Microwave-Assisted Detection of Physical Intrusions in Commercial Food Packaged Products via Machine Learning

This study introduces a novel approach using microwaves (MW) combined with a machine learning (ML) classifier to detect physical intrusions within food packaging. Our objective is to validate our previous works [1], [2], and [3], by selecting real commercial water-based and oil-based food packaged products, specifically tomato and pesto sauce. The non-invasive MW sensing system is designed for real-time operation within a food production chain. Experimenting with both the support vector machine (SVM) and the multi-layer perceptron neural network (MLP) algorithms for binary classification, after training on datasets generated from scattering parameters acquired during measurements, resulting in a remarkable precision of 100% accuracy across 200 measurement samples for each food type. This outcome reflects the effectiveness of our previous findings.