A vision sensing system for taking and processing the image of MIG welding pool of aluminum alloy has been setup in this paper. The morphological operator which can effectively wipe off noise and 'cathode pulverization' area in image was used to process the image of the aluminum alloy MIG welding pool. Based on the step response experiment, the welding pool width model of dynamic process in aluminum alloy pulsed MIG welding is identified by least square method. The input parameters of the model are welding wire speed, base current and pulse duty ratio, and the output is beam width of welding pool. The influence of wire speed, base current and pulse duty ratio on welding pool width is analyzed, which provides the theoretical basis to realize the intelligent control for aluminum alloy pulsed MIG welding process.