⛵ sentinel2-xcube-boat-detection
Github repository to detect and count boat traffic on Euro Data Cube using ESA's Sentinel-2 imagery
Setup with EDC
- Create an account on Mapbox and Euro Data Cube
- Clone this repository in a Jupyter Lab environment (Python 3.6+) on Euro Data Cube
- Edit Mapbox and Sentinel Hub token in a .env file
- Run
pip install -r requirements.txt
Pipeline
🛰️ Download data
1. Sentinel 2 L1C bands B03, B08 and CLP in areas of interest (e.g., Ports, Straits, MPA), saved under data/aoi.json
Annotations/Labels: 1 squared km chips with boat counts
☄️ Preprocess
2. Informative snaps: Nans and cloudy images (from CLP) removed
Background NDWI: image computed by fusion over time (max operation over NDWI band)
🔭 Detect and count boat traffic
3. Output: Boat traffic presence/density heatmap (10-100m spatial resolution) & global count
Hidden states: Pixels' embeddings, Patches' encodings
Input (3 channels): NIR band, Background NDWI and CLP band
Links
Tweet and RACE Dashboard
April 2020, ESA🇪🇺
February/March 2020, 1st Covid-19 Lockdowns in Europe Credits
ESA, Copernicus, Euro Data Cube, Sinergise
License
This repository and dataset is made available under the Do No Harm License and Creative Commons Attribution 4.0 International License by Michel Deudon and Zhichao Lin.