A rapid wine brand identification method based on the joint application of SERS and machine learning techniques

In this paper, an innovative approach is proposed to achieve no-preparation, rapid, and accurate identification of red wine brands by combining Surface-Enhanced Raman Scattering (SERS) spectroscopy with machine learning. SERS detects trace molecular signatures (e.g., polyphenols) in wine with high sensitivity. Spectra were preprocessed via principal component analysis, and a convolutional neural network (CNN) model performed feature extraction and recognition. Testing 18 wines (6 each from China, Chile, Italy) generated 1080 spectra. The model can achieve a maximum classification accuracy of 99.27 %. This SERS-CNN approach enables non-destructive, real-time brand identification in under 3 s per sample, offering portability and high efficiency. It holds significant promise for applications in wine brand protection, quality/safety monitoring, and consumer markets.