Image processing method for automatic tracking of the weld line in plasma robotic welding

For automatic control in plasma welding (PAW), the welding situation must sometimes be observed in real-time. However, the welding quality of PAW depends on the performance of the tracking result. Therefore, to determine gaps in the welding, we developed an image processing method that automatically tracks the weld line. Since the plasma electrode is buried in the torch, visual judgment by the human eye is difficult. This method replaces human sight with a visible automatic profile control that exactly teaches the plasma torch during the welding process. The image processing method for automatic tracking adopts a hybrid optimization process with template matching, which identifies small parts of a welding characteristic image that match the template image. In general, template matching provides a partial quality control. In special cases, the welding gap and its boundary are effectively detected by a boundary detection method, and variable gaps detected in the real-time tracking are analyzed to ensure high-welding quality. In verification trials, our method demonstrated efficient welding work and effectively reduced the false torch guidance.