An Enhanced YOLOv11 for Robust Object Detection in UAV Swarms

Abstract

Object detection in unmanned aerial vehicle (UAV) swarm environments remains challenging due to extremely small target sizes, dense distributions, and frequent occlusions. AI-though deep learning-based detectors have progressed, conventional architectures suffer from the loss of fine-grained spatial information caused by repeated downsampling, resulting in unstable localization and degraded detection performance. To address these limitations, this paper proposes an enhanced YOLOv11 framework for UAV swarm detection. The proposed method integrates a Fourier token-based plugin (FourierSR) into the standard C3k2 modules as F-C3k2 to expand the receptive field via frequency-domain operations. In addition, a high-resolution $P_{2}$ detection branch is introduced to preserve fine-grained shallow spatial features for tiny targets. Experimental results on the UAVSwarm dataset demonstrate that our framework substantially improves detection performance while maintaining real-time inference speed.