Prediction of tea leaf characteristics using spectral data and machine learning techniques

Tea production is significantly impacted by abiotic and biotic stresses, necessitating early detection and diagnosis for effective management strategies. Tea leaf characteristics, like relative leaf water (RWC), leaf chlorophyll (LCC), and sugar contents, are significantly impacted by diseases and pest infestation. The remote sensing spectral reflectance has the potential to detect variations in leaf characteristics. This study predicts tea leaf characteristics and detects leaf infestation using spectral reflectance data and machine learning (ML) algorithms. Datt vegetation index and Meris Terrestrial Chlorophyll Index were good predictors of LCC and sugar contents. Random forest and eXtreme gradient boost performed well for predicting leaf chlorophyll and sugar contents, respectively. Support vector machine and Decision tree classifiers accurately identified infested leaves based on LCC and sugar contents, while logistic regression classifier classifies disease well using vegetation indices. The developed ML models will be helpful to map health condition of tea leaves in large areas.