Non-destructive quality estimation and internal pigmentation discrimination of ‘Sanguinelli’ blood oranges using hyperspectral imaging

Abstract

Blood oranges are gaining attention due to their vibrant red pigmentation and associated health benefits, primarily attributed to their high content of anthocyanins. These pigments not only enhance the fruit's visual appeal but also offer potent antioxidant and anti-inflammatory properties, which contribute to the prevention of metabolic disorders such as cardiovascular diseases and diabetes. As consumer demand increase, it is crucial for packing houses and breeding programs to non-destructively assess internal fruit quality parameters, including total soluble solids (TSS), titratable acidity (TA), maturity index (MI) and anthocyanin content. Understanding how climatic conditions influence these parameters is also essential for optimizing fruit quality and meeting market expectations. This study first investigates how chilling temperatures affect anthocyanin biosynthesis in ‘Sanguinelli’ blood oranges. It then aims to develop and compare models for the non-destructive prediction of key quality parameters, including TSS, total anthocyanin content and MI, using NIR hyperspectral imaging. Results indicated that early-season chilling hours were strongly correlated with increased anthocyanin accumulation. In contrast, late-season chilling was associated with a slight reduction in TSS content. Three machine learning models were employed for predictive modeling: Partial least squares (PLS), support vector machine (SVM), and a multi-layer perceptron (MLP). The prediction of TSS and anthocyanin content achieved R2 scores of 0.72 and 0.61, respectively. However, the predictions for TA and MI were not satisfactory. When assessing internal pigmentation, the best model, the MLP achieved accuracies of 79 % and 66 % for two and three different pigmentation groups, respectively. The MLP also achieved high accuracy for internal pigmentation discrimination and demonstrated good predictive capabilities for TSS and internal pigmentation. These findings confirm the potential of NIR as a rapid and reliable tool for the non-destructive quality control and internal pigmentation sorting of ‘Sanguinelli’ blood oranges.