This study proposes a V-shaped seam tracking algorithm based on particle filter with a histogram of oriented gradient (Hog) to address the problem of accuracy loss due arc light, spatter, and other noises in the process of weld tracking. The Hog has good invariance in geometric and optical deformation, and it is used to describe the directionality of the laser stripe at the flex point. The Hog of the target region of the weld seam is extracted to establish the target template. The candidate region state of the weld seam is predicted by the continuity of the position information of the weld seam image sequence. The Hog feature of the candidate region state is calculated through particle sampling. The similarity in the features of the candidate state and the target template is calculated, the optimal weld position is then obtained, and the weld seam tracking is realized. The tracking experiment of typical V-shaped weld image is performed. Experimental results show that the proposed algorithm can accurately locate the target position of weld under the environment of strong arc light, splash and other noise interferences, and the tracking error is less than 0.24 mm.