YOLO11-RLN: An aerial UAV algorithm for forest fire detection

Abstract

To address the limitations of existing forest fire detection models, including suboptimal unmanned aerial vehicle adaptability, low detection accuracy, and high false detection rates, we propose a new a UAV-oriented forest fire detection algorithm: YOLO11 with RepVGG, long fire-line texture fusion attention, and nano optimization, Our approach introduces several key improvements: RepVGG replaces the YOLO11 backbone to enhance feature extraction and detection accuracy; a novel long fire-line texture fusion (LTF) module is designed to improve fire feature perception in complex forest environments; the WIoU loss function is integrated to enhance small fire detection and accelerate convergence; and YOLOv8-nano parameterization is employed to reduce model complexity and mitigate overfitting. Experimental results demonstrate that YOLO11-RLN outperforms YOLO11, achieving 7.338%, 5.392%, 7.862%, and 7.019% improvements in precision, recall, mAP50, and mAP50-75, respectively. Statistical significance analysis confirms the robustness of these improvements.