Study On Identification Of Egg's Varieties And Origins Based On Electronic Nose And Visible-nearinfrared Spectrum

With the gradual improvement of people's quality of life,the variety and origin of eggs requirements are also constantly improving.More and more peple are inclined to “Native eggs”,“Organic eggs” et al.Due to the high price of such eggs and driven by interests,illegal traders use Egg-hen eggs to serve as Native eggs,which not only harm the interests of consumers,but also undermine the fairness of the market.In addition,the phenomenon of the origin of the disorderly marked chicken eggs is serious in the market,which not only affects the healthy development of the market,but also is not conducive to the protection of famous Chinese original egg varieties and brands.Therefore,there is an urgent need for an accurate and rapid non-destructive detecting method for egg varieties and regions.In this study,we used electronic nose technology and visible-near infrared technology to measure the response data of volatile substances and visible-near infrared spectra of eggs from different varieties and origins.The differences of electronic nose data and spectral data of eggs from different varieties and origins are explored,and an identification model of egg varieties and origins based on electronic nose and visible-near infrared spectrum is established,and the software for detecting egg varieties and origins was designed and developed.The research content and conclusions are as follows:(1)To study the differences of odor substances in different eggs and establish the identification model.GC-MS technology was used to detect volatile components of eggs,there were significant differences in volatile components of eggs from different origins.Fox 4000 electronic nose system was used to obtain the volatile components of eggs from different breeds and producing areas.Through the analysis of data radar chart,the differences of volatile components of eggs from different breeds and producing areas were found.Spss was used to analyze the electronic nose data significantly,except 4 sensors showed no significant difference,the other 14 sensors showed significant difference.Partial Least Squares-Discriminant Analysis and k nearest neighbor algorithm were used to establish the model of variety and origin identification.PLS-DA modeling based on significant difference features was the best.The recognition accuracy of training set and prediction set was 100 % for variety and origin,and the cross validation accuracy of training set was 98.36 % and 97.65 %.(2)The identification model of egg varieties was studied and established.Build a spectral information acquisition platform,collect different varieties of eggs(Wuhan eggs,Wuhan native eggs,Xiaogan eggs and Xiaogan native eggs)spectral information.Standard normal variate was performed on the original spectral data to eliminate the effects of noise and baseline drift,and 35 characteristic wavelengths were selected by Competitive adaptive reweighed sampling.Then principal component analysis is used to reduce the dimension of the characteristic wavelength data,and the first two principal component information with cumulative contribution rate of 97.22 % are selected.An egg variety identification model was established by using Boosting regression trees algorithm,the correct rates of model training set and prediction set were 98.33 % and 97.00 %,and the cross validation accuracy of training set was 96.54 %.(3)The identification model of egg origin was studied and established.The spectral information of eggs from four different regions(Wuhan,Xiaogan,Enshi and Huangshi)were collected.Direct drthonormal signal correction was used to preprocess the original spectral data,and t-Stochastic neighbor embedding was used to reduce the dimension of the preprocessed data.The first four dimensions of information were selected as feature information.The original spectral data ware preprocessed using dosc,and t-SNE was used to reduce the dimension of the preprocessed data,and the former 4-dimensional information was selected as the characteristic information.A random forest was used to establish an egg origin identification model.The identification accuracy of training set and prediction set of the model was 100 % and 97.78% respectively,the cross validation accuracy of training set was 93.50 %,and the error rate of the out-of-bag error was 0.03.(4)Developing egg variety and origin detection software.On the basis of egg variety and origin identification model,the software design is completed by using the idea of Java and MATLAB mixed programming.Firstly,MATLAB software is used to write preprocessing,dimension reduction and modeling methods,and deploytool publishing tool is used to package and publish functions in the form of Java language.Secondly,Java programming is done with eclipse tools,and import the previously released function jars.Finally,the spectral data processing is realized by invoking function,using the database system to manage the data,to complete the rapid identification of the egg varieties and origin.This detection software has high stability.The average consumption was 30 ms,when one sample was processed.