Enhanced Multilevel Ground Segmentation for 3-D LiDAR Point Clouds Based on Shape Analysis
IEEE Access · 2026 · 14 · 75088-75100
DOI: 10.1109/access.2026.3693841 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Ming Gou und 8 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Hubei Provincial Natural Science Joint Foundation.
Abstract (Originalsprache)
Conventional ground filtering methods for large-scale 3D Light Detection and Ranging (LiDAR) point clouds often show insufficient adaptability to complex terrain and non-ground interference in transmission-corridor scenes, making it difficult to obtain stable and accurate ground-segmentation results. To address this issue, this paper proposes an enhanced multilevel ground segmentation method for transmission-corridor point clouds. The method constructs an initial candidate ground set using grid-based minimum-elevation sampling, refines reliable ground seeds through local PCA and a planarity criterion, and then employs two-dimensional Gaussian process regression (2D GPR) together with connectivity-based refinement to achieve continuous and spatially consistent ground extraction. Experiments on two real transmission-corridor scenes, including a relatively flat scene and a mountainous scene, show that the proposed method achieves the lowest total error and the highest Kappa coefficient among the compared methods. In Scene 1, it obtained a Type I error of 4.8%, a Type II error of 0.3%, a total error of 1.7%, and a Kappa coefficient of 96.6%; in Scene 2, the corresponding values were 9.9%, 1.6%, 4.4%, and 90.0%, respectively. These results indicate that the proposed method can better preserve valid ground points while suppressing non-ground interference in complex transmission-corridor environments.
Förderung
- Hubei Provincial Natural Science Joint Foundation
Zitieren
APA
Gou, M., Xu, W., Chen, Z., Liu, C., Zhang, L., Tang, H., Liu, J., Zhang, P., & Chen, X. (2026). Enhanced Multilevel Ground Segmentation for 3-D LiDAR Point Clouds Based on Shape Analysis. IEEE Access, 14, 75088-75100. https://doi.org/10.1109/access.2026.3693841
Vancouver
Gou M, Xu W, Chen Z, Liu C, Zhang L, Tang H, et al. Enhanced Multilevel Ground Segmentation for 3-D LiDAR Point Clouds Based on Shape Analysis. IEEE Access. 2026;14:75088-75100. doi:10.1109/access.2026.3693841
BibTeX
@article{Gou2026,
author = {Gou, Ming and Xu, Weizhong and Chen, Zhong and Liu, Chunyu and Zhang, Liguang and Tang, Hao and Liu, Jiwu and Zhang, Puyu and Chen, Xi},
title = {Enhanced Multilevel Ground Segmentation for 3-D LiDAR Point Clouds Based on Shape Analysis},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {75088--75100},
doi = {10.1109/access.2026.3693841}
}
EUVIMED