Vision Based Identification and Detection of Initial, Mid and End Points of Weld Seams Path in Butt-Welding Joint using Point Detector Methods

Abstract

The main challenge in using welding robot is the time taken to program robot path for a new job in low to medium volume manufacturing industries or repair work. Because of that, it is cheaper and efficient to weld the part manually. This project intends to identify and detect the initial, mid and end point of weld seams path in straight line joints of stainless steel piecework which are typical to welding applications. The image piecework is snapped by charge coupled device (CCD) camera which is perpendicular between piecework. Weld seams path identification method is implemented in three stages; (1) preprocessing (2) reduced domain and (3) the weld seams path is identified. The point detection techniques is used to find the point of the images. Point detection techniques such as Harris Binominal, Harris, Lepetit and Sojka points were analyed to remove those points that does not belong to seam path. The experimental results show this systems can identify and detect weld seam path location in terms of x-y pixels coordinates. Result show that, Harris points detector is the best point detector compared to the other detectors which the identification error is around ±2.0 pixels for both coordinates which is row and column in starting and ending points of weld seams path.