Abstract:

This study reports on the application of near infrared reflectance spectroscopy to determine the major chemical
components in milk. Totally 150 milk samples were collected from the same dairy farm, and the contents of protein, fat,
lactose, total solids and non-fat solids in milk were measured. Based on the spectral data, a regression model for each component
was proposed by partial least square regression combined with cross-validation. The results showed that middle- and
long- wavelength (1300–2500 nm) absorption contributed much to the regression models. The coefficient of determination
for the models for protein, fat, lactose, total solids and non-fat solids in validation sets were 0.96, 0.90, 0.90, 0.91 and 0.92,
respectively, with relative errors of 4.97, 3.19, 3.15, 3.37 and 3.54, respectively. The models for protein, lactose and non-fat
solids were effective in predicting 10 unknown samples, with relative error less than 1.40%.

Key words: milk, near-infrared transflectance spectroscopy, milk components, partial least squares