In this paper, we proposed a laser welding quality monitoring system based on vector machine, which uses the light and sound sensors access to the various signals of the welding and Gabor transform features to extract the vector. The basic features of Laser welding is the formation of small holes, pool and photo-induced plasma, and the result is quality of the welding of light, sound and the potential of poor signal. On this basis, making use of the advantage and objectivity of machine leaning, SVM classification and more comprehensive information to determine the parameters of laser welding quality of the welding seam.