An Ensemble Learning Method to Estimate the Leaf Nitrogen Concentration in Citronella Crop Using UAV-Multispectral Imaging

Remote sensing with unmanned aerial vehicles (UAVs) has emerged as a cutting-edge technology for precision crop management. Developing an accurate model to estimate Leaf Nitrogen Content (LNC) in citronella and predict nitrogen stress levels is essential for optimizing crop management systems. Nitrogen stress can be inferred from LNC using rapid, cost-effective, and non-destructive techniques such as spectral reflectance. A MicaSense Altum multispectral sensor was used to capture high-resolution image data of citronella (December 2022). This research addresses this limitation by employing three ensemble learning techniques bagging, boosting, and stacking, which can extract the best model by comparing these models. This framework integrates UAV-multispectral imaging vegetative indices (VIs) and ground sample data to assess nitrogen concentration in citronella leaves by employing ensemble learning methods. The results indicate the stacking ensemble model predicted LNC with the highest accuracy $\mathrm{R}^{2}$ of 0.8804 and RMSE of $0.3965\left(\mathrm{gm} / \mathrm{m}^{2}\right)$ for the citronella crop.