Abstract: In order to realize the rapid and real-time monitoring of the changes in the number of microorganisms on cucumbers during storage based on the endogenous fluorescence information of microorganisms on them, and to provide a basis for predicting microbial spoilage, three-dimensional fluorescence information was collected on the surface of cucumbers at different storage times. The polynomial interpolation method was used to remove the Rayleigh scattering in the original spectra, and Savitzky-Golay polynomial smoothing and denoising were performed. Meanwhile, in order to avoid the metabolism of microorganisms on cucumbers from interfering with the fluorescence signal analysis, the core consistency diagnostic (CORCONDIA) method was used to estimate the number of components. By using the alternating trilinear decomposition (ATLD) algorithm, the three-dimensional fluorescence array was decomposed according to the number of components, yielding the relative excitation intensity spectrum, relative emission intensity spectrum and relative concentration matrix of each component. The results showed that the number of components was 4 when the function value was more than 60%. Among the four components obtained by the ATLD algorithm, components 1 and 3 had a special bimodal structure, and their characteristic excitation and emission spectra were consistent with the fluorescence fingerprints of tryptophan and tyrosine as the major endogenous fluorescent substances in microorganisms, and the tryptophan represented by component 1 had a high fluorescence quantum yield. The fluorescence region integration method was used to quantitatively analyze the total fluorescence of tryptophan-like regions with high and low excitation in the characteristic spectrum of component 1, and the functional relationship between the integral value of standardized region and the number of microorganisms was constructed by multiple stepwise regression analysis. The regression results showed that the determination coefficient of the regression model constructed by two-element four-power stepwise regression analysis was 98.309 8%. The prediction model was validated using samples not involved in its establishment, and the relative error of the obtained number of microorganisms was 1.037 1%. These results indicated that the model for monitoring changes in the microbial load on the surface of cucumbers during storage could provide basis for monitoring vegetable spoilage in real time.

Key words: three-dimensional spectral information; microbial quantity; alternating trilinear decomposition; fluorescence regional integration; multiple stepwise regression; monitoring model