UAV-Based Retrieval Of Wheat Canopy Chlorophyll Content Using A Hybrid Machine Learning Approach

Hyperspectral data having ultrahigh spatial resolution captured using unmanned aerial vehicles (UAVs) facilitate operational delivery of canopy chlorophyll content (CCC) for efficient crop management in smart agricultural practices. In this study, a hybrid model of Gaussian process regression (GPR) trained using simulated data from PROSAIL was used to generate accurate predictions of the CCC of wheat crops. The sequential backward band removal (SBBR) algorithm and active learning method were selected as dimensionality reduction methods in spectral and sampling domains. The five most informative bands obtained from SBBR were used to develop the hybrid GPR model. On validated using in-situ field measurements, it shows a high prediction accuracy of $\mathrm{R}^{2}$ of 0.69 and an NRMSE of 16.90%. The prediction results suggest the use of a hybrid GPR model for the operational delivery of CCC estimates.