Abstract: For welding process information sensing, the visual information of weld pool in aluminum alloy tungsten arc welding was investigated.By CCD optical sensor designed independently, the weld pool image was observed under arc light.After recovering and smoothing the original degraded image, the recovered image show more characters for next processing steps.Series of algorithm including automatic threshold segmentation, noise removing based on 8-neighbourhoods area threshold, total points statistics, were performed to calculate gap area in the image coordinate system.After thinning, Hough transformation was used to fit the gap points.The fitted line direction is the welding direction.With the gap area and the calibration result before welding, the gap size can be obtained.