https://github.com/arturogonzalezm/fire_smoke_detector
https://github.com/arturogonzalezm/fire_smoke_detector
Last synced: 2 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/arturogonzalezm/fire_smoke_detector
- Owner: arturogonzalezm
- License: mit
- Created: 2024-08-12T01:07:18.000Z (9 months ago)
- Default Branch: main
- Last Pushed: 2024-08-12T04:46:02.000Z (9 months ago)
- Last Synced: 2025-01-02T08:14:35.312Z (4 months ago)
- Language: Python
- Size: 10.7 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Fire Detection System
This project implements a real-time fire detection system using computer vision and machine learning techniques. It combines traditional computer vision methods with a deep learning model to detect fires in video streams.
## Table of Contents
- [Overview](#overview)
- [System Architecture](#system-architecture)
- [Installation](#installation)
- [Usage](#usage)
- [Key Components](#key-components)
- [Algorithm](#algorithm)
- [License](#license)
## Overview
The Fire Detection System uses a combination of computer vision techniques and a machine learning model to detect fires in real-time video streams. It processes each frame using both approaches and combines their results for more accurate fire detection.
## System Architecture
```mermaid
graph TD
A[Video Input] --> B[Frame Processing]
B --> C[Computer Vision Detection]
B --> D[Machine Learning Detection]
C --> E[Result Combination]
D --> E
E --> F[Fire Detection Decision]
F --> G[Tracking and Visualization]
G --> H[Display Output]
```
## Installation
1. Clone this repository:
```
git clone https://github.com/arturogonzalezm/fire_smoke_detector.git
cd fire_smoke_detector
```
2. Install the required dependencies:
```
pip install opencv-python numpy keras pillow scipy torchvision
```
3. Download the pre-trained model file `fire_smoke_detection_model.h5` and place it in the `models/` directory.
## Usage
Run the main script to start the fire detection system:
```
python main.py
```
Press 'q' to quit the application.
## Key Components
1. `FireDetectionModel`: A singleton class that loads and manages the machine learning model.
2. `KalmanTracker`: Implements Kalman filtering for smooth tracking of detected fire regions.
3. `detect_fire_cv`: Performs computer vision-based fire detection using color and motion analysis.
4. `update_tracked_fires`: Manages the tracking of fire regions across frames.
## Algorithm
The fire detection algorithm follows this sequence:
```mermaid
sequenceDiagram
participant Video Input
participant Frame Processing
participant CV Detection
participant ML Detection
participant Result Combination
participant Tracking
participant Display
loop For each frame
Video Input->>Frame Processing: Capture frame
Frame Processing->>CV Detection: Process frame
Frame Processing->>ML Detection: Preprocess and infer
CV Detection->>Result Combination: CV fire regions
ML Detection->>Result Combination: Fire probability
Result Combination->>Tracking: Combined detection
Tracking->>Display: Update fire regions
Display->>Video Input: Show processed frame
end
```
1. Capture a frame from the video input.
2. Perform computer vision-based fire detection:
- Create a fire mask using color thresholding.
- Detect motion by comparing with the previous frame.
- Combine color and motion information to identify potential fire regions.
3. Run the machine learning model on the preprocessed frame to get a fire probability.
4. Combine the results from both methods using a weighted sum.
5. Apply temporal smoothing to reduce false positives.
6. Update tracked fire regions using the Hungarian algorithm and Kalman filtering.
7. Visualize the results on the frame and display it.
## License
This project is licensed under the MIT License. See the [LICENSE](LICENSE) file for details.