Sesame Yield Prediction Using Hyperspectral Reflectance: Determining Spectral Features and Their Timeline Trends

In this work, we use hyperspectral reflectance for a field-grown sesame (Sesamum indicum) dataset to determine the spectral features that best correlate with its yield post-harvesting. The spectral reflectance acquired at leaf and crop canopy levels for selected dates during the growing season. It provides us an understanding of the role of visible and near-infrared regions that can model sesame yield. The contrast among the red edge (starting from 700 nm), green reflectance (550 nm) with the blue and red absorptions (420 - 600 nm) indicates strong correlations determined using spectral band analysis, normalized indices, and random forest-based predictors' importance. The associations of these spectral features with the yield are high for a selected duration of sesame growth.