In this note, we propose an approach to initialize the Iterative Closest Point (ICP) algorithm to match unlabelled point clouds related by rigid transformations. The method is based on matching the ellipsoids defined by the points' covariance matrices and then testing the various principal half-axes matchings that differ by elements of a finite reflection group. We derive bounds on the robustness of our approach to noise and numerical experiments confirm our theoretical findings.
@article{arxiv.2212.05332,
title = {An approach to robust ICP initialization},
author = {Alexander Kolpakov and Michael Werman},
journal= {arXiv preprint arXiv:2212.05332},
year = {2023}
}
Comments
9 pages, 18 figures, 1 table; GitHub repository at (https://github.com/sashakolpakov/icp-init)