Skip to content

A fast and robust global registration library for outdoor LiDAR point clouds.

License

Notifications You must be signed in to change notification settings

HKUST-Aerial-Robotics/G3Reg

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

27 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

G3Reg:

Pyramid Graph-based Global Registration using Gaussian Ellipsoid Model

Zhijian Qiao, Zehuan Yu, Binqian Jiang, Huan Yin, and Shaojie Shen

IEEE Transactions on Automation Science and Engineering

News

  • 03 Apr 2024: Accepted by IEEE TASE!
  • 19 Dec 2023: Conditionally Accept.
  • 22 Aug 2023: We released our paper on Arxiv and submit it to IEEE TASE.

Abstract

G3Reg is a fast and robust global registration framework for point clouds.

Features:

  • Fast matching: We utilize segments, including planes, clusters, and lines, parameterized as Gaussian Ellipsoid Models (GEM) to facilitate registration.
  • Robustness: We introduce a distrust-and-verify scheme, termed Pyramid Compatibility Graph for Global Registration (PAGOR), designed to enhance the robustness of the registration process.
  • Framework Integration: Both GEM and PAGOR can be integrated into existing registration frameworks to boost their performance.

Note to Practitioners:

  • Application Scope: The method outlined in this paper focuses on global registration of outdoor LiDAR point clouds. However, the fundamental principles of G3Reg, including segment-based matching and PAGOR, are applicable to any point-based registration tasks, including indoor environments.
  • Segmentation Check: If the registration does not perform as expected on your point cloud, it is advisable to review the segmentation results closely, referring to Segmentation Demo.
  • Alternative Matching Approaches: For practitioners preferring not to use GEM-based matching, point-based matching is a viable alternative. For implementation details, please refer to the configuration file at fpfh_pagor.
  • Limitations: Segment-based matching may be less effective in environments with sparse geometric information, such as areas with dense vegetation. In such scenarios, enhancing segment descriptions through hand-crafted or deep learning-based descriptors is recommended to improve matching accuracy.

Getting Started

Qualitative results on datasets

KITTI-08

kitti08.mp4

Apollo-Highway

apollo20.mp4

Apollo-Sunnyvale

apollo21.mp4

Livox-HIT-1

hit-1-1120.mp4

Livox-HIT-3

hit-3-861.mp4

Application to Multi-session Map Merging

map_merging

Acknowledgements

We would like to show our greatest respect to authors of the following repos for making their works public:

Citation

If you find G3Reg is useful in your research or applications, please consider giving us a star 🌟 and citing it by the following BibTeX entry.

@ARTICLE{qiao2024g3reg,
  author={Qiao, Zhijian and Yu, Zehuan and Jiang, Binqian and Yin, Huan and Shen, Shaojie},
  journal={IEEE Transactions on Automation Science and Engineering}, 
  title={G3Reg: Pyramid Graph-Based Global Registration Using Gaussian Ellipsoid Model}, 
  year={2024},
  volume={},
  number={},
  pages={1-17},
  keywords={Point cloud compression;Three-dimensional displays;Laser radar;Ellipsoids;Robustness;Upper bound;Uncertainty;Global registration;point cloud;LiDAR;graph theory;robust estimation},
  doi={10.1109/TASE.2024.3394519}}
@inproceedings{qiao2023pyramid,
  title={Pyramid Semantic Graph-based Global Point Cloud Registration with Low Overlap},
  author={Qiao, Zhijian and Yu, Zehuan and Yin, Huan and Shen, Shaojie},
  booktitle={2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  pages={11202--11209},
  year={2023},
  organization={IEEE}
}

About

A fast and robust global registration library for outdoor LiDAR point clouds.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published