AI Uncorks a Wine's True Identity in Seconds

Discerning a fine wine's true origin from a counterfeit can be a complex affair, but a new technique promises to identify a red wine's brand with remarkable speed and accuracy. Scientists have combined a powerful light-based analysis with artificial intelligence to create a system that can verify a wine in under three seconds.

The method uses Surface-Enhanced Raman Scattering (SERS), a highly sensitive technology that detects the unique molecular signatures of substances like polyphenols within the wine, creating a distinct spectral 'fingerprint'. This fingerprint is then analysed by a specialised machine learning model known as a convolutional neural network (CNN).

In a recent study, researchers trained their model using 1080 spectra from 18 different red wines originating from China, Chile, and Italy. The system proved highly effective, achieving a maximum classification accuracy of 99.27%. Because this approach is non-destructive and requires no sample preparation, it offers a portable and efficient tool for brand protection and quality control in the wine market.