Identification of liquid milk adulteration using Raman spectroscopy combined with lactose indexed screening and support vector machine

Abstract

Pasteurised milk is frequently adulterated with ultra-high temperature (UHT), posing significant safety risks for consumers. Raman spectroscopy combined with lactose indexed screening (LIS) and support vector machine (SVM) were used to rapidly identify pasteurised milk and UHT milk. Adulterated samples were prepared at 6 different mixing ratios of pasteurised milk and UHT milk for classification modelling. The characteristic Raman shift of lactose standard was used as the extraction index for processing the spectra data. The Maillard reaction between lactose and protein was confirmed using high-resolution mass spectrometry, which verified the correlation between lactose and heating treatment of liquid milk. After comparing four feature extraction algorithms, CARS coupled with SVM classifier showed better performance. The final LIS-CARS-SVM model achieved 100% classification accuracy in the test set, demonstrating the potential for the rapid and accurate identification of pasteurised milk and UHT milk, as well as detecting liquid milk adulteration in various forms.