Support Vector Machine for Multiclass Contaminant Classification in Food Products using Microwave Sensing

This paper introduces an innovative approach using microwaves (MW) enhanced with a machine learning (ML) clas- sifier to identify products contaminated with external intrusions and distinguish the contaminant type. The non-invasive MW sensing system is intended to function in real-time in a food production chain. For the multi-class classification, we use the Support Vector Machine (SVM) algorithm, training it on datasets generated from the scattering parameters acquired during the measurements phase. A precision of 100% accuracy is achieved for 200 measurement samples.