Part-in-whole point cloud registration for aircraft partial scan automated localization

Abstract

Skin deformation measurement is quite important for the aerodynamic performance of aircrafts. To achieve this, 3D scanning is usually adopted to capture the 3D point cloud, which reflects the geometric information of partial skin surface. The scanned point cloud is then compared with the designed CAD model to quantify the shape deformation. However, accurate localization of the partial scanned point cloud, which is the prerequisite for the shape comparison, still remains an open problem. This can be formulated as a part-in-whole registration problem in the free-form surface, which is of significant difficulty due to the featurelessness and local similarity problems on the aircraft skin surface. In this paper, we propose a “coarse-to-fine” registration framework to address these two challenges. First, a Multi-Descriptor Voting (MDV) scheme is presented to roughly locate the partial scanned point cloud in the whole aircraft skin model. A voting mechanism using multiple 3D descriptors leads to the high possibility of correct localization under the featurelessness situation on the free-form surface. Then, we observed in the aircraft skin that there are seam structures which can assist in localization, we accordingly design a Seam Structure Aware ICP (SSA-ICP) for fine registration to eliminate the localization ambiguity in local regions, based on the detected seam points. The proposed algorithm is implemented using Point Cloud Library (PCL), and the results show that the proposed MDV and SSA-ICP achieve promising performance on both synthetic and real scanned point clouds.