Defect detection of three-dimensional structures is essential for intelligent laser robot system. A method to detect structural damages in robot laser remanufacturing is presented, and a high precision algorithm is proposed to predict edge locations of damaged part. Initial edge points of the damaged parts are identified using normal-vector technique based on accumulation of the characteristic areas of the mesh vertices. A three-step optimization strategy is proposed to solve gaps, branches, and serrated boundaries of the initial edges. Finally, closed defect edges are predicated with high accuracy. The proposed method is fast for structural damage detection with high precision, which is necessary for industrial applications of intelligent laser robot system.