Abstract: Forty eligible pure milk samples were collected and 40 adulterated milk samples were prepared by adding urea with various concentrations (1－20 g/L). Two-dimensional correlation spectroscopy was calculated under the perturbation of adulteration. Based on the 2D correlation infrared spectroscopic characteristics, the spectra in the range of 4200－4800 cm-1  were selected for quantitative and qualitative by partial least squares. There results showed that a 100% recognition rate of samples was achieved by partial least square-discriminate analysis (PLS-DA). The correlation coefficient (R) of calibration sets was 0.999, and the root mean square errors of cross validation (RMSECV) were 0.242. The R between the predicted values and actual values was 0.999, and the root mean square errors of prediction (RMSEP) were 0.57. Therefore, the developed models have good prediction capacity.

Key words: near-infrared spectroscopy, milk, urea, adulteration, partial least square-discriminate analysis