Gang et al. (2025) Plant-specific crop evapotranspiration estimation system for greenhouse tomatoes using convolutional neural network and rail-based monitoring device

Identification

• Journal: Computers and Electronics in Agriculture
• Year: 2025
• Date: 2025-10-16
• Authors: Min-Seok Gang, Hak-Jin Kim, Sung Kwon Park, Woo-Jae Cho, Tae-Hyeong Kim, Tae In Ahn, Joon Yong Kim, Kue-Seung Hwang
• DOI: 10.1016/j.compag.2025.111079

Research Groups

• Department of Biosystems Engineering, College of Agriculture and Life Sciences, Seoul National University, Seoul, South Korea
• Integrated Major in Global Smart Farm, College of Agriculture and Life Sciences, Seoul National University, Seoul, South Korea
• Smart Farm Research Center, Korea Institute of Science & Technology, Gangneung, South Korea
• Research Institute of Agriculture and Life Sciences, Seoul National University, Seoul, South Korea
• Department of Biosystems Engineering, Gyeongsang National University, Jinju, South Korea
• Department of Plant Science, College of Agriculture and Life Sciences, Seoul National University, Seoul, South Korea
• FarmOS Co., Ltd., Anyang, South Korea
• Kyung Nong Co., Ltd., Seoul, South Korea

Short Summary

This study developed a plant-specific crop evapotranspiration (ET) estimation system for greenhouse tomatoes, integrating a rail-based monitoring device with a convolutional neural network (CNN) for leaf area index (LAI) estimation and a simplified Penman–Monteith model, achieving high accuracy in both LAI and ET predictions.

Objective

• To develop and validate a plant-specific crop evapotranspiration estimation system for hydroponic tomato cultivation in greenhouses during the harvest season, enabling precise management and sustainable resource use.

Study Configuration

• Spatial Scale: Individual tomato plants within a hydroponic greenhouse cultivation system.
• Temporal Scale: Harvest season.

Methodology and Data

• Models used:
  • Simplified Penman–Monteith model (for evapotranspiration)
  • ResNet-based Convolutional Neural Network (CNN) (for LAI estimation)
  • You Only Look Once version 8 Nano (YOLOv8 Nano) (for object detection and rapid image acquisition)
• Data sources:
  • Side-view RGB images of individual tomato plants acquired by a rail-based crop-monitoring device.
  • Solar radiation distribution measured by the rail-based device.
  • Leaf area index (LAI), air temperature, and relative humidity (inputs for the ET model).

Main Results

• Strong correlation between estimated and measured LAI: R² = 0.89, Root Mean Square Error (RMSE) = 0.06.
• Strong correlation between predicted and actual evapotranspiration values: R² = 0.88, RMSE = 26.43 g h⁻¹ plant⁻¹.
• The developed system successfully generated distribution maps for LAI and evapotranspiration, demonstrating its capability for accurate plant-specific assessment.

Contributions

• Development of a novel plant-specific crop evapotranspiration estimation system that integrates a rail-based monitoring device with a CNN-based LAI estimation and a simplified Penman–Monteith model.
• Enables accurate, individualized assessment of evapotranspiration, supporting precision crop management in greenhouse cultivation.
• Facilitates improved productivity and reduced resource consumption by allowing early detection of plant heterogeneity and targeted treatments.

Funding

• Funding information is not available in the provided paper text.

Citation

@article{Gang2025Plantspecific,
  author = {Gang, Min-Seok and Kim, Hak-Jin and Park, Sung Kwon and Cho, Woo-Jae and Kim, Tae-Hyeong and Ahn, Tae In and Kim, Joon Yong and Hwang, Kue-Seung},
  title = {Plant-specific crop evapotranspiration estimation system for greenhouse tomatoes using convolutional neural network and rail-based monitoring device},
  journal = {Computers and Electronics in Agriculture},
  year = {2025},
  doi = {10.1016/j.compag.2025.111079},
  url = {https://doi.org/10.1016/j.compag.2025.111079}
}