Zhang et al. (2025) A pixel-aligned co-registration and DSM-grid fusion framework for UAV multispectral and thermal imagery and point-cloud data: 3D Characterization of crop canopy water status

Identification

• Journal: Computers and Electronics in Agriculture
• Year: 2025
• Date: 2025-12-11
• Authors: Lechun Zhang, Pengchao Chen, Ruqiang Ma, Miao He, Haiyan Zhu, Yubin Lan
• DOI: 10.1016/j.compag.2025.111290

Research Groups

• National Key Laboratory of Automotive Chassis Integration and Bionics/School of Mechanical and Aerospace Engineering, Jilin University, Changchun 130022, China
• National Center for International Collaboration Research on Precision Agricultural Aviation Pesticides Spraying Technology (NPAAC), South China Agricultural University, Guangzhou 510642, China
• College of Electronic Engineering, College of Artificial Intelligence, South China Agricultural University, Guangzhou 510642, China
• College of Engineering and Technology, Jilin Agricultural University, Changchun 130118, China
• School of Agricultural Engineering and Food Science, Shandong University of Technology, Zibo 255049, Shandong, China

Short Summary

This study proposes a pixel-aligned co-registration and DSM-grid fusion framework to integrate UAV multispectral, thermal imagery, and point-cloud data for 3D prediction and visualization of cotton canopy leaf water content (LWC) and equivalent water thickness (LEWT), demonstrating its effectiveness for precision agricultural management.

Objective

• To develop and validate a pixel-aligned co-registration and DSM-grid fusion framework for integrating UAV multispectral, thermal imagery, and point-cloud data to enable three-dimensional prediction and visualization of cotton canopy leaf water content (LWC) and equivalent water thickness (LEWT).
• To evaluate and select an optimal interpolation method for low spatial resolution thermal imagery.
• To compare the performance of various machine learning models for predicting LWC and LEWT.
• To characterize the 3D distribution of canopy water status and assess the influence of nitrogen treatment.

Study Configuration

• Spatial Scale: Individual crop canopy (cotton), pixel-aligned grid cells, 3D point cloud units.
• Temporal Scale: Monitoring plant senescence, temporal sensitivity of water parameters.

Methodology and Data

• Models used:
  • Interpolation methods: Nearest neighbor, Bilinear, Bicubic.
  • Machine learning models: Random Forest (RF), Partial Least Squares Regression (PLSR), Support Vector Machine (SVM), Extreme Learning Machine (ELM).
  • Feature selection: Recursive Feature Elimination (RFE).
• Data sources:
  • Unmanned Aerial Vehicle (UAV) acquired data.
  • High-resolution point clouds (generated from Cross-circling oblique (CCO) photogrammetry).
  • Multispectral (MS) images.
  • Thermal imagery.
  • Digital Surface Model (DSM).
  • Vegetation and thermal indices (extracted from mapped features).

Main Results

• Bicubic interpolation was identified as the best method for thermal imagery downsampling-upsampling, effectively preserving spatial details and minimizing errors (quantified by RMSE and PSNR).
• The proposed framework, combined with RFE and Random Forest models, achieved R² values of 0.792 for LEWT and 0.752 for LWC, with relative root mean square error (rRMSE) values of 13.84% and 9.68% respectively, outperforming PLSR, SVM, and ELM models.
• A 3D representation of canopy water parameters revealed a typical top-down gradient of water loss in cotton.
• Nitrogen treatment significantly influenced the vertical distribution of water content, with high-nitrogen application delaying water loss in middle and lower canopy layers.
• LEWT exhibited higher sensitivity than LWC across both temporal and spatial scales, making it a more robust indicator for canopy water monitoring.
• Point clouds generated from Cross-circling oblique (CCO) photogrammetry outperformed UAV LiDAR systems in terms of point density, structural completeness, and image fusion potential.

Contributions

• Development of a novel pixel-aligned co-registration and DSM-grid fusion framework for integrating UAV multispectral, thermal, and point-cloud data for 3D crop water status characterization.
• Quantitative evaluation of interpolation methods for low-resolution thermal imagery in the context of UAV remote sensing.
• Validation of the feasibility and effectiveness of the framework for 3D crop water monitoring, providing a reliable technical foundation for drought detection, irrigation management, and yield prediction in precision agriculture.
• Identification of LEWT as a more robust indicator for canopy water monitoring compared to LWC across temporal and spatial scales.
• Demonstration of the superior performance of CCO photogrammetry for point cloud generation over UAV LiDAR for this application.
• Revelation of the coupled regulation between nitrogen and water in cotton canopy.

Funding

• Not specified in the provided text.

Citation

@article{Zhang2025pixelaligned,
  author = {Zhang, Lechun and Chen, Pengchao and Ma, Ruqiang and He, Miao and Zhu, Haiyan and Lan, Yubin},
  title = {A pixel-aligned co-registration and DSM-grid fusion framework for UAV multispectral and thermal imagery and point-cloud data: 3D Characterization of crop canopy water status},
  journal = {Computers and Electronics in Agriculture},
  year = {2025},
  doi = {10.1016/j.compag.2025.111290},
  url = {https://doi.org/10.1016/j.compag.2025.111290}
}