Robotic measurement has emerged as a prominent method for the quality control of large-scale metal workpieces. However, it faces two significant challenges: optical measurement of homogeneous high-reflective metal surfaces and point cloud registration on local featureless surfaces. To address these challenges, this study presents a cooperative optical profiler that integrates a certain number of light-emitting diodes into a conventional fringe projection profiler. This integration enables the use of photometric stereo, providing comprehensive and accurate surface normals of machined metal surfaces with intricate details such as tool marks. The surface normal was then fused with the initial point cloud from the fringe projection profiler to yield complete local measurement results. In addition, a geometry descriptor incorporating the surface normal was designed to improve data registration accuracy. Experiments based on six small workpieces with the combination of three shapes and two materials and a large freeform workpiece demonstrated that the proposed method achieved measurement accuracy close to that of a coordinate measuring machine.
Read full abstract