Vision-based control of wire extension in GMA welding

This paper presents a vision-based approach for wire extension measurement and melt pool detection. The image analysis applies on the image sequences, which are extracted from a recorded welding video. The algorithms have been developed using the image processing combining profile analysis, intensity thresholding, region growing, filtering, and morphology. The goal is to use these computed parameters to help in welding control and defect examination. The software runs automatically and has been verified on various videos from different welding experiments. The resulting data are also represented and compared with chronological knowledge to improve the reliability of the system.