In-situ and non-destructive grape quality discrimination via field spectroradiometer and machine learning models

The grape composition impacts the subsequent wine quality. Timely non-destructive assessing grape quality demands rapid and highly accurate estimation, which is often quantified using spectroscopic analyses under the lab condition. However, non-destructive assessing grape quality in vineyard remains challenging. This work evaluated using field spectroscopy to assess total soluble solids (TSS) in grape non-destructively. We used in-field spectral database of Pinot Noir to develop classification models to discriminate grape TSS. Classification models of TSS had accuracy ranged from 0.73 to 0.9 for a test dataset. This study showed field spectroscopic analysis could discriminate grape TSS rapidly and non-destructively.