Kimchi is a traditional Korean dish made from fermenting vegetables. The fermentation process is crucial for enhancing its quality and flavor during storage. Approaches such as hyperspectral imaging (HSI) and chemometrics (PLS, partial least square; SVR, support vector regression) including principal component analysis (PCA), and 2-dimensional correlation spectroscopy (2D-COS) can detect key physical and chemical components and changes in total soluble solids (TSS), pH, titratable acidity (TA), salinity, and lactic acid bacteria (LAB). Multivariate analytical models were developed to predict the quality properties using full and characteristic wavelengths and preprocessed data. The results showed that the ratio of prediction to deviation (RPD) values of the PLS prediction model constructed using the full wavelengths of TSS, salinity, pH, TA, and LAB were 1.57, 2.33, 2.79, 2.91, and 2.73, respectively. The Savitzky Golay 1st derivative preprocessed SVR model established based on characteristic wavelengths (951, 1020, 1139, 1174, 1216, 1321, and 1384 nm) extracted by PCA and a 2D-COS matrix showed the best results and increased efficiency in predicting pH (Rp2 = 0.9166, RPD = 3.281) and the number of LAB (Rp2 = 0.8488, RPD = 2.466). Additionally, the visualization process accurately illustrated the distribution of various quality indicators of kimchi across different periods. These results demonstrate that our proposed HSI strategy successfully assessed the degree of kimchi fermentation.
Read full abstract