PlasticPatrol

PlasticPatrol AI detects ocean plastic via Sentinel-2 and FDI. Boaters validate patches using Galileo to earn Eco-Points, turning citizen science into real impact.

  • 0 Raised
  • 54 Views
  • 0 Judges

Tags

  • Romania

Categories

  • Challenge #2: Tracking and preventing water pollution​

Description

PlasticPatrol is an AI-based and satellite-powered platform for detecting, tracking, and incentivizing the cleanup of plastic debris in oceans and on coastlines.

The problem:

Over 8 million tonnes of plastic enter the oceans every year, devastating marine ecosystems, contaminating water supplies, and threatening biodiversity. Current cleanup efforts are fragmented, reactive, and rely on chance encounters with debris. Municipalities and environmental organizations lack a real-time, scalable system to:

- Detect where plastic accumulates (both in open water and along coastlines),

- Verify that reported cleanups actually happened,

- Motivate citizens to actively participate in removal.

The solution:

PlasticPatrol bridges the gap between space-based remote sensing and citizen-led action on the ground through a three-layer architecture:

1. Satellite Detection Pipeline — A custom evalscript running on Copernicus Sentinel-2 L2A imagery applies a multi-stage spectral analysis (NDWI water masking → FDI Floating Debris Index → NDVI vegetation rejection → land-kill filters using SWIR/NIR bands) to identify floating plastic accumulations at 10-meter resolution across configurable hotspot regions worldwide (Rotterdam, Bosphorus, Constanța, Singapore Strait, LA/Long Beach, Rio de Janeiro, Malta Channel, and more).

2. ML Image Verification — When a user physically reaches a debris cluster and submits a photo as proof, an on-device EfficientNet/MobileNetV2-based binary classifier (trained on ocean debris vs. clean water imagery) verifies that the photo genuinely shows plastic. This prevents fraudulent "cleanup" claims. The cleanup is then independently re-verified by the next satellite pass — if the satellite now sees clean water where debris was previously detected, the cleanup is confirmed and eco-points are awarded.

3. Gamified Citizen Engagement — Users register, receive a live map of debris clusters near their location (loaded via geolocation within a configurable radius), reserve a cluster for 24 hours, physically collect the waste, submit photographic proof, and earn eco-points upon satellite verification. A leaderboard, notification system, and point tiers (small/medium/large clusters) drive sustained participation.

-> The platform features a FastAPI backend with PostGIS spatial queries, an Angular 21 frontend with Leaflet-based interactive maps, and a fully automated data pipeline that periodically scans global hotspots and inserts new detections while deduplicating within 100-meter proximity.


Data & Signals Used

EU Space AssetHow It's UsedValue Added
Copernicus Sentinel-2 L2A (Multispectral Imagery)We consume 5 spectral bands (B03 Green, B04 Red, B06 Vegetation Red Edge, B08 NIR, B11 SWIR) via the Sentinel Hub Processing API to run a custom evalscript that detects floating debrisProvides free, open, global, 10-meter resolution imagery with a 5-day revisit time — enabling near-real-time monitoring of plastic hotspots without any proprietary sensor costs
Copernicus Data Space Ecosystem (CDSE)Our sentinel_fetcher.py pipeline authenticates via OAuth2 against identity.dataspace.copernicus.eu and fetches processed imagery tiles through the CDSE API endpointGives us scalable, cloud-based access to petabytes of Earth observation data without needing to download and store raw satellite scenes locally
NDWI / FDI Spectral Indices (derived from Sentinel-2)Our evalscript computes the Normalized Difference Water Index (B03−B08)/(B03+B08) to confirm water pixels, then applies the Floating Debris Index using B08, B06, B11 to isolate anomalous surface reflectance caused by plasticThese indices, purpose-built for Sentinel-2's band configuration, allow us to distinguish floating plastic from clean water, algae, foam, and land — a capability only possible with multispectral satellite data
WGS84 / EPSG:4326 Coordinate Reference SystemAll geospatial data is stored in PostGIS with SRID 4326, and bounding boxes for Sentinel queries use WGS84 coordinatesEnsures interoperability between satellite imagery, our database, and third-party mapping tools (Leaflet, QGIS, Google Earth via GeoJSON export)

Challenge Addressed: Protecting Water Resources from Pollution

PlasticPatrol directly tackles the challenge of protecting and managing water resources by combining EU space technology with citizen action to remove one of the most pervasive and damaging pollutants — plastic waste.

How We Contribute

  1. Detection → Prevention: By identifying plastic accumulation zones before debris breaks down into microplastics, we enable proactive intervention. Microplastics (particles < 5mm) are nearly impossible to recover once dispersed and contaminate drinking water sources, aquatic food chains, and sediment. Our 10-meter satellite detection catches debris while it's still recoverable.

  2. Data-Driven Water Management: Our platform's GeoJSON export and statistics dashboard (total detected, collected, verified debris; monthly detection trends; collection rates) give environmental agencies actionable intelligence on pollution patterns. Municipalities can identify which rivers, ports, or coastlines are chronic sources of plastic input and target infrastructure investments (e.g., river barriers, waste management upgrades) accordingly.

  3. Verified Cleanup = Measurable Impact: Unlike awareness campaigns, PlasticPatrol provides satellite-verified proof that water bodies are actually getting cleaner. Each verified cleanup is a data point that feeds back into our system — confirmed clean zones are deprioritized in future scans, allowing satellite resources to focus on new pollution fronts.

  4. Scalable Citizen Infrastructure: By gamifying cleanup through eco-points and leaderboards, we transform every smartphone user near a coastline into a potential water quality agent. The reserve-collect-verify cycle ensures accountability: you have 24 hours to act on your reservation, photo proof is ML-verified, and the satellite delivers final confirmation.

  5. Coastal & Open Water Coverage: Our system distinguishes between beach debris (manual reports, instant verification by authorized personnel) and ocean debris (satellite-detected, requires satellite re-verification). This dual approach addresses both coastal zone management and open water pollution monitoring, covering the full lifecycle of marine plastic.


Team

NameGitHubRoleBio
Rares Neacsu@RaresNeacsu🛰️ Backend & Satellite Pipeline LeadBuilt the complete Sentinel-2 data pipeline including the custom evalscript for floating debris detection, the multi-region scanning engine, and the spectral analysis logic (NDWI, FDI, NDVI, land-kill filters). Implemented the backend API for satellite scan triggers, debris clustering, and satellite verification of cleanups.
Matei Necula@matei-necula🤖 Full-Stack & ML Developer
Integrated the ML image classifier that verifies photo proof of cleanups, and built the Angular frontend including the collect overlay, JWT auth flow, notifications, and stats dashboard.
Andrei  Stan@rthh21🗺️ Frontend & Map DeveloperImplemented the interactive Leaflet map component with real-time geolocation tracking, debris marker rendering with color-coded severity, and the satellite tile layer integration. Built the map UI including the info overlay and location display.
Calin
Murariu
@Swiorx🔌 Backend & API Developer
Set up the FastAPI application and Uvicorn server, implemented the core GET and POST endpoints for debris data, and built the user authentication routes (register and login). Also handled the database configuration and initial schema setup.

Tech Stack: Python (FastAPI, SQLAlchemy, GeoAlchemy2) · Angular 21 · Leaflet · PostGIS · TensorFlow/Keras · Sentinel Hub API · Copernicus Data Space

Attachments