Active Fire Detection using VIIRS Data

In this project, Level-1 VIIRS imageries from 2020 are used to detect active fire in California. Mid-infrared band (M13) is used to detect the hotspots and the geolocations are used to form fire clusters which are then vertorized into polygons.

Data Source: Images from the VIIRS instrument onboard the Suomi-NPP satellite from California on 2020-08-20 and 2020-08-21 Spectral Band: MWIR (M13) Band

Workflow

• Step 0: Exploratory Data Analysis
• Step 1: Data Preprocessing and Transformation
• Step 2. Filtering thermal anomaly for hotspot detection (Otsu's method)
• Step 3: Clustering of fire hotspots (DBSCAN Clustering Algorithm)
• Step 4: Vector generation (Convex Hull Algorithms)

Output:

1. Jupyter notebook for the full workflow
2. A CSV file on coordinates of the fire hotspots detected
3. A geojson file storing the cluster polygons

01.11.2021