This paper proposes an image sensing system for the TIG (Tungsten Inert Gas) welding process of aluminum pipe. In automatic welding process, it is important to recognize the molten pool during the welding process. The use of CCD camera as vision sensor obtains the information of molten pool. The research was conducted for welding of aluminum alloy Al6063S-T5 with controlled welding speed and the AC welding polarity. Firstly, histogram analysis and range of brightness threshold was established. Secondly, by using the differential value of brightness of molten pool, the edge of molten pool can be detected. Finally, to overcome the unsteady detected edge, the ellipse approximation was used to obtain the image parameters, which are width, length and area. From the experiments it is shown that the recognition of molten pool using vision sensor is an effective method.