Wu et al. (2026) UAV multispectral data and mixed-pixel segmentation for citrus water deficit diagnosis: Integrating spectral–texture features with machine learning for precision irrigation management

Identification

• Journal: Agricultural Water Management
• Year: 2026
• Date: 2026-08-06
• Authors: Zongjun Wu, Hongzhao Shi, Liwen Xing, Shenglin Wen, Ningbo Cui, Wei Zhao, Yaosheng Wang, Quanshan Liu, Shidan Zhu, Lu Zhao, Rangjian Qiu
• DOI: 10.1016/j.agwat.2026.110668

Research Groups

• State Key Laboratory of Hydraulics and Mountain River Engineering & College of Water Resource and Hydropower, Sichuan University
• Institute of Mountain Hazards and Environment, Chinese Academy of Sciences
• State Key Laboratory of Efficient Utilization of Agricultural Water Resources, Key Laboratory of Dryland Agriculture, Ministry of Agriculture and Rural Affairs of China, Institute of Environment and Sustainable Development in Agriculture, Chinese Academy of Agricultural Sciences
• State Key Laboratory of Water Resources Engineering and Management, School of Water Resources and Hydropower Engineering, Wuhan University

Short Summary

This study developed a novel citrus water status monitoring strategy by integrating UAV multispectral imagery and Gaussian mixture model based mixed-pixel segmentation algorithm. The results demonstrated that the Random Forest model achieved the highest prediction accuracy for estimating citrus water content during four critical growth stages.

Objective

• To implement a GMM-based mixed-pixel segmentation algorithm for precise extraction of citrus canopy pixels, thereby minimizing spectral interference from bare soil and weed backgrounds.
• To compare the performance of four representative ML models (ELM, SVM, RF, XGBoost) in estimating VWC under three different input feature schemes: spectral indices only, texture features only, and their fusion.

Study Configuration

• Spatial Scale: Centimeter-level spatial resolution using UAV-based multispectral remote sensing.
• Temporal Scale: Four critical growth stages of citrus (shoot-flowering, young fruit, fruit enlargement, and fruit maturation).

Methodology and Data

• Models used: Extreme Learning Machine (ELM), Support Vector Machine (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost).
• Data sources: UAV multispectral data, ground sampling points, and meteorological conditions.

Main Results

• The correlations between vegetation indices and vegetation water content were observed to vary across different growth stages.
• The spatial distributions of spectral vegetation indices and texture indices differed significantly.
• The Random Forest model achieved the highest prediction accuracy for estimating citrus water content during four critical growth stages.

Contributions

• This study provides a reliable technique for monitoring water deficit and supporting efficient agricultural water management.
• The results demonstrate the effectiveness of integrating UAV multispectral imagery and Gaussian mixture model based mixed-pixel segmentation algorithm for accurate vegetation water status assessment.

Funding

• This research was supported by the State Key Laboratory of Hydraulics and Mountain River Engineering & College of Water Resource and Hydropower, Sichuan University.
• The Institute of Mountain Hazards and Environment, Chinese Academy of Sciences provided additional funding.

Citation

@article{Wu2026UAV,
  author = {Wu, Zongjun and Shi, Hongzhao and Xing, Liwen and Wen, Shenglin and Cui, Ningbo and Zhao, Wei and Wang, Yaosheng and Liu, Quanshan and Zhu, Shidan and Zhao, Lu and Qiu, Rangjian},
  title = {UAV multispectral data and mixed-pixel segmentation for citrus water deficit diagnosis: Integrating spectral–texture features with machine learning for precision irrigation management},
  journal = {Agricultural Water Management},
  year = {2026},
  doi = {10.1016/j.agwat.2026.110668},
  url = {https://doi.org/10.1016/j.agwat.2026.110668}
}