Unsupervised Large-Scale Point Cloud Registration via Spherical Projection Consistency
IEEE Signal Processing Letters · 2026 · 33 · 2190-2194
DOI: 10.1109/lsp.2026.3695797 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Signal Processing Letters, 2026. Verfasst von Yejun Shou und 4 weiteren Autor:innen. Die Arbeit nennt 44 Literaturangaben. Gefördert von: NARI Technology Co., Ltd., Research and Application of Key Technologies for Efficiency Improvement of Power Operation Robots in Diverse Distribution Network Line Scenarios and Complex Op….
Abstract (Originalsprache)
In real-world applications, large-scale outdoor LiDAR point cloud registration faces significant challenges, including data sparsity, occlusions, and reliance on expensive ground-truth pose labels. To address these issues, this paper proposes an unsupervised registration framework based on spherical projection consistency. Specifically, both the input point cloud and its spatially transformed counterpart are projected into range images, and their spatial consistency is exploited as a supervision signal. Furthermore, a multi-scale patch-to-patch feature fusion module is introduced to effectively integrate features from point clouds and range images, thereby enhancing feature discriminability. In addition, a dynamic masking strategy is applied in the range image domain to mitigate the impact of sparsity variations and further strengthen the spatial consistency of the supervision signal. Extensive experiments on the KITTI and NuScenes datasets demonstrate that the proposed method outperforms state-of-the-art unsupervised approaches, validating its effectiveness and robustness in large-scale outdoor scenarios.
Förderung
- NARI Technology Co., Ltd.
- Research and Application of Key Technologies for Efficiency Improvement of Power Operation Robots in Diverse Distribution Network Line Scenarios and Complex Op…
Zitieren
APA
Shou, Y., Li, S., Wang, H., Xu, Z., & Cao, Y. (2026). Unsupervised Large-Scale Point Cloud Registration via Spherical Projection Consistency. IEEE Signal Processing Letters, 33, 2190-2194. https://doi.org/10.1109/lsp.2026.3695797
Vancouver
Shou Y, Li S, Wang H, Xu Z, Cao Y. Unsupervised Large-Scale Point Cloud Registration via Spherical Projection Consistency. IEEE Signal Processing Letters. 2026;33:2190-2194. doi:10.1109/lsp.2026.3695797
BibTeX
@article{Shou2026,
author = {Shou, Yejun and Li, Shuai and Wang, Haocheng and Xu, Zhijie and Cao, Yanlong},
title = {Unsupervised Large-Scale Point Cloud Registration via Spherical Projection Consistency},
journal = {IEEE Signal Processing Letters},
year = {2026},
volume = {33},
pages = {2190--2194},
doi = {10.1109/lsp.2026.3695797}
}
EUVIMED