Application Of Three Dimensional Fluorescence Combined With Quaternion Principal Components In Vinegar And Edible Oil Detection

In recent years,food safety problems have frequently occurred,which have seriously endangered people's physical health and social harmony.Therefore,it is significant for the detection of food safety.In order to solve the problems of traditional food safety with complicated detection process and large amount of equipment maintenance.This paper studies traceability of edible vinegar and rapid detection of edible oil doping based on three-dimensional fluorescence spectroscopy analysis technology and quaternion principal component analysis method.The main contents of this paper are as follows:First,the research of theories of three-dimensional fluorescence spectroscopy and quaternion signal processing algorithms at home and abroad are reviewed.Second,three-dimensional fluorescence spectra of 120 vinegar samples from four brands of Jiangsu Hengshun,Shandong Luhua,Shanxi Zilin and Tianjin Tianli were collected.Three-dimensional fluorescence spectra are preprocessed by a zero-setting method and a triangle interpolation method.The results are compared by parallel factor analysis and quaternion principal component analysis on vinegar brands.Among them,the classification accuracy of quaternion component principal component analysis combined with KNN classifier reached 100%.Then,three-dimensional fluorescence spectra of 240 oil samples from rapeseed oil,soybean oil,peanut oil,blend oil and corn oil are measured after being mixed with frying oil.Spectral data of the samples are expressed with quaternion parallel theory.And the feature is extracted and compared by quaternion principal component analysis algorithm and principal component analysis.A support vector machine classification model based on particle swarm optimization algorithm is established with the information after feature extraction to classify edible oil and adulterated frying oil.And,the classification accuracy of reached 100%.Finally,the three-dimensional fluorescence spectrum of 140 oil samples data of rapeseed oil,soybean oil,peanut oil,blend oil and corn oil are respectively analyzed by quaternion principal component analysis.The frying oil concentration predictive model is compared by partial least-squares regression and support vector regression algorithm.Among them,the correlation coefficient,mean square error and root mean square error of the model validation set based on support vector machine regression method can reach 0.9999,0.1163,and 0.341,which are better than the partial least squares regression method.