Non-destructive determination of grape berrysugar concentration using visible/near infraredimaging and possible impact on wine quality

Reducing heterogeneity in harvest material may be
beneficial for wine quality and this goal may be achieved through
advanced berry sorting systems. The general aim is to assess if a
relationship could be found between berry sugar concentration
and hyperspectral images to determine the possible impact on
wine quality. Grapes were picked at different stages of maturity
in a one-year time interval and the berries were sorted according
to their size and density. Hyperspectral images of the groups
were obtained in the vis/NIR wavelength range with a complete
spectrum from 400 nm to 2500 nm. Our results showed that
vis/NIR images can be used as a tool to improve the segregation
of berries from all tested grape varieties based on their sugar content.
The PLSR algorithm is trained on all grape varieties together
and later validated on each variety separately, proving the possibility
of using a general regression model with constant parameters
to predict sugar concentrations. Finally, the impact on quality
was tested for red wines. Pinot noir berries with higher sugar
concentrations presented more color since anthocyanin concentration
was higher. Nevertheless, tannin concentration in skins
and seeds tended to decrease with increasing sugar concentration.
Groups with higher sugar concentration resulted in wines
with higher anthocyanin and lower tannin concentration.