Study On Detection Of Egg Powder Adulteration Based On Nirs And Hyperspectral Analysis

Egg powders are widely used in the food industry because of high quality protein,balanced minerals,vitamins and good functional characteristics.However,in order to reduce production costs and get greater profits,some business often incorporates cheap substances into egg powder.Thus,this study focused on egg powder(whole egg powder,egg white powder,egg yolk powder)and attempted to detect adulteration in binary adulteration system and multiple adulteration system of egg powder with near infrared spectroscopy and hyperspectral technology.At the same time,two kinds of spectroscopic techniques were used to establish the detection model of protein and fat content in the egg powder,the main results are as follows:(1)The adulteration in binary adulteration system of egg powder was detected by near infrared spectroscopy analysis.The adulterant starch,soybean meal and maltodextrin was added into three kinds of egg powder respectively and constituted the binary adulteration system.The qualitative discriminant model(partial least squares discriminant model)and quantitative detection model(partial least squares regression model)were established in adulteration detection of three kinds of egg powder.The results showed that the partial least squares discriminant model(PLS-DA)could be used to distinguish pure egg powder and adulterated egg powder in the discriminant analysis of three kinds of egg powder adulteration.In the quantitative model of whole egg powder adulteration,the best model for the detection of starch,soybean meal and maltodextrin was the regression coefficient-partial least squares regression model(RC-PLSR),and the correlation coefficient(Rp2)Reached at 0.990,0.996 and 0.998,respectively.In the quantitative model of egg white powder adulteration,the best detection model of soybean meal and maltodextrin was RC-PLSR,Rp2 reached at 0.962 and 0.979,respectively.For the model of egg white powder adulterated with starch,the performance of RC-PLSR was significantly lower than PLSR.Therefore,the PLSR model was selected and the Rp2 was 0.921.In the quantitative model of egg yolk adulteration,the performance of RC-PLSR model was poor.So,the PLSR model at the best band was selected and Rp2 reached at 0.998,0.997 and 0.986,respectively.(2)The adulteration in multiple adulteration system of egg powder was detected by near infrared spectroscopy.The quantitative model of starch,soybean meal,maltodextrin and the total adulteration content was established,respectively.The optimal conditions were selected by studying the effects of different pretreatment and different bands on each model.Based on the best pretreatment and optimal band,principal component regression(PCR)quantitative model was established and the performance between PLSR and PCR was compared.The results showed that the model of starch adulteration and maltodextrin adulteration were not good,while the predictive performance of soybean meal adulteration and total adulteration was good,and Rp2 reached more than 0.950.The performance of PLSR was better than PCR.(3)The hyperspectral technique was used to detect the binary adulteration system of egg powder.Firstly,the hyperspectral images of the samples and the adulterated samples were collected and the average spectra were extracted.The SVM model was used to distinguish the pure and adulterated egg powders.The results showed that the correct rate of adulteration of three kinds of egg powder(whole egg powder,egg yolk powder and egg white powder)reached more than 90%.In order to detect the content of adulterants quantitatively,the study used the PLSR model to establish the relationship between spectral data and adulterated content.In whole egg powder adulteration model,the Rp2 of starch,soybean meal and maltodextrin adulteration was 0.931,0.981 and 0.990,respectively.In egg white powder adulteration model,The Rp2 reached at 0.832,0.994 and 0.984,respectively.In the study of egg yolk adulteration,the Rp2 reached at 0.998,0.986 and 0.975,respectively.Then the simplified models RC-PLSR and SPA-PLSR were established by RC method and SPA method.The results showed that there was no significant difference in the performance of the simplified models,but the number of wavelengths was greatly reduced.RC-PLSR model showed a better performance.(4)The adulteration in multiple adulteration system of egg powder was detected by hyperspectral technique.In the model of egg powder adulteration,the accuracy of the predictive set of random forest model reached more than 90%.Compared with the SVM model,the results showed that the correctness of the random forest model was better.In the quantitative regression model,the Rp2 of the PLSR model of starch,soybean beal and maltodextrin reached at 0.986,0.992 and 0.989,respectively.In order to simplify the model,the RC-PLSR,RC-MLR and SPA-PLSR and SPA-MLR models were established based on the characteristic wavelengths.The results indicated that the simplified models were more suitable for detection requirements.(5)The quantitative model of protein and fat content in whole egg powder were established by near infrared spectroscopy and hyperspectral technique.For the near infrared spectroscopy,the optimal model for protein detection was the model established by the characteristic wavelengths based on RC method,Rp2 reached at 0.996,and the optimal model for fat detection was based on the model established by the full band,Rp2 reached at 0.974.For the hyperspectral technique,the optimal model for protein detection was the model established by the characteristic wavelengths based on loading factor method,Rp2 reached at 0.995,and the optimal model for fat detection was the model established by the wavelengths extracted by RC method,Rp2 reached at 0.964.