Abstract: This experiment used a hyperspectral image system in the wavelength range of 400–1 000 nm to predict the total viable count in sausage. Spectral data of 450 sausage samples were selected as the training set, and another 50 samples as the test set. The spectra was preprocessed by multiplicative scatter correction (MSC) method and principal component analysis (PCA) was used to reduce the dimensionality of the spectral data. Support vector regression (SVR) and gradient boosting decision tree (GBDT) were separately used to establish quantitative analysis models for the training and test sets, and the optimal model was selected. The results showed that the GBDT models were better than the SVR models. The root mean square error (RMSE) of the GBDT models were 0.001 and 0.003 for the training and test sets, respectively, and the coefficients of determination (R2) were 0.998 and 0.996, respectively. This study proved that the GBDT method based on hyperspectral imaging technology was feasible and effective to predict the total viable count in sausage.

Key words: hyperspectral imaging technology, sausage, total viable count (TVC), support vector regression (SVR), gradient boosting decision tree (GBDT)