Author :
Bo Cai ; Chen, Huacai ; Zhang, Yongjun ; Jiang, Jiaxin
Author_Institution :
Coll. of Opt. Electron. Technol., China Jiliang Univ., Hangzhou, China
Abstract :
Two kinds of sampling methods of optics integrating sphere diffuse reflectance chamber and optical fiber diffuse reflectance probe were employed to collect the near-infrared spectra of 72 fresh lean pork meat samples, respectively. The amount of fat, protein and water in pork meat were determined by the national standard analysis methods as reference. The quantitative analysis models of the above constituents in pork meat were developed by using partial least squares (PLS) regression with internal cross-validation and optimized through spectra pretreatment and spectra region selection. The correlation coefficients (R2) of the calibration models of fat, protein and water based on integrating sphere diffuse reflectance were 90.24%, 91.95% and 90.15%, respectively. And the values of RMSECV of these models were 0.377%, 0.011% and 0.323%, respectively. For the models on optical fiber bundle diffuse reflectance spectra, the values of R2 were 83.68%, 87.97% and 83.27% respectively, the values of RMSECV were 0.321%, 0.0099% and 0.277%, respectively. The calibration models were further validated by the prediction sample sets. The prediction mean square error (RMSEP) of the three constituents were 0.254%, 0.011% and 0.168% for the models established on integrating sphere diffuse reflectance spectra, the value of R2 were 96.64%, 91.51% and 97.45%, respectively. For the models on diffuse reflectance optic fiber probe spectra, the values of RMSEP were 0.361%, 0.012% and 0.193%, respectively, the values of R2 were 81.03%, 84.48% and 91.39%, respectively. T test results showed that there was no significant difference between actual values and the values predicted by the NIR models. Although the optical integrating sphere got a little higher prediction accuracy and stability than the optical fiber diffuse reflectance probe, both sampling methods could meet the rapid non-destructive determination of the amounts fat, protein and water in po- - rk meat, and the latter is more suitable for hand-held and convenient.
Keywords :
food products; infrared spectra; mean square error methods; nondestructive testing; probes; reflectivity; regression analysis; sampling methods; NIR models; calibration models; correlation coefficients; fresh pork meat; national standard analysis methods; near diffuse reflectance spectroscopy; near-infrared spectra; nondestructive quality determination; optical fiber bundle diffuse reflectance spectra; optical fiber diffuse reflectance probe; partial least squares regression; prediction mean square error; quantitative analysis models; sampling methods; spectra pretreatment; spectra region selection; sphere diffuse reflectance chamber; Calibration; Integrated optics; Least squares methods; Optical fibers; Predictive models; Probes; Proteins; Reflectivity; Sampling methods; Spectroscopy;