Qualitative Analysis of Melamine in Fresh Milk Based on near Infrared Spectroscopy

The national standard stipulates that melamine content in infant milk products must not exceed 1 mg/kg. To rapidly and accurately identify melamine contamination in raw milk, this study collected near-infrared transmission spectra from 110 samples of raw cow's milk and 330 samples of melamine-contaminated milk within the 900-1700 nm wavelength range. Using 512 full-spectrum (FS) wavelengths, 10 effective wavelengths obtained through principal component analysis (PCA), and 11 effective wavelengths acquired via successive projection algorithm (SPA) as input variables, we established qualitative analysis models for melamine content using support vector machines (SVM) and partial least squares discriminant analysis (PLS-DA). Results demonstrated that the SVM model based on FS exhibited optimal predictive performance with a 96.36% accuracy rate. This study therefore proves that near-infrared spectroscopy technology can effectively identify melamine-contaminated raw cow's milk.