This repo contains the code for our ECMR2021 paper "Online Range Image-based Pole Extractor for Long-term LiDAR Localization in Urban Environments" and RAS paper "Online Pole Segmentation on Range Images for Long-term LiDAR Localization in Urban Environments".
Developed by Hao Dong and Xieyuanli Chen.
Overview of our approach. A. we project the LiDAR point cloud into a range image and B. extract poles in the image. C. based on the extracted poles, we then build a global pole map of the environment. D. we finally propose a pole-based observation model for MCL to localize the robot in the map.
If you use our implementation in your academic work, please cite the corresponding conference paper and journal paper:
@InProceedings{dong2021ecmr,
author = {H. Dong and X. Chen and C. Stachniss},
title = {{Online Range Image-based Pole Extractor for Long-term LiDAR Localization in Urban Environments}},
booktitle = {Proceedings of the European Conference on Mobile Robots (ECMR)},
year = {2021}
}
@article{dong2023jras,
title = {Online pole segmentation on range images for long-term LiDAR localization in urban environments},
journal = {Robotics and Autonomous Systems},
volume ={159},
pages = {104283},
year = {2023},
issn = {0921-8890},
doi = {https://doi.org/10.1016/j.robot.2022.104283},
author = {H. Dong and X. Chen and S. S{\"a}rkk{\"a} and C. Stachniss}
}
The code was tested with Ubuntu 20.04 with its standard python version 3.8.
-
Python dependencies
pip install numpy matplotlib open3d-python progressbar2 pyquaternion transforms3d scipy scikit-image networkx numba arrow pykitti
Download the dataset and extract the data in the /nclt/data folder following the recommended data structure, and then run:
python src/ncltpoles.pyDownload the KITTI raw data 2011_09_26_drive_0009 by navigating to the /kitti/raw_data folder and run:
./kitti_downloader.shthen run:
python src/kittipoles.pyDownload the KAIST 01 and KAIST 02 dataset and extract the data in the /mulran/data folder following the recommended data structure, and then run:
python src/mulranpoles.pyThe pole datasets are stored in the /data/pole-dataset/KITTI and /data/pole-dataset/NCLT folders. The data are stored in .npz format with the shape N*2. Each row represents the x and y position of one pole. You can evaluate the pole extraction with the groud-truth pole map by running:
python src/test_match.py python src/ncltpoles_learning.pyCopyright 2021, Hao Dong, Xieyuanli Chen, Cyrill Stachniss, Photogrammetry and Robotics Lab, University of Bonn.
This project is free software made available under the MIT License. For details see the LICENSE file.
Many thanks to the excellent open-source projects polex and SalsaNext.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent