基于三维荧光光谱的清香型原酒品质鉴别模型

aking the light aroma and flavor types of based liquor of various grades after sensory oral evaluation as samples, the three-dimensional fluorescence spectra were measured for analysis. The results show that the fluorescence spectrum of the original wine is in 290nm/310nm (excitation wavelength/emission wavelength, Ex/Em, the same below) and 335/340nm, there are two fluorescence peaks, but the fluorescence intensity of the two fluorescence peaks is different between different wine samples . The original spectral data of 390 samples is used as the training set for support vector machine modeling, and the classification effect is poor. Parallel factors were used to analyze the differences between the spectra of the wine samples, and the emission spectra at excitation wavelengths of 290 nm, 310 nm, 340 nm and 360 nm were extracted as new characteristic quantities. After dimensionality reduction by principal component analysis and linear discriminant analysis, a support vector machine was used to model thefirstbased liquor and the second basedliquor , and the five-fold cross-validation method was used to verify that the modeling accuracy was 86.9%. Then 78 wine samples that were not involved in the modeling were selected for external testing. The accuracy of the test was 85.9%, and the identification effect was good. It can provide a rapid auxiliary manual oral evaluation method for the quality identification of based liquor.