Abstract: The principal components analysis performed on the centered data was used to study the near-infrared transflectance reflectance spectra of milk. It eliminated the interference of water. The new variables identified the characteristic wavelengths of fat, protein and lactose very well. The spectral range 2030～2350 nm was studied by principal components analysis. The principal component regression was used to predict the concentrations of fat,protein and lactose of the samples with standard predicted errors 1.61,1.94,2.08 g/L respectively.

Key words: near-infrared transflectance reflectance spectrum, milk, nondestructive analysis