Abstract: This paper presented a method to synchronously detect all-position weld bead geometry and the welding torch in lap joint GMAW with galvanized sheets, in which laser vision sensing was used. In this method a laser was used to simultaneously detect the weld bead and the welding torch, and the scale-invariant feature transform algorithm and orientation feature detection algorithm were used to extract the weld profile as well as the welding torch. The feature points of the profile were identified with the Harris corner detection algorithm. Since the weld area is relatively small, a sub pixel level measurement method was proposed to obtain the all-position bead height, width and area. The Weld center of gravity was detected with the zero and first moment. Experimental results showed the strong adaptability and high accuracy of that this proposed method. It provides the possibility of adjusting the attitude of the welding torch and welding process parameters to control weld formation online.