A novel light scanning system based on circular laser trajectory for welding robot is developed. With the help of image processing technique, intelligent laser welding could be realized. According to laser triangulation algorithm and Scheimpflug condition, mathematical model for circular laser vision is built. This scanning system projects circular laser onto welded seams and recovers the depth of the welded seams, escapes from shortcomings of less information, explains ambiguity and single tracking direction inherent in “spot” or “line” type laser trajectory. Three-dimensional (3D) model for welded seams could be recognized after depth recovery. The imaging error is investigated also. A novel light scanning system based on circular laser trajectory for welding robot is developed. With the help of image processing technique, intelligent laser welding could be realized. According to laser triangulation algorithm and Scheimpflug condition, mathematical model for circular laser vision is built. scanning system projects circular laser onto welded seams and recovers the depth of the welded seams, escapes from shortcomings of less information, illustrating ambiguity and single tracking direction inherent in “spot ” or “line ” type laser trajectory. Three-dimensional (3D) model for welded seams could be recognized after depth recovery. The imaging error is been also.