Sonar Context

• Robust Imaging Sonar-based Place Recognition and Localization in Underwater Environments

  • Aceppted to ICRA 2023
  • Our paper is available at IEEE

• What is Sonar Context and polar key?

  • Sonar context is a global descriptor, which encodes geometric characteristics of underwater environments.
  • Polar key is a 1D vector collected of average values in each row direction of the sonar context.
  • The descriptor consists of coarse (polar key) and fine description (Sonar Context) for efficient loop closure detection.
  • The descriptor is robust for rotational and translational differences by adaptive shifting and matching algorithms.

• Author & Contributor

  • Hogyun Kim, Gilhwan Kang, Seungjun Ma, Seokhwan Jeong and Younggun Cho

Overview of our method

• Place Recognition & Pose Graph SLAM in Aracati 2017 Datasets

Datasets

• HOLOOCEAN

  • googledrive

• KRISO (KOREA RESEARCH INSTITUTE OF SHIPS AND OCEAN ENGINEERING)

  • not public

• Aracati 2017

  • github

How to use sonar context?

0. Download sonar context

    $ git clone https://github.com/sparolab/sonar_context.git

1. Requirements

• We implement our place recognition method in python3.

    $ pip3 install -r requirements.txt

2. Generate sonar context and polar key

• Modify the patch size parameters to fit your datasets in the img2scpk.py.
  • Set your input path (rectangular image called encoded polar image) and output paths (sonar context and polar key).

    $ cd sonar_context/generate/
    $ python3 img2scpk.py

3. Place Recognition and Localization

• Create your datasets.txt file similar to examples.txt and modify the parameters to fit your datasets.

    $ cd sonar_context/place_recognition/
    $ python3 main holoocean.txt

Contact

• Hogyun Kim ([email protected])

Supplementary

• Project Page
• Submission video

Cite Sonar Context

• Our paper has arxiv version on Google Scalar because of the Google Scalar issue. So, use the following bibtex for the ieee version.

@inproceedings{kim2023robust,
  title={Robust Imaging Sonar-based Place Recognition and Localization in Underwater Environments},
  author={Kim, Hogyun and Kang, Gilhwan and Jeong, Seokhwan and Ma, Seungjun and Cho, Younggun},
  booktitle={2023 IEEE International Conference on Robotics and Automation (ICRA)},
  pages={1083--1089},
  year={2023},
  organization={IEEE}
}

Thanks Sonar Context

@inproceedings{kim2018scan,
  title={Scan context: Egocentric spatial descriptor for place recognition within 3d point cloud map},
  author={Kim, Giseop and Kim, Ayoung},
  booktitle={2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  pages={4802--4809},
  year={2018},
  organization={IEEE}
}

@inproceedings{potokar2022holoocean,
  title={Holoocean: An underwater robotics simulator},
  author={Potokar, Easton and Ashford, Spencer and Kaess, Michael and Mangelson, Joshua G},
  booktitle={2022 International Conference on Robotics and Automation (ICRA)},
  pages={3040--3046},
  year={2022},
  organization={IEEE}
}

@inproceedings{jang2021multi,
  title={Multi-session underwater pose-graph slam using inter-session opti-acoustic two-view factor},
  author={Jang, Hyesu and Yoon, Sungho and Kim, Ayoung},
  booktitle={2021 IEEE International Conference on Robotics and Automation (ICRA)},
  pages={11668--11674},
  year={2021},
  organization={IEEE}
}

@article{dos2022cross,
  title={Cross-view and cross-domain underwater localization based on optical aerial and acoustic underwater images},
  author={Dos Santos, Matheus M and De Giacomo, Giovanni G and Drews-Jr, Paulo LJ and Botelho, Silvia SC},
  journal={IEEE Robotics and Automation Letters},
  volume={7},
  number={2},
  pages={4969--4974},
  year={2022},
  publisher={IEEE}
}

License

• For academic usage, the code is released under the BSD 3.0 license. For any commercial purpose, please contact the authors.