Technology for the Quantitative Identification of Dairy Products Based on Raman Spectroscopy, Chemometrics, and Machine Learning

Abstract

The technologies used for the characterization and quantitative analysis of dairy products based on Raman spectroscopy have developed rapidly in recent years. At the level of spectral data, there are not only traditional Raman spectra but also two-dimensional correlation spectra, which can provide rich compositional and characteristic information about the samples. In terms of spectral preprocessing, there are various methods, such as normalization, wavelet denoising, and feature extraction. A combination of these methods with appropriate quantitative techniques is beneficial to reveal the differences between samples or improve predictive performance. Quantitative evaluation can be divided into similarity measurement methods and machine learning algorithms. When evaluating small batch samples, similarity measurements can provide quantitative discrimination results. When the sample data are sufficient and matched with Raman spectroscopy parameters, machine learning algorithms suitable for intelligent discrimination can be trained and optimized. Finally, with the rise of deep learning algorithms and fusion strategies, some challenges in this field are proposed.

Keywords

• dairy products
• Raman spectroscopy
• quantitative identification
• chemometrics
• machine learning