Hierarchical Viewpoint Planning for Complex Surfaces in Industrial Product Inspection
Shaohu Wang, Yuchuang Tong, Xiuqin Shang, Zhengtao Zhang
Abstract
Appearance quality assessment plays a crucial role in industrial product manufacturing, where traditional inspection methods struggle with the diverse surface structures and complex shapes. This article introduces a novel hierarchical viewpoint planning paradigm designed to tackle the challenges associated with mapping camera viewpoints onto complex object surfaces featuring demanding imaging quality requirements. Based on multilevel geometric representation of the object point cloud, the proposed paradigm utilizes a divide-and-conquer strategy for systematic and effective viewpoint planning, which encompasses both bottom–up and top–down methodologies tailored for regular and irregular surfaces, respectively. The bottom–up method employs gridization-based viewpoint generation for rapid acquisition of high-quality viewpoints, while the top–down method utilizes hybrid confidence and dynamic viewpoint optimization to achieve complete and high-quality surface coverage. By combining these complementary strategies, the proposed method ensures high-quality surface coverage for different types of geometric primitives. Experimental evaluations demonstrate superior performance compared with existing approaches, and real-world experiments validate the efficiency and effectiveness of the proposed method in industrial settings.