding seam identification methods and system experimental results:1)For straight edge geometry weld workpiece,after the acquisition of image data,enhancement processing,smoothing processing,edge detection and sharpening processing are performed,and advantages and disadvantages of various methods are analyzed in the recognition and treatment of the weld.The Hough change method is used to detect the straight line in the expected region,and the weld path specific point of the fixed model workpiece is obtained by using geometric methods,such as obtaining the midline and obtaining the intersection point of the two straight lines.2)For the welds that are easily deformed locally,the article establishes the workpiece coordinate system by using the template matching to position the workpiece.In the relative expected region,the boundaries are detected with equal spacing and vertical boundary directions,and the boundary feature data arrays of multiple scales are acquired.The required weld path is obtained after selecting the path data and filtering the arrays of the them according to the characteristics of the welds.The experimental result shows that the application of the two schemes in the visual system is feasible,which can extend the scope of the application of machine vision in the field of welding,and provide an reference for other projects and researches.