EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Weakly Supervised Parcel Delineation of Smallholder Tree Crops with Seasonal Radar-Optical Time Series

Lethabo Sithole, Rethabile Ndlovu

International Journal of Computational and Biological Sciences · 2026

DOI: 10.66238/ijcbs104 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Computational and Biological Sciences, 2026. Verfasst von Lethabo Sithole und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The accurate delineation of smallholder tree crop parcels is a critical component for ensuring global food security, promoting sustainable agricultural management, and enforcing zero-deforestation supply chain regulations. Traditional supervised deep learning approaches for agricultural boundary extraction rely heavily on exhaustive polygon annotations, which are prohibitively expensive and time-consuming to acquire, especially in heterogeneous tropical landscapes. This paper proposes a comprehensive weakly supervised learning framework designed to delineate smallholder tree crops using seasonal radar and optical time-series data. By leveraging sparse point-level annotations combined with classical region-growing heuristics, the framework generates reliable pseudo-labels that guide the training of a dual-stream convolutional neural network. The integration of Sentinel-1 synthetic aperture radar and Sentinel-2 optical imagery effectively mitigates the pervasive issue of cloud cover in tropical regions while capturing distinct phenological and structural signatures of tree canopies throughout the growing season. The methodology eliminates the need for manual boundary digitization without sacrificing spatial accuracy. Comprehensive experimental evaluations demonstrate that the proposed weakly supervised approach achieves delineation accuracy comparable to fully supervised methods. The synergistic fusion of seasonal radar volumetric scattering information and optical multispectral indices provides a robust feature space that enhances the separation of adjacent parcels. This research offers a scalable and cost-effective solution for large-scale agricultural monitoring, supporting international initiatives for sustainable agroforestry and transparent supply chain traceability.

Zitieren

APA

Sithole, L., & Ndlovu, R. (2026). Weakly Supervised Parcel Delineation of Smallholder Tree Crops with Seasonal Radar-Optical Time Series. International Journal of Computational and Biological Sciences. https://doi.org/10.66238/ijcbs104

Vancouver

Sithole L, Ndlovu R. Weakly Supervised Parcel Delineation of Smallholder Tree Crops with Seasonal Radar-Optical Time Series. International Journal of Computational and Biological Sciences. 2026. doi:10.66238/ijcbs104

BibTeX
@article{Sithole2026,
  author = {Sithole, Lethabo and Ndlovu, Rethabile},
  title = {Weakly Supervised Parcel Delineation of Smallholder Tree Crops with Seasonal Radar-Optical Time Series},
  journal = {International Journal of Computational and Biological Sciences},
  year = {2026},
  doi = {10.66238/ijcbs104}
}
Infokarte | EUVIMED

Artikeldaten derzeit nicht verfügbar.

Recherche