Dual-Modality Wildfire Detection with Visible and Infrared Images From UAVs

Wildfires pose significant risks to the environment and human safety. To address the limitations of single-modality detection, this paper proposes a dual-modality wildfire detection framework that integrates visible and infrared images captured by UAV-mounted cameras. Multimodal images are first registered and then processed through an enhanced YOLOv8based object detection model with a dual-stream architecture. Experimental results demonstrate improved performance over single-modality baselines in terms of precision, recall, and mean Average Precision (mAP). The proposed method highlights the effectiveness of dual-modality method for accurate and robust wildfire detection.