FFSN-DETR: A Feature-Enhanced UAV Object Detection Algorithm

With the rapid development of unmanned aerial vehicle (UAV) technology, its applications in urban security, traffic monitoring, disaster rescue, and agricultural inspection have become increasingly widespread. However, UAV imagery often faces challenges such as small target scales, severe occlusions, and complex backgrounds. Therefore, how to improve object detection accuracy while ensuring real-time performance has become a key issue in current research. To this end, this paper proposes a realtime object detection model with multi-scale feature fusion based on RT-DETR, named FFSN-DETR. Firstly, by optimizing and reconstructing the network backbone structure, the model’s feature extraction capability for small-scale targets from UAV perspectives is enhanced. Secondly, this paper proposes a novel multi-scale feature fusion module (MFFM) to enhance the model’s synergistic ability in extracting local details and global semantic information. Finally, through targeted optimization of the loss function, the model’s detection sensitivity and accuracy for smallscale targets are significantly improved. Experimental results show that, on the VisDrone2019 dataset, this method can maintain accuracy comparable to the original network while reducing the model parameters and computational load by $31.58 \%$ and $32.45 \%$, respectively.