Short-wave near-infrared (SWNIR) technique was applied to determine the protein contents in milk powder. NIR spectra in the 800-1025nm region of 350 milk powder samples were analyzed. Based on the whole SWNIR spectra, least-square support vector machine (LS-SVM) obtained the best performance than partial least squares (PLS). Determination coefficient for prediction (Rp2) and root mean square error of prediction (RMSEP) were 0.98 and 0.23, respectively. Loading weights and regression coefficients of PLS and LS-SVM were used to determine the sensitive bands for protein content, as 906, 949, 987, 1002, and 1025nm. Based on these sensitive bands, Rp2 of 0.977 and RMSEP of 0.178 were obtained. It is concluded that some wavelengths at short-wave NIR region can be used to predict protein content in milk powder.
Content Determination of Proteins in Milk Powder Using Short-Wave Near-Infrared Spectroscopy
2008 Congress on Image and Signal Processing ; 5 ; 258-261
2008-05-01
321705 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Near-infrared spectroscopy locates epileptic sites
British Library Online Contents | 1998
Classification of Damaged Soybean Seeds Using Near-Infrared Spectroscopy
Tema Archiv | 2002
|