Direct Imaging Based Seam Tracking for Welding Control

In this paper a method for seam tracking in automated welding systems using direct imaging with advanced signal processing rather than optical filtering is presented. Optimal algorithms for prefiltering, edge detection and segmentation have been searched for. A variation of the FIR median hybrid (FMH) filter proved to be the best solution for the prefiltering task. Edge detection is performed using Sobel 3times3 convolution mask. A new variation of Hough transform called angle limited Hough transform is presented for segmentation. The results show that direct imaging with advanced image processing can provide accurate and reliable measurement of groove edges. The developed system requires less hardware than traditional laser stripe triangulation systems and therefore reduces the cost of automatic seam tracking.