Characterization of winegrape berries’ composition on sorting tables using hyperspectral imaging and AI

Comprehensive evaluation of grape composition at winery receiving areas often requires multiple measurements to ensure representativeness, as well as the use of analytical techniques that are time-consuming and involve sample preparation. Recent advances in non-destructive sensing technologies, particularly hyperspectral imaging (HSI), offer promising alternatives for rapid and reliable grape quality assessment. In this context, the present study proposes a novel, non-invasive methodology for the characterization of winegrape composition directly on sorting tables. Specifically, hyperspectral imaging (HSI) in the visible to near-infrared range (4001000 nm), combined with multivariate statistical analysis and artificial intelligence (AI), was applied to estimate key compositional parameters including total soluble solids (TSS), pH, chromatic characteristics, anthocyanins, malic acid, tartaric acid, and yeast assimilable nitrogen (YAN). The highest predictive performance was obtained for pH (R²CV = 0.90), with malic acid (R²CV = 0.76) and total soluble solids (TSS; R²CV = 0.64) also showing strong predictive capacity. For the remaining parameters, models achieved moderate R²CV values (0.300.50), sufficient to support binary classification between high and low levels. These findings highlight the potential of HSI as a powerful approach for grape quality assessment and decision-making in outdoor settings, such as grape sorting tables.