Recognition and tracking of weld line by visual sensing system using artificial neural network

In order to develop a robust control system to detect a weld line and to track the weld line automatically by means of a welding robot with a visual sensor, a new control system applying the artificial neural network was constructed. Before the image taken was input to the neural network, the LOG filter was applied to the original image to enhance the weld line image, and then a binary image was generated. The binary image was input to the artificial neural network, and the weld line direction was detected. Then seam tracking was performed using the weld line direction. Because the neural network was used for detecting the weld line, a more robust control system was obtained. Moreover, as the result of seam tracking simulation, the applicability of the system to automatic welding was confirmed.