Approaches for Recognition and Interpretation of Workpiece Surface Features Using Structured Lighting

摘要

Vision sensing of the exact location and orientation of a workpiece and determination of its detailed surface geometry are often necessary for robot guidance, inspection, quality evaluation, or process control in manufacturing applications of robots, as well as for sensing and operator feedback in telerobotic tasks. In the work reported in this paper, con trolled illumination in the form of projected planes of laser light is used in order to determine such three-dimensional (3D) workpiece geometry. A data-driven approach for the rec ognition of general stripe contour features is proposed and implemented. A computationally efficient scheme for 3D surface modeling using a local ruled surface approximation is also developed for the determination of cross-sectional features and dimensions. The interpretation of the visually detected workpiece surface geometry and features typically requires specialized engineering knowledge and expertise. The feasibility of encoding such knowledge in the form of production rules is demonstrated through the development of a prototype defect interpretation system for weld inspection.