Abstract: Obtaining the dynamic characteristic information of molten pool and intelligent controlling in the welding process are focus of research in welding field.As to the TIG welding, machine vision technique is adopted to obtain the image of welding molten pool, and penetration quality controlling techniques in welding process are deeply researched.Mathematic morphologic theory is used to process welding image, and fuzzy inference and PID controlling theory are combined to form a whole image processing and closed-loop controlling system.In the controlling experiment, the abruptly-change Q235 steel workpieces of 2 mm thickness are used.The results indicated that the quality controlling system can conquer outside disturb in welding process effectively and ensure the stabilization of welding formation.