Zhao et al. (2026) Estimation of crop canopy nitrogen content using deep transfer learning with PROSAIL-PRO model and UAV hyperspectral imagery

Identification

• Journal: Computers and Electronics in Agriculture
• Year: 2026
• Date: 2026-09-14
• Authors: Jing Zhao, Hong Li, Junping Liu, Wei Chen, Xin Guo, Yunlong Wu, Menglong Zhao, Junaid Nawaz Chauhdary, Zhaoxia Yan
• DOI: 10.1016/j.compag.2026.112385

Research Groups

• Research Center of Fluid Machinery Engineering and Technology, Jiangsu University
• Faculty of Agricultural Engineering, Jiangsu University
• School of Agricultural Engineering, Jiangsu University

Short Summary

This study proposes a novel inversion framework integrating the PROSAIL-PRO physical model with deep transfer learning to estimate crop canopy nitrogen content using UAV hyperspectral imagery. The results demonstrate that transfer-learning-optimized deep models significantly outperform traditional methods in terms of accuracy and generalizability.

Objective

• Investigate the feasibility of estimating crop canopy nitrogen content using UAV-based hyperspectral remote sensing and deep transfer learning

Study Configuration

• Spatial Scale: Field-scale
• Temporal Scale: Dynamic monitoring over time

Methodology and Data

• Models used: PROSAIL-PRO physical model, convolutional neural network (CNN), Residual Neural Network (ResNet18)
• Data sources: UAV hyperspectral imagery, simulated spectra from PROSAIL-PRO

Main Results

• The CNN-based transfer model achieved the best performance on the wheat dataset (R2 = 0.8597 and RMSECV = 1.7642), while the ResNet18-based transfer model achieved the best performance on the maize dataset (R2 = 0.6921 and RMSECV = 2.7944)
• Transfer-learning-optimized deep models significantly outperformed traditional methods in terms of accuracy and generalizability

Contributions

• The proposed physics-guided data-driven fusion approach provides a promising solution for crop nitrogen monitoring by ensuring high-precision inversion and robust generalization

Funding

• This research was funded by the National Natural Science Foundation of China (Grant No. 52177124)

Citation

@article{Zhao2026Estimation,
  author = {Zhao, Jing and Li, Hong and Liu, Junping and Chen, Wei and Guo, Xin and Wu, Yunlong and Zhao, Menglong and Chauhdary, Junaid Nawaz and Yan, Zhaoxia},
  title = {Estimation of crop canopy nitrogen content using deep transfer learning with PROSAIL-PRO model and UAV hyperspectral imagery},
  journal = {Computers and Electronics in Agriculture},
  year = {2026},
  doi = {10.1016/j.compag.2026.112385},
  url = {https://doi.org/10.1016/j.compag.2026.112385}
}