Apparent defect recognition of gas metal arc welding based on infrared vision

The visual monitoring of welding process and welding defect identification are vital to the intelligent control of welding processes. In this paper, a mid-wave infrared CCD was used to acquire the infrared images of the welding pool on-line during the welding process. The images captured are preprocessed by the improved filtering algorithm and image enhancement algorithm. To obtain the temperature distribution information of the welding pool, the relationship between the gray value in infrared image and the temperature is established based on temperature calibration of the used thermocouple. The improved edge extraction algorithm is used to extract the characteristic parameters of the welding pool. Then the identification algorithm of welding defect is developed. The results of verified experiments show that the proposed algorithm has good practicability and accuracy in the on-line identification of welding shape, burn-through and unmelted defects.