A Precise Method for Identifying 3-D Circles in Freeform Surface Point Clouds

Jingdan Deng, Shangzhi Liu, Hong Chen, Youquan Chang, Yingtian Yu, Wenting Ma, Yaonan Wang, He Xie

Abstract: Accurate recognition and precise positioning of surface circular holes in aerospace components, which are crucial for automatic drilling and rivet leak detection, ensure the reliability of manufacturing and maintenance processes. Existing circular hole detection and positioning methods mainly focus on planar point clouds, leading to significant inaccuracies when identifying circular holes on complex surfaces such as fairings, wings, and fuselages. To address this challenge, we propose a novel approach specifically designed to accurately identify and position circular holes on irregular surfaces. First, an orthogonal-guided rotational sphere (OGRS)-based margin point extraction method is introduced to precisely extract the boundaries of circular holes, while incorporating normal vector constraints to adapt to point cloud data with varying curvatures. After clustering and segmenting the margin points, a novel anisotropic overdetermined equation (AOE) method is applied for identifying circular holes, yielding their precise position information. This method consists of two key parts: the first is a proximal averaging and residual optimization (PR) approach to determine the overall orientation of circular holes, and the second is an overdetermined 3-D circle identification (OC) method for accurate hole localization. In the experimental evaluation, the proposed algorithm’s effectiveness and general applicability to freeform surfaces are verified through three distinct experiments: those involving surface point clouds with varying curvatures, cylindrical models, and scanned point clouds of aircraft skin. Across the simulated point cloud data, the algorithm demonstrates a maximum radius error of 0.005 mm, with a maximum error of 0.015 mm for scanned point clouds, thus meeting aerospace manufacturing standards.