A Lightweight Anti-Unmanned Aerial Vehicle Detection Method Based on Improved YOLOv11

Abstract

Research on anti-UAV (anti-unmanned aerial vehicle) detection techniques is essential, since the widespread use of UAVs, while improving convenience, poses several hidden risks to privacy, security, air control, etc. This paper focuses on the challenges of long-distance UAV identification and proposes a lightweight anti-UAV detection method based on improved YOLOv11. Firstly, HWD is imported as the backbone’s downsampling module, which lowers feature loss in the feature extraction procedure while using fewer parameters. A lighter CCFM structure is then used in place of the original neck portion, to improve the model’s capacity to detect small targets and adjust to scale changes. The detection effect on small targets is greatly enhanced by removing the original large-scale detection head and adding a new small-scale detection head in response to the small size of UAV targets. In this paper, experimental validation was carried out using the DUT ANTI-UAV dataset, and, compared to the baseline model YOLOv11, the method we propose improved the P, R, mAP50, and mAP50-05 metrics by 4%, 4.5%, 4.1%, and 4.9%, respectively, and decreased the parameters by 38.4%. However, the FPS declined by roughly 5%. The experimental results show that the improved method we propose has better performance in anti-UAV detection tasks, and the model is more lightweight.