Non-destructive prediction of pomegranate juice quality: near-infrared vs. mid-infrared spectroscopy

This study evaluated the application of non-destructive technologies such as Fourier transform near-infrared spectroscopy (FT-NIRs) and mid-infrared spectroscopy (FT-MIRs) in predicting the quality attributes of pomegranate juice ‘Wonderful’. Various pre-processing algorithms were applied to the spectral data, such as 1^st derivative, 2^nd derivative, multiplicative scattering correction, straight-line subtraction, vector normalization, and constant offset elimination. Partial least squares regression (PLSr) and principal component analysis (PCA) were used to develop regression models. Model’s performances were selected based on the pre-processing algorithms that provided the highest coefficient of determination (R²), the lowest root mean square error of prediction (RMSEP), and higher residual predictive deviation (RPD). The results suggest that pre-processing algorithms such as derivatives (1^st), vector normalization, and multiplicative scattering correction improved the model’s performance. FT-NIRs provided better prediction statistics for the total soluble solids content (TSS) (R²=0.92, RPD=3.62) and titratable acidity (TA) (R²=0.86, RPD=2.70) compared to FT-MIRs, TSS (R²=0.91, RPD=3.49) and TA (R²=0.82, RPD=2.40). While TSS:TA ratio (R²=0.81, RPD=2.35) was best predicted with FT-MIRs. Similarly, color components (a* and C*) were predicted in the MIR region with prediction statistics of (R²=0.855, RMSEP=3.22, and RPD=2.66) and (R²=0.827, RMSEP=4.17 and RPD=2.45), respectively. For vitamin C content, based on the RPD values, rough predictions for both FT-NIRs (RPD=1.85) and FT-MIRs (RPD=1.77) spectrophotometers were observed. The result of this study suggests that nondestructive technologies like Fourier transform near and mid-infrared spectroscopy can be implemented either at-line or online in the agro-processing industries to evaluate pomegranate juice quality attributes.