Aircraft Seam Feature Extraction from 3D Raw Point Cloud via Hierarchical Multi-structure Fitting

Abstract

Advanced 3D scanners and depth cameras facilitate aircraft skin assembly, by acquiring high-precision point clouds. The measurement of skin seam quality directly affects the aerodynamic performance of aircraft. The extraction of the seam feature is an essential process in aircraft assembly. We propose a robust seam point extraction approach from raw point clouds of aircraft skins. Motivated by the non-uniform distribution of the acquired point cloud, we first devise a new tensor voting algorithm to exclude most of the points not belonging to seam structures. Our main contribution is the design of a hierarchical multi-structure fitting algorithm to classify the extracted points into non-seam points and seam points. We design the line constraint and the misjudgment strategy to cope with the seam feature extraction. In addition, we show that our approach can be applied into the gap-and-flush measurement (GAFM) of skin seams: it avoids the error of fitting two close lines into one line, which is a common and arduous case for the state of the arts. A number of experiments demonstrate the effectiveness and reliability of our approach on both synthetic and real-world raw data.