Skip to content

A 3D Global Descriptor For Loop Closure Detection. NDT-Map-Code.

License

Notifications You must be signed in to change notification settings

liuXinGangChina/NDTMC

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

NDT-Map-Code

NDT-Map-Code: A 3D global descriptor for real-time loop closure detection in lidar SLAM

You can find the paper here NDT-Map-Code.

Demo

NDT-Map-Code

Example

  • We implemented the SLAM package of NDTMC and LIOSAM integration, which can be found at NDTMC-LIO-SAM.

  • We tested our descriptor on KITTI sequences 00, 02, 05, 06, 07 and 08:

image

image

  • F1 score and extended precision results on the KITTI dataset:

image

How to use

  1. Create GT of LCD

    Use script/create_ground_truth.py to generate the LCD GT value, you need to modify the two file paths in the script/create_ground_truth.py (odometry ground truth file path, LCD GT value result path).

  2. Save descriptors

    Modify the main function in run_demo.cpp as follow:

    int main(int argc, char **argv)
    {
        saveDesc(argv);
        // matchForKitti(argv);
        // matchForNIO(argv);
        return 0;
    }
    
    

    Build and run script/auto_test.py to generate descriptors, where input_folder and output_folder need to be modified to the voledyne folder path of the currently processed sequence and the folder path where you want to save the descriptor results, respectively.

  3. Descriptors matching

    Modify the main function in run_demo.cpp as follow:

    int main(int argc, char **argv)
    {
        // saveDesc(argv);
        matchForKitti(argv);
        // matchForNIO(argv);
        return 0;
    }
    
    

    Build and run script/auto_match.py to perform descriptor matching, where input_folder and output_txt need to be modified to the folder path where the descriptor was saved in the previous step and the txt file path where you want to save the matching result, respectively.

At this point, you can get the matching result of each frame, and you can process it yourself to get more result information, such as: precision-recall curve, F1 Score and Extended Precision.

Cite NDTMC

@misc{liao2023ndtmapcode,
      title={NDT-Map-Code: A 3D global descriptor for real-time loop closure detection in lidar SLAM}, 
      author={Lizhou Liao and Li Sun and Xinhui Bai and Zhenxing You and Hongyuan Yuan and Chunyun Fu},
      year={2023},
      eprint={2307.08221},
      archivePrefix={arXiv},
      primaryClass={cs.RO}
}

Contact

Contributors

  • Lizhou Liao: completed the code

Acknowledgement

  • Thanks for NIO low-speed localization and mapping group.

About

A 3D Global Descriptor For Loop Closure Detection. NDT-Map-Code.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C++ 87.7%
  • Python 11.0%
  • CMake 1.3%