Chemometric Solutions | AI-Powered Food Fraud Detection with NIR

AI-Powered Food Fraud Detection with NIR Spectroscopy

Project Overview

Food fraud is a pervasive issue in today's market, posing financial risks and threats to public health. Common fraudulent practices include the adulteration of products such as Arabica coffee with lower-grade Robusta or Chicory, and the dilution or contamination of milk with substances like water or cleaning solutions. To address these challenges, we have developed an innovative solution leveraging cutting-edge Spectroscopy techniques, specifically Near-Infrared Spectroscopy (NIR) for ground coffee and Raman Spectroscopy for milk.

Innovative Solution

Our cutting-edge approach involves using Spectroscopy — specifically, NIR for ground coffee and Raman for milk — to obtain the spectral footprint of an unadulterated product. We then employ an Autoencoder trained on these spectra. During inference, we analyze the spectra of a sample and process it through the autoencoder, using the reconstruction error to ascertain whether the sample is pure or adulterated. For cases requiring quantification of adulteration, we use a regression model like PLSR (Partial Least Squares Regression).

Impactful Results

Our method has shown remarkable success in detecting adulteration:

In Arabica coffee, we could detect as little as 5% Robusta contamination. For Chicory, the detection level was even lower, at less than 1%. In milk, dilution with water was identifiable down to a 5% level. Impressively, cleaning detergents could be detected in any amount, even as low as 0.01%. These results not only demonstrate the precision of our approach but also highlight our commitment to ensuring food integrity and public health safety.