Near-optimal selection of views and surface regions for ICP pose estimation

Abstract

This paper presents an innovative approach for the selection of well-constrained views and surface regions for efficient ICP pose estimation using LIDAR range scanning. The region selection is performed using the Principal Component Analysis technique with derived predictive indices that can be used to assess a view/region for pose estimation. Localized scanning has been proposed for spacecraft rendezvous operations, particularly in the "last mile" scenario where whole object scanning is not possible. The paper illustrates the PCA approach for selection of optimal scanning views and localized regions using (a) CAD models of several spacecraft structures with supporting simulation results based on large amount of data, and (b) a model of a faceted shape, cuboctahedron, which was scanned using Neptec's TriDAR laser scanner. The results confirm the hypothesis that the selected views or regions deliver accurate estimates for the pose norm and also for each component of the pose. © 2010 Springer-Verlag.