Packaged food products can hide contaminants that threaten product quality, causing product recalls or endangering consumers’ health. To meet this challenge, a team of students from Alta Scuola Politecnica – a two-year joint excellence programme of Politecnico di Milano and Politecnico di Torino – supported by Wavision s.r.l., a spin-off of Politecnico di Torino, have worked on the Wavisionproject, an innovative solution using microwave sensors and Machine Learning algorithms.
Food companies are producing ever larger volumes of packaged food to meet the growing and taking advantage of increased process automation, and this means a proportional increase in the risk of food contamination. Protecting consumer safety allows companies to retain customer confidence and safeguard brand reputation, which is crucial for them. It is therefore essential that any contaminants are detected before products are marketed.
The project’s underlying technology uses an innovative detection principle: microwaves can see the difference between the product to be inspected and any foreign body contained in it. When contamination occurs, the microwaves are altered in such a way that the algorithms developed by Wavision can detect the contamination. The capabilities of this system are an innovation aimed at overcoming the intrinsic limitations of already available devices, as the detection principle is based on a physical property that has never been considered for this purpose before, namely, dielectric contrast. For example, X-ray-based devices exploit the contrast in density between product and contaminant, and therefore, their detection capabilities only cover the most frequent classes of contaminants found in the food industry, such as plastics, glass, wood, etc.
The project now runs in five directions. Initially, it will work to improve the setup of the prototype for experiments, proposing an alternative that uses cheaper components without jeopardising efficiency. Next, the dataset will be expanded to strengthen robustness tests and improve the ability to identify contaminants. A theoretical analysis of biological contaminants will be also performed to identify prevailing ones. Advanced Machine Learning models will also be evaluated to improve detection accuracy and reduce calibration time. Finally, a trained Neural Network model will be introduced to detect and manage anomalies in the industrial production chain.
Much progress has been made but some research questions still remain open. However, the Wavision team’s innovative approach addresses many of the limitations of current contaminant detection methods, aiming to reduce costs and waste and ensure consumer safety. This advancement in technology could revolutionise the industry, ensuring safer products and consumer confidence.
