EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Field-scale soil salinity mapping from UAV multispectral imagery by disentangling soil and vegetation spectral responses

Yuting Deng, Lu Xu, Ying Yang, Xifeng Zhang, Siheng Cao, Tingting Kang, Jianguo Li

Smart Agricultural Technology · 2026-11 · 102737

DOI: 10.1016/j.atech.2026.102737 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Smart Agricultural Technology, 2026-11. Verfasst von Yuting Deng und 6 weiteren Autor:innen. Die Arbeit nennt 74 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Soil salinization severely constrains sustainable agricultural development by degrading soil properties and inhibiting crop growth, making accurate soil salinity monitoring essential for farmland quality assessment and management. However, previous farmland-scale soil salinity inversion studies have commonly relied on full-image analysis, which mixes soil and vegetation signals and may increase errors in sparsely vegetated area. To address this issue, this study proposed a soil salinity inversion framework integrating UAV multispectral imagery and image segmentation. Soil and vegetation pixels were separated through image segmentation to construct soil, vegetation, and original-image datasets for characterizing their respective spectral response patterns. Pearson correlation coefficient (PCC), recursive feature elimination (RFE), and feature importance ranking (FI) were used for feature selection, while XGBoost, Random Forest Regression, and Support Vector Regression were applied to build soil salinity inversion models. Results showed that soil-pixel-based models achieved higher accuracy and stability than vegetation-pixel-based and original-image-based models. The XGBoost model using the soil dataset and RFE-selected variables performed best, with R², RMSE, RPD, and Lin’s CCC values of 0.800, 0.180 dS m⁻¹, 2.239, and 0.870, respectively. SHAP analysis indicated that salinity-sensitive information was mainly derived from visible, red-edge and near-infrared bands. Overall, this study highlights the critical role of accurate soil pixel extraction in reducing vegetation spectral interference and improving model accuracy, providing an effective framework for high-precision UAV-based monitoring of salinized farmland under vegetation-covered conditions.

Förderung

  • National Natural Science Foundation of China

Zitieren

APA

Deng, Y., Xu, L., Yang, Y., Zhang, X., Cao, S., Kang, T., & Li, J. (2026). Field-scale soil salinity mapping from UAV multispectral imagery by disentangling soil and vegetation spectral responses. Smart Agricultural Technology, 102737. https://doi.org/10.1016/j.atech.2026.102737

Vancouver

Deng Y, Xu L, Yang Y, Zhang X, Cao S, Kang T, et al. Field-scale soil salinity mapping from UAV multispectral imagery by disentangling soil and vegetation spectral responses. Smart Agricultural Technology. 2026:102737. doi:10.1016/j.atech.2026.102737

BibTeX
@article{Deng2026,
  author = {Deng, Yuting and Xu, Lu and Yang, Ying and Zhang, Xifeng and Cao, Siheng and Kang, Tingting and Li, Jianguo},
  title = {Field-scale soil salinity mapping from UAV multispectral imagery by disentangling soil and vegetation spectral responses},
  journal = {Smart Agricultural Technology},
  year = {2026},
  pages = {102737},
  doi = {10.1016/j.atech.2026.102737}
}
Kostenlos starten →