Microwave Sensing for Food Safety: a Neural Network Implementation

Customers’ attention in packaged food quality is arising in the last few years, as physical intrusions can still be found in commercialized products. Frightful newspapers articles attract consideration on this matter, and pose a serious health issue: accidental ingestion of foreign bodies can severely damage the digestive system, or even cause choking, with seniors and children being particularly exposed. Existing devices to monitor food products have lacks, missing certain classes of contaminant, low-density ones in particular, due to intrinsic limitations to their working principle, based on materials density in the case of x-rays inspection. Here, we propose a novel detection principle, based on microwave-sensing and exploiting the dielectric contrast between the potential intrusion and the surrounding matter. The realized microwave-sensing device is, then, combined with a machine-learning approach, with a classification mechanism capable in discerning clean from contaminated samples. The microwave-sensing device is applied to an industrial food production line, showing a remarkable precision in correctly detecting millimetric-sized intrusions made of plastic, glass or wood, which are the classes of materials unlikely to be located by existing inspection devices.