Non-destructive Measurement Of Eggs Quality Using Near Infrared Spectroscopy

Along with the improvement of the society and people’s living standard, eggs are themajor source of nutrition during people’s daily life. The internal quality and egg freshness aregood or bad directly related to the impact of human body.Now, the grading and internal quality detection method of eggs have much defect such asdestroyed samples, time consuming, low efficiency. So we need to seek a testing accurate,rapid and efficient measurement technique.The non-destructive measurement technology using near infrared spectroscopy can berealized the internal nondestructive testing of eggs without destroying samples, since it isaccurate, efficient and fast. This paper combined nondestructive testing technology with theapplication of near infrared spectral theory in food testing. Internal quality (Haugh unit, theheight of albumen and albumen pH, etc.) served as the measurement index of egg quality. Thefast testing method and modeling of internal quality and freshness of eggs were studied usingnear infrared spectral nondestructive testing technology. The main research contents andresults were as follows:1. Egg samples corresponding to the best instrument matching parameters, measuredposition of samples, eggs shell color and other factors on the response characteristicsinfluence to eggs internal quality of near infrared spectral was measured under differentinstruments and collecting method. The research gained that the measured position of sampleshad effect on the eggs internal quality near infrared spectral response characteristics. Theequatorial region of egg was the best measured position which can reflect the near infraredspectral characteristics of eggs samples. The eggs shell color had little effect on eggsvisible/near infrared spectrum response characteristics.2. The original spectral data which was collected on diffuse reflectance by mean ofFourier transform near-infrared (FT-NIR) spectrometer and diffuse transmission by onlinevisible/near infrared spectrum detection device from eggs samples was pretreatment testingstudied respectively. Each of ideal near infrared spectral pretreatment method which wascorrespond to the eggs samples quality index was obtained from different instrument.3. With using the backward interval partial least squares (BiPLS), the optimal wavebandswere extracted from the spectrum of each eggs quality index after pre-processing. Researchshowed that on diffuse reflectance by Fourier transform near infrared spectrometer, Spectrumdivided into20,20,25,25,25sub-intervals and the variables were1714,1556,1371,1245,1381can dominantly reflex original spectral information of Haugh unit, egg weight, albumenpH, yolk index and the height of albumen, respectively. On diffuse transmission by online visible/near infrared spectrum detection device, Spectrum divided into20,15,25,25sub-intervals and the variables were658,879,431,479can dominantly reflex originalspectral information of albumen pH, Haugh unit, egg weight and the height of albumen,respectively. Obviously, quality index was different, corresponding to the sub-intervalsnumber and variables were different.4. Rapid and quantitative mathematical models were established for the near infraredspectral of every quality index, multivariate calibration methods of PLSR, MLR and PCRwere used. The result indicated that on diffuse reflectance by Fourier transform near infraredspectrometer, the best model established by partial least squares (PLS) for Haugh unit, eggweight, albumen pH and the height of albumen. The results of the correlation coefficient ofprediction (Rp) were0.854,0.844,0.726,0.878, respectively and the root mean square errorsof prediction (RMSEP) were9.595,3.295,0.323,1.077, respectively. And the multiple linearregression (MLR) was the ideal model for yolk index. The result of the correlation coefficientRp was0.909, RMSEP was0.036. On diffuse transmission by online visible/near infraredspectrum detection device, the best models of Haugh unit and the height of albumen wereestablished by PLS. The results of the correlation coefficient Rp were0.786and0.778,RMSEP were2.569and0.349. The best models of egg weight and albumen pH wereestablished by principal component regression (PCR). The results of the correlationcoefficient Rp were0.805and0.876, RMSEP were2.616and0.077. The results showed thatunder the quantitative analysis, the collecting methods were different, the quality indexeswere different, and the ideal models were different. The effect of models constructed with eggspectral information collected by Fourier transform near infrared spectrometer were superiorto the online visible/near infrared spectrum detection device.5. The qualitative analysis mathematical models of eggs samples freshness wereestablished. Samples were divided into six little batch and stored in0,3,6,9,12,15days,respectively. According to the Haugh unit value, freshness can divide eggs samples into freshand non-fresh. The qualitative analysis results which were obtained by partial least squarediscrimination analysis (PLS-DA) showed that on diffuse reflectance by Fourier transformnear infrared spectrometer, the identification rates of egg samples which were stored0,3,6,9,12,15days, were all100%, the identification rates of fresh eggs and unfresh eggs were both100%. On diffuse transmission by online visible/near infrared spectrum detection device, theidentification rates of egg samples which were stored0,3,6,9,12,15days, were100%,95%,88%,92%,86%,100%, respectively and the identification rates of fresh eggs and unfresheggs were81%and83%. The result indicated that the effect of near infrared spectrumcollected by Fourier transform near infrared spectrometer and the accuracy were better thanthose by online visible/near infrared spectrum detection device.