Kuang et al. (2026) Intelligent diagnosis of winter wheat water stress based on UAV multi-modal remote sensing

Identification

• Journal: Smart Agricultural Technology
• Year: 2026
• Date: 2026-01-02
• Authors: Xiaohui Kuang, Qian Cheng, Deshan Chen, W.-W. Fu, Hao Li, Zhen Chen
• DOI: 10.1016/j.atech.2026.101781

Research Groups

• State Key Laboratory of Crop Stress Adaptation and Improvement, Henan University, China.
• Institute of Farmland Irrigation, Chinese Academy of Agricultural Sciences (CAAS), China.
• The Shennong Laboratory, Zhengzhou, China.

Short Summary

This study develops a machine learning-based classification model for winter wheat water stress using UAV-mounted multispectral and thermal infrared sensors. The research demonstrates that fusing vegetation indices with canopy temperature data, particularly using a Support Vector Machine (SVM) algorithm, significantly improves diagnostic accuracy across critical growth stages.

Objective

• To develop a non-destructive, accurate, and rapid monitoring framework for winter wheat water stress by comparing multi-modal sensor fusion and machine learning algorithms (RF, XGBoost, SVM) across different phenological stages.

Study Configuration

• Spatial Scale: Field-scale experiment at the Xinxiang Comprehensive Experimental Base, Henan Province, China (35.2°N, 113.8°E), involving 120 experimental plots (1.2 m × 4 m each) with four irrigation levels (0 mm, 60 mm, 180 mm, and 300 mm).
• Temporal Scale: 2022–2023 growing season, with data collection during three critical growth stages: heading (April 20), flowering (May 6), and filling (May 23).

Methodology and Data

• Models used: Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Support Vector Machine (SVM).
• Data sources:
  • UAV-based multispectral imagery (MicaSense RedEdge MX) providing five spectral bands (Blue, Green, Red, Red Edge, NIR).
  • UAV-based thermal infrared (TIR) imagery (Zenmuse XT2) for canopy temperature.
  • Ground-truth plant water content (PWC) obtained through destructive sampling and oven-drying (85°C for 72 hours).
• Features: 12 Vegetation Indices (VIs) including NDVI, EVI, NDREI, and OSAVI, combined with TIR-derived canopy temperature.

Main Results

• Sensor Fusion: Multi-source data fusion (VIs + TIR) consistently outperformed single-source data (VIs-only or TIR-only) in classification accuracy across all stages.
• Model Performance: The SVM model was the most effective classifier. During the flowering stage, the SVM model achieved an accuracy of 97.22%, precision of 97.78%, and F1-score of 97.27%.
• Multi-stage Analysis: When merging data from all growth stages, the SVM model maintained the highest accuracy (82.41%) and F1-score (81.93%).
• Correlations: PWC showed strong correlations with remote sensing features, particularly during the filling stage, where NDREI reached a correlation of 0.94 and TIR reached -0.91.
• Optimal Stage: The flowering stage was identified as the most sensitive and accurate period for monitoring water stress.

Contributions

• Synergistic Monitoring: Establishes the superiority of combining spectral (structural/chlorophyll) and thermal (transpiration/stomatal) features for robust water stress diagnosis.
• Algorithm Optimization: Systematically identifies SVM as the superior classifier for high-dimensional, multi-modal UAV data in small-to-medium sample sizes.
• Precision Agriculture: Provides a practical, non-destructive technical framework for real-time irrigation decision-making and water-saving management in winter wheat production.

Funding

• National Key Research and Development Program of China (2023YFF1002200, 2023YFD1900705).
• Central Public-interest Scientific Institution Basal Research Fund (No. IFI2024-01).
• Science and Technology Research Project of Henan Province (242102110355).
• Natural Science Foundation of Henan Province (242300421101).
• Key Research Project of the Shennong Laboratory (SN02-2024-02).

Citation

@article{Kuang2026Intelligent,
  author = {Kuang, Xiaohui and Cheng, Qian and Chen, Deshan and Fu, W.-W. and Li, Hao and Chen, Zhen},
  title = {Intelligent diagnosis of winter wheat water stress based on UAV multi-modal remote sensing},
  journal = {Smart Agricultural Technology},
  year = {2026},
  doi = {10.1016/j.atech.2026.101781},
  url = {https://doi.org/10.1016/j.atech.2026.101781}
}

Generated by BiblioAssistant using gemini-3-flash-preview (Google API)