Multispectral UAV imaging and machine learning for estimating wheat nitrogen nutrition index.

Introduction: Monitoring nitrogen nutrition indices is crucial for assessing current wheat growth conditions and guiding nitrogen fertilizer application. Methods: To estimate the wheat nitrogen nutrition index (NNI) and explore the effects of planting density and nitrogen application rates on NNI, this study employed UAVs to capture multispectral canopy imagery of wheat at key growth stages (tillering, jointing, booting, and filling) under varying planting densities and nitrogen application rates. Vegetation indices were selected using Pearson correlation and feature importance analysis. A Bayesian optimized random forest model was constructed to estimate the NNI. Results: Experimental results indicate that vegetation indices DVI, MDD, NGI, MEVI, NDVI, EVI, and ENDVI exhibit strong resistance to interference, enabling the construction of highly robust models. The NNI estimation model developed under nitrogen application level N2 (210 kg/hm Discussion: The research findings systematically reveal the patterns of planting density and nitrogen application levels affecting wheat NNI. The constructed NNI estimation model plays a crucial role in assessing wheat growth status and also provides reference for rationally determining planting density and nitrogen application levels for spring wheat.