Determination Of Adulterated Honey Based On Near-infrared And Mid-infrared Spectroscopy Combined With Data Fusion

China is a big beekeeping country,and its honey export volume ranks first in the world.In recent years,however,China's honey reputation is negatively affected due to the serious honey adulteration,and making the export and domestic sales of honey are not good.Because of the shortcomings of the existing adulterated honey detection methods,such as time-consuming and labor-intensive,requiring chemical reagents,causing sample contamination,high cost and so on,the problem of honey adulteration has not been effectively curbed.Therefore,exploring a real-time,fast and accurate identification method of adulterated honey has very important practical significance and application value.In this paper,20 kinds of honey samples from 10 different provinces and 6 kinds of syrup samples commonly used in honey adulteration are the research objects.The NIR spectra and FTIR-ATR spectra combined with data fusion technique are used to establish an ideal identification method of adulterated honey.The main contents are as follows:1.Establish the identification model of adulterated honey based on NIR spectroscopy.NIR spectra of 112 honey samples and 112 adulterated honey samples with syrup are collected.Firstly,the NIR spectra are subjected to four kinds of spectral pretreatments(first derivative,second derivative,multiplicative scatter correction,standard normal variate transformation),and then combined with PLS-DA and SVM to establish the identification model of adulterated and authentic honey.The penalty parameter c and the kernel function parameter g of the SVM model are optimized by three optimization algorithms(grid search,genetic algorithm,particle swarm optimization).The accuracy of the best PLS-DA model is 87.50%,and the accuracy of the best SVM model is 94.64%.2.Establish the identification model of adulterated honey based on FTIR-ATR spectroscopy.The FTIR-ATR spectra of 112 honey samples and 112 adulterated honey samples with syrup are collected.Similarly,the FTIR-ATR spectra are subjected to four kinds of spectral pretreatments first,and then combined with PLS-DA and the parameter-optimized SVM to establish the identification model of adulterated and authentic honey.The accuracy of the best PLS-DA model is 89.29%,and the accuracy of the best SVM model is 91.07%.3.Establish the identification model of adulterated honey based on data fusion technique.NIR spectra and FTIR-ATR spectra are adopted low-level and intermediate-level data fusion approaches,respectively.In low-level data fusion,two types of spectral data are first subjected to four kinds of spectral pretreatments and then correspondingly fused.In intermediate-level data fusion,principal component analysis(PCA)is used to extract the principal components of pretreated NIR spectra and FTIR-ATR spectra,and then the first 10 principal components are correspondingly fused.After data fusion,the identification model of adulterated and authentic honey is established combined with PLS-DA and parameter-optimized SVM algorithm.With low-level data fusion,the accuracy of the best PLS-DA model is 94.64%,and the accuracy of the best SVM model is 91.07%.With intermediate-level data fusion,the accuracy of the best PLS-DA model is 94.64%,and the accuracy of the best SVM model is 100%.The results of this paper show that the NIR spectra and FTIR-ATR spectra combined with data fusion technique can establish a rapid,efficient and accurate identification method of adulterated honey,which is of great significance for maintaining the honey market order and protecting the legitimate rights and interests of consumers.