Hyperspectral Fingerprinting for Plant Health

This paper investigates the efficacy of hyperspectral imaging and various vegetation indices for assessing plant health via predicting yield. We compare Bandwise and Pixelwise application methods for index calculation using the DeepPotato and HyperLeaf datasets. Results show strong correlations between specific indices and yield metrics: GNDVI ($R^2$=0.837) and CIGreen ($R^2$=0.836) were highly correlated with tuber dry mass in DeepPotato, while CIRed ($R^2$=0.868) showed the strongest link to grain weight in HyperLeaf. These findings underscore HSI's potential to provide valuable, non-destructive insights into crop health, serving as crucial proxies for yield prediction in precision agriculture.