Wildfire detection for transmission line based on improved lightweight YOLO

Wildfires in transmission line passages are a severe threat to power security. Two wildfire detection models, which are based on the YOLOv5, are proposed in this paper. Due to the limited computing power of embedded terminals, the proposed models simplify the network structure of YOLOv5. Specifically, one is that only the overall structure of the neck and head parts in the original network structure is simplified, and the second method is based on the first method to delete the modules of backbone, which greatly reduces the number of parameters of the model. The experimental results show that the proposed lightweight models can achieve real-time monitoring in embedded devices while the accuracy and recall remain high.