Robotic Path Planning for Inspection of Complex-Shaped Objects

摘要

In 3D measurement inspection systems, precise registration between measured point clouds is required to obtain high quality results. In such cases, it is critical that there be proper overlaps between the measurements and that the overall shapes be measured without any blank areas. Thus, if the inspection system does not reflect the shape of the object, unmeasured areas may remain, causing the registration to fail or deteriorate. To solve this problem, a robotic path planning method to measure all areas of complex shaped objects is proposed. First, a segmentation-based view planning to extract a viewpoint that properly reflects the object shape is presented. In addition, occlusions that may occur in the extracted viewpoints are prevented, and path planning is performed to make the viewpoint available to a measurement system comprising a robot and rotary table. Furthermore, it is shown that a complex-shaped object can be measured without occlusions using the proposed method.