Non-destructive quick detection model based on Vis-NIR spectra for TVB-N of chilled beef

Abstract.This paper studied on using visible and near-infrared (Vis-NIR) spectroscopy and the mathematical modeling method to predict thetotal volatile basic nitrogen (TVB-N) of chilled beef non-destructively. The TVB-N is an important index of beef freshness, the detection of the beef TVB-N can achieve the purpose of detecting the freshness of beef quickly and non-destructively.The Vis-NIR diffuse reflectance spectra of chilled beef surfacewas obtained by spectrometer, and TVB-N was measured by Kjeldahl nitrogen determination apparatus. The multiplicative scattering correction (MSC) and Szvitzky-Golay (SG) smoothing method were used as pre-treatment methods for the raw sample spectral data. The prediction model of beef TVB-N was established by using partial least squares regression (PLSR), the most reasonable prediction model was chosen according to the correlation coefficient and standard error of prediction. The results showed that treating spectral data with the MSC after SG smoothing finished was the most efficient pre-treatment method with 7 variables selected were employed to establish the model of PLSR. The correlation coefficient (Rc) of the calibration set was 0.9467, calibration standard deviation (SEC) was 1.8847, the correlation coefficient (Rv) of the validation set was 0.8392, and the validation standard deviation (SEP) was 3.5736. The results demonstrated that predicting thebeef TVB-N with visible and near-infrared spectroscopy is feasible, and can obtain satisfactory prediction accuracy, laid the foundation for further development of online testing system.