Machine vision system plays an important role in improving the quality of laser robot system. A system to detect three-dimensional (3D) shallow spot defects by combining two-dimensional (2D) images with 3D point clouds is carried out, which solves the problem of 3D positioning of shallow spot defects. With gray image processing, including image enhancement in the frequency domain, region marking and region merging, the system achieves the 2D positioning of defects. In order to obtain the 3D point cloud, laser remanufacturing part surfaces are scanned with binocular stereo vision system, and the 2D defect boundaries are then converted into 3D ones to realize the positioning of defects in the point cloud. Experimental results show that the system can effectively identify shallow spot defects, which further enables high precision remanufacturing of parts.