Feature Recognition And Accurate Messurement Of Welding Seam

This dissertation is based on the practical engineering projects. It studies the methodsof welding seam image feature recognition and image processing in the industries whichrequire the welding process, including the process of welding the safety shell of nuclearstation, helix energy-exchange board and tube-plate energy exchange board, etc. It uses thewelding robotics system to capture and process the practical welding seam images, and thenreturns the processed result to the robotics system, guiding the robot end effector tocomplete the process of real-time tracking and welding.This dissertation mainly uses digital image processing technology and computer visiontechnology to process the image of welding seam, including median filter, thresholdprocessing and morphological algorithm, etc. Some other algorithms are used to process thespecial welding seam images, such as sliding window searching method, clusteringalgorithm, breadth first search algorithm and convex hull algorithm, etc.In the process of practical welding process, arc light splash and reflection light willalways appear, this dissertation describes two methods to overcome such interferences, oneis using physical method to insulate the welding seam image, another one is using imageprocessing technology. The physical method contains adding the light-blocked board, filterand polarization glasses, etc. The image processing method contains using the robustalgorithm which is insensitive to the local interferences to process the image, such astemplate matching algorithm.The software system in the experiment is programmed using VC++6.0and OpenCV1.0, with the API of CG-Card which is designed by DaHeng image company and the API ofDMC which is designed by Leadshine technology company. After capturing images, CCDsends the images into the industry computer through the video line, then the industrycomputer processes the images and gets the difference between the welding gun and thewelding seam, sending the difference need to be adjusted into DMC and using DMC tocontrol the robot end effector for real-time seam tracking.It proves the high robustness of the image processing method introduced by thisdissertation and improves the efficiency in the process of welding.