Abstract: | In this paper, we propose a pre-filtering method of geometric measurement for LiDAR odometry, which reduces a localization error by data association based on a statistical property. Recently, LiDAR odometry has been mainly developed in such a way that geometric feature points are matched for pose estimation. One of the critical problems affecting localization performance is false association between measurements that deviate from the assumption of matching geometrically identical features in 3D. We propose a pre-filtering method using stable plane feature points. To reduce the risk of false association, a stable plane feature is selected using the covariance of the feature point group on a local plane. After the pre-filtering process, odometry is performed by solving the data association of the measurement. To verify the proposed method, we formulate the LiDAR odometry in the LOAM framework, compare the accuracy of localization with LOAM, and show the improved result through Monte Carlo simulation. |
Published in: |
Proceedings of the 36th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2023) September 11 - 15, 2023 Hyatt Regency Denver Denver, Colorado |
Pages: | 2134 - 2142 |
Cite this article: | Lee, Hanyeol, Jung, Jae Hyung, Park, Chan Gook, "LiDAR Odometry with Pre-Filtering of Plane Feature Points," Proceedings of the 36th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2023), Denver, Colorado, September 2023, pp. 2134-2142. https://doi.org/10.33012/2023.19411 |
Full Paper: |
ION Members/Non-Members: 1 Download Credit
Sign In |