Abstract: A new classification of the vision sensing technology in seam recognition is proposed based on the feature of the target. It is different from the two kinds of usual classification:passive vision and active vision, or one-dimensional, two-dimensional and three-dimensional vision.According to this classification, the existing methods of seam vision sensing are separated into two classes: vision sensing based on the macroscopical three-dimensional structure feature of workpiece, and vision sensing based on the local optical feature of the workpiece.Furthermore, with this classification, a new method of seam recognition based on texture analyzing is proposed, according to the obvious texture feature of the seam especially in the multilayer welding and the multipass welding.In this method, the texture feature of the seam, which is obvious but ignored before, is analyzed and utilized to recognize the seam automatically.