An Adaptive Tracking of Weld Joints Using Active Contour Model in Arc Welding Processes

摘要

This paper presents a vision processing scheme to automatic weld joint tracking in robotic arc welding process. Particular attention is concentrated on its robustness against various optical disturbances, such as arc glares and weld spatters radiating from the melted weld pool. Underlying the developed vision processing is a kind of model-based pattern searching, which is necessarily accompanied by two separate stages of modeling and tracking. In the modeling stage, a syntactic approach is adopted to identify unknown weld joint structure. The joint profile identified in the mdoeling stage is used as a starting point for successive tracking of variations in the geometry of weld joint during welding, which is automatically achieved by an active contour model technology following feature-based template matching. The performance of the developed scheme is investigated through a series of practical welding experiments.