Drone Fire and Smoke Detection System

Drone Fire Detection System (Project Dravion)

In response to the growing threat of forest fires, the "Dravion" project, primarily designed for preventive burning, also integrates a specialized software system for real-time environmental surveillance. A key component of this system is a dedicated script for the detection of fire.

Technical Implementation: fire_detection.py

The project utilizes three distinct Python scripts to process video data; one of these is specifically named fire_detection.py.

Unlike the project's person detection script, the fire detection module does not rely on an artificial intelligence model. Instead, it employs classic color-based image processing techniques to identify potential fire sources or bright hotspots ("foyers lumineux").

Detection Process

The fire_detection.py script follows a precise workflow for each image captured by the drone's camera:

• Color Space Conversion: The script first converts the captured image from the standard RGB model to the HSV (Hue, Saturation, Value) color space. This conversion allows for a more effective separation of color components.
• Color Masking: It then applies a color mask centered on the red, orange, and yellow hues that are characteristic of flames.
• Thresholding: The system analyzes the detected colored zones and filters out irrelevant noise. Only areas that exceed a minimum size (a surface area of at least 500 pixels) are considered relevant detections.

Output and Integration

Once a relevant fire source is identified, the script frames the area with a rectangle. The coordinates of this rectangle are then transmitted in JSON format, image by image.

This data structure allows the system to be combined with other scripts in the main interface. When a fire is identified, the user interface is designed to display a rectangle over the video feed with the label "feu" (fire). This architecture provides a comprehensive overview of the environment perceived by the drone in real-time.