LW-UAV–YOLOv10: A lightweight model for small UAV detection on infrared data based on YOLOv10
T.P. Nguyen, Nguyễn Long Giang, Duy D. Bui
Abstract
Advancements in unmanned aerial vehicle (UAV) technology have driven their widespread use in both civil and military sectors. Among various UAV types, small UAVs pose significant threats to global security, necessitating effective detection solutions. Real-time detection of small UAVs, especially under challenging conditions, remains a critical issue in computer vision. This study introduces LW-UAV–YOLOv10, an enhanced YOLOv10-based detection model optimized for small UAV detection using infrared data in mountainous terrain. Architectural improvements in the Backbone and Head modules enhance detection accuracy while maintaining a lightweight structure. Experimental results show that LW-UAV–YOLOv10 surpasses existing YOLO models in accuracy, speed, and suitability for real-time applications, offering a promising solution for UAV detection in complex environments. • Improved YOLOv10 model for detecting small UAVs on infrared data, bringing high efficiency. • Achieved outstanding accuracy in detecting small UAV targets in mountainous terrain conditions. • Provided real-time detection of small UAVs with fast inference speed and reliable results. • Provided effective solutions to address security challenges caused by small UAVs.