Improved YOLOv11n-Based Small Object Detection Algorithm from UAV Perspective

With the rapid development of the low-altitude economy, drone technology is playing an increasingly important role in various fields such as environmental monitoring, agricultural protection, and emergency rescue. Aerial images captured by drones provide valuable data resources due to their unique perspectives and wide coverage. However, small object detection in drone aerial images remains a major technological bottleneck due to challenges such as small target sizes, inconspicuous features, and complex backgrounds. To address this issue, this study proposes an improved YOLOv11n-based small object detection algorithm to enhance detection accuracy from the UAV perspective. First, we introduce the C3k2-AD module, which enhances feature extraction through local perception, attention mechanisms, and multilayer perceptrons (MLP) to better capture object shape features at different viewing distances in UAV imagery. Second, to improve the model’s ability to detect small objects, we incorporate the SBA module in the Neck part of the model and add a P2 small object detection head in the Head section, thereby enhancing the model’s ability to distinguish and accurately detect targets in different regions. Experimental results on the VisDrone2019 dataset demonstrate that the improved YOLOv11n algorithm outperforms other models in small object detection accuracy. Specifically, the proposed model achieves an $\mathrm{mAP} \text{@} 0.5$ of $39.4 \%$, a precision of $48.8 \%$, and a recall of $38.6 \%$, representing improvements of $6.7 \%, 4.8 \%$, and $5.8 \%$, respectively, compared to the baseline model.