An approach to robust ICP initialization

TL;DR

This paper introduces a novel initialization method for the ICP algorithm that improves matching accuracy of unlabelled point clouds by leveraging covariance ellipsoids and principal axes matching, with proven robustness bounds.

Contribution

It presents a new robust ICP initialization technique based on covariance ellipsoids and principal axes matching, with theoretical noise robustness analysis.

Findings

Method improves ICP initialization accuracy.

Theoretical bounds on robustness to noise are derived.

Numerical experiments confirm theoretical predictions.

Abstract

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.