Abstract: In open arc welding process, it is difficult for the common vision system to achieve clear images due to the arc interference. In order to get high-quality images, a vision system using a digital camera and a welding glass was proposed. Through the image fusion of multi-exposure image sets, the weld pool area and the neighboring dark areas were both clear. A software-based method for capturing multi-exposure image sets was investigated by using software trigger and online change of the exposure parameters. A fusion algorithm accumulating images after gamma transformation was used to obtain images with enhancement of shadow detail and removing the false edge. The fusion method was shown to be rational and effective by using the approximate radiometric response curve. The results show that a 20FPS frame rate could be obtained, and the images are able to meet the needs of remote monitoring, which provided sufficient information for intelligent welding.