Plastic Debris Detection via Satellite Imagery
Detecting plastic debris in the Mediterranean Sea using Sentinel-2 satellite data and spectral analysis.
This Omdena open-source collaboration tackles environmental monitoring by detecting marine plastic debris using satellite imagery. The project processes Sentinel-2 multispectral data through a pipeline that includes atmospheric correction via ACOLITE, cloud masking with s2cloudless, and spectral index computation.
The detection workflow generates patch/mask pairs from satellite tiles, filtering detections using spectral indices including NDVI (Normalized Difference Vegetation Index) and FDI (Floating Debris Index). The pipeline integrates with MARIDA-type dataset workflows for training data preparation and uses NetCDF windrow filtering for debris candidate identification.
Key contributions include implementing the ACOLITE correction workflow, RGB patch generation, mask rasterization, cloud detection integration, and patch-level processing within the process_tile loop. The work demonstrates rigorous data pipeline engineering applied to environmental satellite remote sensing.
- Sentinel-2 satellite data processing with ACOLITE atmospheric correction
- Cloud masking integration using s2cloudless for clean imagery
- Spectral index filtering with NDVI and FDI for debris detection
- Patch/mask pair generation pipeline with NetCDF windrow filtering
- MARIDA-type dataset workflow integration for training data preparation