Broadband diffuse reflection spectroscopy and Linear Discriminant Analysis enabling a non-destructive milk identification

Milk is one of the most important dietary products thanks to its high nutritional value and economic importance, but unfortunately also among the top products facing food fraud and adulteration. Authentication of milk is therefore of high importance, motivating the development of novel sensing techniques. We present the use of broadband diffuse reflection spectroscopy (400 - 1700 nm) in combination with Linear Discriminant Analysis (LDA) to achieve a rapid and non-destructive milk identification. Two case-studies are presented: (1) the differentiation between whole/high fat cow milk, semi-skimmed cow milk and skimmed cow milk, and (2) the differentiation between semi-skimmed regular cow milk and lactose-free milk. Considering a hold-out test set, classification performances exceeding 99.8% were achieved. Consequently, we believe these presented results contribute to an improved milk classification, enhancing food quality monitoring.