EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

LGD-PeakNet: a density-aware point-supervised framework for UAV-based wheat seedling counting

Bowen Zhang, Wenhao Zhao, Xiaolei Wang, Chenyu Xie, Shun Wang, Yinchao Che, Lei Shi, Xinming Ma, Shuping Xiong, Lei Xi

Smart Agricultural Technology · 2026-12 · 15 · 102594

DOI: 10.1016/j.atech.2026.102594 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Smart Agricultural Technology, 2026-12. Verfasst von Bowen Zhang und 9 weiteren Autor:innen. Die Arbeit nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Henan Provincial Science and Technology Research Project, Key Research and Development Program of Hunan Province of China, Key Scientific Research Project of Colleges and Universities in Henan Province.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate assessment of wheat seedling emergence and spatial distribution is essential for evaluating sowing quality and supporting early-stage field management. Unmanned aerial vehicle (UAV) RGB imagery provides an efficient approach for high-throughput seedling monitoring; however, reliable seedling counting remains challenging due to small target size, weak texture, complex background interference, and heterogeneous density distribution under field conditions. Existing low-annotation-cost methods often struggle to simultaneously achieve accurate counting and stable point-level prediction, particularly when fixed candidate response structures are applied to regions with different local densities. To address these challenges, this study proposes LGD-PeakNet, a density-aware point-supervised framework for UAV-based wheat seedling counting. The proposed method uses visible seedling center points as supervision signals and integrates local–global feature representation, multi-scale feature fusion, and density-aware adaptive sparse point response regulation. Specifically, local–global feature representation enhances discriminative seedling features under complex backgrounds, while multi-scale feature fusion improves feature integration across different spatial scales. Furthermore, the proposed adaptive sparse point response mechanism dynamically adjusts candidate response capacity according to local density variations, thereby improving point-level prediction quality under heterogeneous seedling distributions. Experiments were conducted on the CSD dataset, including low-density wheat, high-density wheat, manually sown wheat, and maize seedling imagery. The results demonstrate that LGD-PeakNet achieves competitive counting accuracy and point-level matching performance compared with representative density-based and point-supervised methods, while maintaining comparable model complexity and inference efficiency. In addition, the generated point-level predictions can be further aggregated into grid-based spatial representations, providing additional information for characterizing seedling distribution variability at the field scale.Overall, LGD-PeakNet provides an effective point-supervised framework for UAV-based wheat seedling counting, while the generated point-level predictions further support spatial characterization of seedling distribution in agricultural fields.

Förderung

  • Henan Provincial Science and Technology Research Project
  • Key Research and Development Program of Hunan Province of China
  • Key Scientific Research Project of Colleges and Universities in Henan Province

Zitieren

APA

Zhang, B., Zhao, W., Wang, X., Xie, C., Wang, S., Che, Y., Shi, L., Ma, X., Xiong, S., & Xi, L. (2026). LGD-PeakNet: a density-aware point-supervised framework for UAV-based wheat seedling counting. Smart Agricultural Technology, 15, 102594. https://doi.org/10.1016/j.atech.2026.102594

Vancouver

Zhang B, Zhao W, Wang X, Xie C, Wang S, Che Y, et al. LGD-PeakNet: a density-aware point-supervised framework for UAV-based wheat seedling counting. Smart Agricultural Technology. 2026;15:102594. doi:10.1016/j.atech.2026.102594

BibTeX
@article{Zhang2026,
  author = {Zhang, Bowen and Zhao, Wenhao and Wang, Xiaolei and Xie, Chenyu and Wang, Shun and Che, Yinchao and Shi, Lei and Ma, Xinming and Xiong, Shuping and Xi, Lei},
  title = {LGD-PeakNet: a density-aware point-supervised framework for UAV-based wheat seedling counting},
  journal = {Smart Agricultural Technology},
  year = {2026},
  volume = {15},
  pages = {102594},
  doi = {10.1016/j.atech.2026.102594}
}
Kostenlos starten →