Design and evaluation of a miniature milk quality detection system based on UV/Vis spectroscopy

A miniature milk quality detection system based on ultraviolet/visible (UV/Vis) spectroscopy was designed. One hundred and eighty milk samples were obtained, and each sample was divided into two parts. One part was utilized as fresh raw (FR) milk, and another part was processed using a high-pressure homogenizer. The UV/Vis absorption spectra of milk samples were collected, preprocessed, and analyzed. The models for predicting fat, protein, lactose, and total solids concentrations of FR and high-pressure homogenized (HPH) milk were established by using partial least squares regression algorithm. The accuracy of the system was evaluated with another 60 milk samples. The validation experiment showed that the root-mean-squares errors of the developed system were 0.35% %, 0.19 %, 0.13 %, and 0.46 % for fat, protein, lactose, and total solids concentrations of FR milk, and were 0.17 %, 0.14 %, 0.09 %, and 0.27 % for HPH milk. Homogenization can significantly improve the accuracy of the milk quality detection system.