Non-destructive Freshness Assessment Of Shell Eggs Using NIR Spectroscopy

Eggs No-destruct detection researches include: egg weight, eggshell crack, blood spots, hatch, freshness and so on. The focus and difficulties detection is freshness of the study. Combined non-destructive testing technology of eggs and near infrared spectroscopy technology which is used to detect agricultural products, this study aims to explore and research feasibility and application prospects in detecting the egg freshness using near infrared spectroscopy non-destructive testing technology, and to develop a new method for the detection of egg freshness.In this research, diffuse reflectance spectrums of egg samples were obtained by mean of Fourier transform near-infrared (FT-NIR) spectrometer (VECTOR 33). Among these spectral data, those of 96 egg samples were analyzed and evaluated with modern metrology method. The specific details are as follows:Analysis of original spectral data of the egg samples using the principal components analysis method. The aim was to study the relationship among various principal components, and to point out corresponding information of egg near-infrared spectrum according to first principal components, second principal components, and third principal components, respectively.1. Removal of the outlier's samples. Compared the Multivariate Curve Resolution alternating least squares(MCR-ALS) iterative method with the traditional method in removed outliers samples, the result showed that the former had an advantage over the latter in removing the outliers samples of near-infrared spectrum.2. Original samples were divided into calibration samples set and validation samples set using spatial distribution analysis of principal component method.3. A comparing study to filter effects of near-infrared spectral data using the four methodologies (moving average smoothing, Savitzky-Golay filter smoothing, Gaussian filter and wavelet transform) was conducted. The results showed that wavelet transform was the best filter way among them and was able to effectively remove the high-frequency noise of the spectrum and to improve the model accuracy.4. A comparing study to the impact of the model using the pre-processing methodologies of spectra (baseline correction, multiple scatter correction, unit vector normalized, derivative and others) was conducted. The results showed that unit vector normalized is the best way of pre-processing methods of spectra. The model set up by PLS (partial least squares) in validation set produced a satisfactory result with a good coefficient of determination (R²=0.8247) of near-infrared spectrum and the egg freshness.5. To obtain and established the validation models of near-infrared spectrum and the egg freshness, the three methodologies (multiple linear regression, principal component regression, and PLS regression) were used in this study. The results showed that PLS regression provided a better accuracy than the others.6. After unit's vector normalized pre-processing, the near infrared spectrum was divided into 20 sub-intervals. With interval partial least squares, spectrum intervals that dominantly reflecting egg freshness features were extracted. The results showed that optimum spectrum intervals were validated from 7185.47 cm^-1 to 6788.20 cm^-1, and from 5580.98 cm^-1 to 4381.48cm^-1.7. To predict and sort the egg samples using partial least squares discriminant analysis, the model was set up with PLS regression analysis to predict and discriminate the egg freshness. The results showed that the accuracy of identification was 71%.The study showed that the near-infrared diffuse reflectance spectroscopy technology could reflect the internal quality of the egg, and result in a rapid non-destructive testing in detecting egg freshness in experimental conditions.