An Embedded Portable Lightweight Platform for Real-Time Early Smoke Detection.

Area of Science:

• Computer Vision
• Embedded Systems
• Fire Safety Engineering

Background:

• Advanced smoke detection algorithms require significant computational power, limiting portability and efficiency.
• Existing solutions often necessitate personal computers or workstations, hindering real-time, on-site applications.

Purpose of the Study:

• To design and develop a lightweight, portable remote smoke front-end perception platform.
• To enable efficient, real-time smoke detection and monitoring in challenging environments like forests.

Main Methods:

• Utilized a Raspberry Pi with a Linux operating system for the platform's core.
• Trained a Local Binary Pattern (LBP) cascade classifier using the Adaboost algorithm in OpenCV with public datasets.
• Integrated four modules: video input, target detection, display, and alarm for comprehensive functionality.

Main Results:

• The developed platform demonstrated strong robustness and high detection accuracy on test datasets.
• Real-time smoke detection and dynamic display of results were achieved.
• The system successfully sent warning messages for detected smoke, enabling immediate alerts.

Conclusions:

• The portable platform provides an affordable and efficient lightweight smoke detection option.
• It overcomes the limitations of computationally intensive algorithms by using embedded systems.
• The system is suitable for practical applications such as forest fire monitoring.