In robotic 3D inspection systems, it is essential to obtain precise measurement results by covering the entire surface area of the target objects. To this end, various view planning methods have been proposed to guarantee complete coverage in order to ensure precise measurements of the target object. However, for precise inspection, in addition to the coverage of the object, globally consistent registration should also be considered. If the target object has a complex shape, occlusions may occur and registration often fails. Therefore, this study proposes a computer-aided design-based automatic inspection system that simultaneously guarantees both the coverage and globally consistent registration model for objects with complex shapes. To this end, view planning that covers all areas of the complex-shaped objects without occlusions and guarantees globally consistent registration is proposed. Experiments were performed on a piston head, casting mold for a smartphone case, and gear; these objects were inspected using an error map and a quantitative error report. Experimental results show that complex-shaped objects can be successfully inspected without unscanned areas and registration failure by using the proposed method.

在机器人 3D 检测系统中，通过覆盖目标物体的整个表面区域来获得精确的测量结果至关重要。为此，已经提出了各种视图规划方法来保证完全覆盖以确保对目标对象的精确测量。但是，对于精确检查，除了对象的覆盖范围外，还应考虑全局一致的注册。如果目标物体形状复杂，可能会出现遮挡，配准往往会失败。因此，本研究提出了一种基于计算机辅助设计的自动检测系统，可同时保证复杂形状物体的覆盖范围和全局一致的配准模型。为此，提出了覆盖复杂形状物体所有区域而没有遮挡并保证全局一致配准的视图规划。在活塞头、智能手机外壳的铸模和齿轮上进行了实验；使用错误图和定量错误报告检查这些对象。实验结果表明，使用所提出的方法可以成功检测复杂形状的物体，没有未扫描区域和配准失败。