A Quick Algorithm to Track Welding Line Based on Computer Vision

In this paper, we propose an algorithm to track welding line of TiG welding in welding pipe or tube. The proposed algorithm consists of detecting edge, binarization, segmenting target area, locating welding torch and locating welding line. Firstly, we use prewitt operator to form the edge of image which is acquired from industrial camera. Secondly, we achieve the binary image by using maximal variance between-class, then segment the image based on region growing and get the target area including welding torch, welding pool and welding line. Finally, the center of welding torch and welding line is computed and the difference of them is used to control the action of motor to track welding line. Experimental results indicate that all steps of the proposed algorithm consume less than 80 ms while location is accurate, and the proposed real-time algorithm can be used for TiG welding robot in practice.