| Summary: | The advent of unmanned aerial vehicles (UAVs) has made a breakthrough in agriculture by capturing high spatial-resolution data of croplands, facilitating the mapping of important crop health parameters. The chlorophyll content is a strong indicator of crop health due to its crucial role in photosynthetic activity. In the present study, a hybrid Gaussian process regression (GPR) model was developed to estimate leaf chlorophyll content (LCC) in mustard crops at two distinct growth stages (60 and 90 days after sowing) in the research farm of the ICAR-Indian Agricultural Research Institute (IARI), New Delhi, India. Imaging hyperspectral data from UAV in the 400–1000 nm spectral range was compared with non-imaging data collected from the field at a 400–2500 nm spectral range. The radiative transfer model (RTM) PROSAIL-4 was used to generate the training data for training the GPR model. An active learning (AL) method and principal component analysis (PCA) were selected for dimensionality reduction and optimizing in sampling and spectral domains. On validating against in-situ LCC measurements, both datasets demonstrated competitive retrieval accuracies in terms of R2 greater than 0.8. The hybrid GPR models led to superior accuracies for the UAV-borne hyperspectral data (NRMSE of 9.15%), followed by field spectroscopy data (NRMSE of 10.54%). When the non-imaging data are spectrally resampled to UAV, they show a robust NRMSE value of 16.07%. This study achieved a UAV-based strategy for operational mapping of chlorophyll content over croplands, facilitating efficient and precise management of fertilizers for smart agricultural practices.
|
|---|