Vision Based Weld Seam Extraction & Parameter Estimation in GTAW

Abstract

TIG welding (GTAW) is one of the most used welding application in shipyard industries, automobile industries, manufacturing sector and so on. It has been the challenge to automate the welding processes to meet the requirement in a given time frame, this study emphasizes on vision system application for tracking the weld seam in mobile welding robot set up. Real time images of the work-piece is captured as target image and current image, target image or reference image is fixed and current image will be changing as robot moves. Our interest of focus on the image is seam edges, so we extract the points from both the edges. Here, we have used SIFT algorithm (scale invariant feature transform) which is invarianttto imageescale, rotationand translation. SIFT algorithm is used tooextract theefeatures fromtthe twooimages and subsequently extracted features are matched. It has four steps – scaleespace peakkselection, key-point localization, orientationnhistogram and key-point descriptor. Once the features are matched, homography matrix from the two real images using four matched points are estimated using DLT algorithm to validate the matching. Once we get the extracted seam edge points, we apply the linear regression for the obtained points on the first edge and second edge. We get the two equation of lines and then equation of centreline is obtained easily. And that is our desired seam trajectory. Homography can be defined as the relation between the two image of the same scene in space, it is also related to the camera motion. Homography element is used to establish the control law based on state space equation. Finally, process parameters like depth of penetration, welding speed, deposition rate, heat input rate are estimated and a real time TIG welding is performed.