Forest Fire Detection and Recognition Using YOLOv8 Algorithms from UAVs Images

In recent years, forest fires have been frequent in many places, but there is a lack of effective and systematic methods to detect forest fires. Therefore, this paper proposes a forest fire detection strategy: a system that can be built with a large number of pre-processed and pre-trained datasets, receive images captured by UAVs and calculate the fire occurrence rate through algorithms and quickly pass them to users, thus speeding up possible rescue operations. For the system equipped with algorithmic models, this paper conducts an in-depth study of the YOLO series technology in image recognition. Based on the dataset of high-definition forest fire images captured by drones, the YOLOv8, YOLOv7 and YOLOv5 models are analyzed and compared from two perspectives: the accuracy of fire detection and the speed of model training, and it is found that YOLOv8 has the best accuracy and achieves a good balance between accuracy and computational speed. Based on this result, the proposed model is dependent on the nano-sized YOLOv8 algorithm, which has a fairly high accuracy, closely matches the original fire-containing test image, and still has a stable performance in detecting small fires. This algorithmic model in combination with the above system can help to accurately identify the occurrence of fires and thus mitigate the damage to forest resources.