In order to develop a rapid, costeffective and highthroughput analysis method for detecting adulterants in milk, adulterated milk discriminant models are established by combing twodimensional (2D) correlation nearinfrared spectroscopy with multiway partial least squaresdiscriminant analysis (NPLSDA). 40 adulterated milk samples with melamine（0.01~3 g/L）and 40 adulterated milk samples with urea （1~20 g/L）are prepared respectively, then the absorption spectra of all samples are measured. Based on quantization of 2D correlation spectrum, the NPLSDA models of ureatainted milk, melaminetainted milk and two types adulterated milk are constructed. The recognition rates of unknown samples are 95%, 90%, and 92.5% by calibration models individually. At the same time, PLSDA and OPLSDA models are established. The results show that NPLSDA model has better predictive ability than PLSDA model and OPLSDA model, and this method can also be applied to other food safety detection areas.