Probabilistic approach to seam tracking of butt welding

The article presents a novel approach to the problem of arc seam tracking during the technological process of welding. A typical solution is using a laser sensor, which is very expensive and usually not ready to use just after removal from the box. Less commonly existing solutions are the applications of passive vision. In this technology an image is captured by a camera and then typically some gradient methods are used for finding the edges, usually after some additional preprocessing operations. The next phase can be the use of Hough transform for detection of straight lines. Unfortunately, due to the strong interference from the welding arc and the need of using some welding filters, the chance for proper recognition of the groove using gradient based methods is strongly limited. Therefore in the proposed approach the Monte Carlo method is used with some points randomly drawn. Then it is determined whether they meet the conditions set for the probable points on the path of welding which are always darker than the surroundings. Having drawn an adequate number of points the area of the welding groove and its position can be estimated with sufficient accuracy. The method presented in the paper has been designed for a specific application involving hard welding automation of long sections of butt joints.