RescueTeam
Team consisting of Sarah Le Net (MLE, TheLab; École Polytechnique/Columbia; MLOps, PyTorch, GCP, RAG) and Paul‑Louis Fouesnant (Data Engineer; INSA Lyon; ETL, streaming).
YouTube Video
Project Description
This project is designed to detect, analyze, and prioritize areas affected by natural disasters using satellite imagery. Leveraging Satellite data and population datasets, it identifies anomalous regions, computes their potential impact based on both environmental and social factors, and outputs a priority-ranked list of clusters for response and monitoring.
Based on these data and rescue guidelines, we are able to generate very important data like interactive map for visualization, enabling real-time analysis and actionable insights. A report is generated to optimize disaster response, planning, and mitigation strategies.
Key Features
Anomaly Detection Using Satellite Data
- Compares radar backscatter between a baseline period and a recent period after the natural disaster.
- Computes Z-score anomalies for VV and VH polarizations.
- Identifies pixels with significant deviations, representing possible flooding, landslides, or other environmental changes.
Vectorization and Cluster Analysis
- Converts anomaly pixels into polygons for easier analysis.
- Calculates cluster-level statistics such as mean anomaly intensity, population affected, and pixel count.
- Prioritizes clusters based on a weighted metric combining anomaly severity and population exposure.
Priority Ranking and Color Mapping
- Ranks the top N clusters according to priority.
- Assigns colors from yellow → orange → red to visually indicate priority levels.
- Provides both a visual map layer and a CSV export of the top clusters for reporting.
Population Integration
- Ensures priority is higher for regions affecting more people.
Interactive Frontend
- Displays VV and VH anomalies, overlapping areas, and top-priority clusters.
- Supports zooming and pan, allowing users to explore affected regions interactively.
Backend API
- Provides endpoints for anomaly analysis, cluster extraction, CSV generation, and reporting.
Automated Reporting
- Generates structured reports summarizing priority clusters, population exposure, and anomaly severity.
- Helps organizations optimize planning, resource allocation, and mitigation strategies in disaster response.