EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

HSF-Net: A Hierarchical Edge Enhancement and Sparse-Aware Fusion Network for Small Object Detection in UAV Aerial Imagery

Jiaqi Li, Qinghua Zeng, Songnan Duan, Junjie Wu, Yilin Li, Yudi Sun, Qian Gao

Drones · 27.08.2026 · 10 (9) · 652

DOI: 10.3390/drones10090652 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Drones, online erschienen am 27.08.2026. Verfasst von Jiaqi Li und 6 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 42 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: the Defense Industrial Technology Development Program.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Object detection in unmanned aerial vehicle (UAV) imagery is severely challenged by extremely small object scales, cluttered backgrounds, and pronounced foreground–background imbalance, which jointly degrade the accuracy of general-purpose detectors. This paper presents HSF-Net, a small-object detection network built upon YOLOv11s through three complementary enhancements. First, a Hierarchical Edge Enhancement Module (HEEM) employs orthogonal strip-convolution decomposition with zero-initialized residual fusion to enhance and re-weight the fine-scale edge and texture cues that are progressively attenuated in deep convolutional backbones. Second, a Sparse-Aware Feature Modulation (SAFM) module replaces concatenation-based fusion in the top-down neck pathway, coupling a sparse foreground gate with channel-wise scale modulation to confine cross-scale aggregation to object-bearing regions. Third, the detection head is restructured from {P3, P4, P5} to {P2, P3, P4}, introducing a high-resolution pathway for tiny targets while removing the original low-resolution P5 branch. On VisDrone, HSF-Net attains 46.4% mAP50 and 28.5% mAP50–95, exceeding the baseline by 7.1 and 4.8 percentage points, respectively, while reducing the parameter count from 9.4 million to 3.6 million. The model achieves an end-to-end throughput of 131.3 FPS on an NVIDIA RTX 3090, although the high-resolution P2 branch increases the computational cost to 39.5 GFLOPs. After dataset-specific training and evaluation on the markedly different TinyPerson benchmark, HSF-Net outperforms YOLOv11s by 5.2 percentage points in mAP50, indicating that its relative performance advantage persists under a substantially different target-scale distribution and maritime background.

Förderung

  • the Defense Industrial Technology Development Program

Zitieren

APA

Li, J., Zeng, Q., Duan, S., Wu, J., Li, Y., Sun, Y., & Gao, Q. (2026). HSF-Net: A Hierarchical Edge Enhancement and Sparse-Aware Fusion Network for Small Object Detection in UAV Aerial Imagery. Drones, 10(9), 652. https://doi.org/10.3390/drones10090652

Vancouver

Li J, Zeng Q, Duan S, Wu J, Li Y, Sun Y, et al. HSF-Net: A Hierarchical Edge Enhancement and Sparse-Aware Fusion Network for Small Object Detection in UAV Aerial Imagery. Drones. 2026;10(9):652. doi:10.3390/drones10090652

BibTeX
@article{Li2026,
  author = {Li, Jiaqi and Zeng, Qinghua and Duan, Songnan and Wu, Junjie and Li, Yilin and Sun, Yudi and Gao, Qian},
  title = {HSF-Net: A Hierarchical Edge Enhancement and Sparse-Aware Fusion Network for Small Object Detection in UAV Aerial Imagery},
  journal = {Drones},
  year = {2026},
  volume = {10},
  number = {9},
  pages = {652},
  doi = {10.3390/drones10090652}
}
Kostenlos starten →