Leveraging UAV Multispectral Imaging and Machine Learning for Sustainable Nitrogen Management

TY - CONF

T1 - Leveraging UAV Multispectral Imaging and Machine Learning for Sustainable Nitrogen Management

AU - Zhou, Yeying

AU - Ata-Ul-Karim, Syed Tahir

AU - Wang, Sheng

AU - Andersen, Mathias Neumann

AU - Cammarano, Davide

PY - 2025/4/29

Y1 - 2025/4/29

N2 - Nitrogen (N) is an essential nutrient for crop growth, influencing photosynthesis, protein synthesis, and overall metabolic activity. Efficient N management is essential for sustaining crop productivity while minimizing environmental impacts. Traditional N management in fields typically relies on farmers’ experience or generic N recommendations, neglecting in-season crop N demand, which can result in inefficient use of N fertilizers. UAV-based multispectral imaging combined with machine learning, offers a high-throughput, data-driven approach to assess crop N status. The objective of this study is to enhance maize N status assessment by identifying key vegetation indices and integrating them into machine learning models for improved predictive accuracy. By leveraging UAV-derived spectral data at critical growth stages, this approach refines N diagnosis and supports more precise fertilization strategies tailored to spatial and temporal crop needs. Among 50 different vegetation indices, RETVI (R2 = 0.72) and NREI (R2 = 0.79) had consistent correlations with nitrogen nutrition index (NNI) at early growth stages. Random Forest outperformed other algorithms in estimating NNI, N concentration, and biomass, achieving D-index between 0.83-0.93 and rRMSE of 4.19%-35.40%.

AB - Nitrogen (N) is an essential nutrient for crop growth, influencing photosynthesis, protein synthesis, and overall metabolic activity. Efficient N management is essential for sustaining crop productivity while minimizing environmental impacts. Traditional N management in fields typically relies on farmers’ experience or generic N recommendations, neglecting in-season crop N demand, which can result in inefficient use of N fertilizers. UAV-based multispectral imaging combined with machine learning, offers a high-throughput, data-driven approach to assess crop N status. The objective of this study is to enhance maize N status assessment by identifying key vegetation indices and integrating them into machine learning models for improved predictive accuracy. By leveraging UAV-derived spectral data at critical growth stages, this approach refines N diagnosis and supports more precise fertilization strategies tailored to spatial and temporal crop needs. Among 50 different vegetation indices, RETVI (R2 = 0.72) and NREI (R2 = 0.79) had consistent correlations with nitrogen nutrition index (NNI) at early growth stages. Random Forest outperformed other algorithms in estimating NNI, N concentration, and biomass, achieving D-index between 0.83-0.93 and rRMSE of 4.19%-35.40%.

M3 - Poster

T2 - START conference: Green Minds Gather 2025

Y2 - 29 April 2025 through 30 April 2025

ER -