Abstract: Aiming at the problem of low positioning accuracy of weld trajectory in automatic arc welding process, a weld tracking method based on gaussian kernel correlation and particle filter was proposed. Image processing technology was adopted to extract the initial feature area of the weld image. The weld feature area was tracked by the gaussian kernel correlation algorithm during welding and the center position of the weld was solved according to the weld feature area. In order to reduce the influence of dynamic noise and observation noise in the welding environment on the positioning accuracy of the weld trajectory, a particle filter model of the weld center position was established to estimate the actual position of the weld. The standard butt weld was tracked in the experiment. The experiment results showed that the gaussian kernel correlation combined particle filter algorithm could further improve the positioning accuracy of the weld trajectory under the interference of the welding ambient noise.