Integration of multispectral imaging with colorimetric sensor arrays for rapid grading and volatile component...

Based on the integration of multispectral imaging (MSI) with a colorimetric sensor array (CSA), this study developed a novel method to rapidly and non-destructively classify grades of Zhenjiang Aromatic Vinegar (ZAV) and predict key volatile components. Grayscale data from the CSA, captured at multiple spectral bands (400 nm, 465 nm, 530 nm, 590 nm, 650 nm, 730 nm, 850 nm, and 6500 K), was used to develop several classification models, including linear (PLS-DA, LDA, KNN) and nonlinear (CART, SVM, BPANN). Notably, the BPANN model outperformed others, achieving 100 % accuracy in the training set and 97.78 % in the prediction set. Quantitative models for acetic acid and n-hexanol were also developed using SVR and BPANN, with BPANN showing superior predictive performance (correlation coefficients of 0.9533 and 0.8741, respectively). These findings confirm that integrating MSI with CSA can effectively model volatile-driven colorimetric changes, enabling precise quality evaluation across fermented products.