Rapid authentication of the geographical origin of milk using portable near-infrared spectrometer and fuzzy uncorrelated discriminant transformation

The quality of milk is related to its geographical origin. In addition, the geographical origin of milk plays an important role in its commercial value. Therefore, this research aimed to design an identification system including a portable near-infrared (NIR) spectrometer and feature extraction algorithms to authenticate the geographical origin of milk. In order to improve the classification accuracy, fuzzy uncorrelated discriminant transformation (FUDT) was used to deal with the NIR spectra of milk collected by a portable NIR spectrometer. In this system, Savitzky-Golay (SG) algorithm and principal component analysis (PCA) were used to remove noise and reduce dimensionality, respectively. Then, three feature extraction algorithms, linear discriminant analysis (LDA), uncorrelated discriminant transformation (UDT), and FUDT were applied to separate the NIR spectra. Finally, K-nearest neighbor (KNN) classifier was utilized to assess the performance of this identification system. The results showed that the maximum classification accuracies of FUDT, UDT, and LDA were 98.67%, 97.33%, and 93.33%, respectively. The results confirmed the great potential for authenticating the geographical origin of milk by the combination of the portable NIR spectrometer and FUDT. Practical Applications The geographical origin of milk is vitally important to evaluate the milk quality in dairy market. Compared with traditional detection methods, NIR spectroscopy is considered to be a fast, accurate, and nondestructive detection method, so it is widely used in the field of food detection. In this article, FUDT with portable NIR spectrometer can be used to detect the adulteration of geographical origin in milk correctly and rapidly. The experimental result indicated that application potential in identifying the geographical origin of milk.