Lightweight Wildfire Location and Recognition for Enhancing Resilience in Energy Internet Systems

Wildfires pose significant threats to energy internet and critical infrastructure, including power transmission lines, necessitating rapid and accurate detection to prevent extensive damage. However, implementing real-time wildfire smoke detection on embedded edge devices with limited computing power presents a considerable challenge due to their constrained resources. To address this problem, we propose a lightweight target detection model based on the YOLO framework, specifically designed for real-time mountain fire smoke detection on embedded devices. By simplifying the network structure of YOLOv5-s, particularly in the feature fusion and detection head components, our model reduces computational demand while enhancing performance. Compared to the original YOLOv5-s, our approach nearly doubles the parameter count and significantly improves recall and accuracy. Experimental results demonstrate that our lightweight model maintains high levels of accuracy and recall, achieving real-time monitoring capabilities on embedded edge devices, making it an effective solution for early wildfire detection in resource-constrained environments.