limosat - Lagrangian Ice Motion from Satellites

limosat is an open source, highly configurable algorithm for calculating sea ice drift from remote sensing imagery. It processes satellite images like so:

1. Keypoint Detection: Identify points representing ice features in a semi-constrained grid.
2. Interpolation: Interpolate positions for unmatched keypoints using homography.
3. Pattern Matching: Refine keypoint positions using template matching and discard points with low correlation.
4. Persistent Storage: Optionally save drift data for later analysis.

Citation

Chua, S. M. T., & Korosov, A. (2025). limosat - Lagrangian Ice Motion from Satellites (v0.1.0). Zenodo. https://doi.org/10.5281/zenodo.15111936

Repository Structure

.
├── LICENSE              # MIT License information
├── README.md            # Project documentation
├── environment.yaml     # Conda environment specification
├── examples/            # Usage examples
├── limosat/             # LiMOSAT library source code
└── tests/               # Tests

Setup Environment

conda env create -f environment.yaml && conda activate limosat

Run limosat

1. Prepare Your Data: Organize your satellite imagery into a folder and run preprocessing.py. If you use the gridded descriptor cache, note that changing stride or descriptor parameters (ORB settings, border size, octave) requires regenerating the cache.
2. Build catalog: Use create_image_gdf to build a catalog of imagery metadata.
3. Set-up database(optional): Enable persistence by providing a SQL engine and Zarr storage path to store both the drift keypoints and pattern matching templates. limosat can also be run without persistence.
4. Run examples/limosat_drift.ipynb
5. Visualise results

License

This project is licensed under the MIT License. See the LICENSE file for details.