Identification of foreign materials in food using passive terahertz imaging and deep learning

Abstract

Food contamination is a major concern in the food industry as it can incur serious health risks to the consumers. For instance, perforation can occur when sharp objects are swallowed. Physical contaminants are the most noticeable type of food contamination and preventive measures can be put in place to ensure that the food products that are sold to the end consumers are free of foreign bodies. Some of the commonly found physical contaminants in food include glass, plastics, stones and metals. Most of the food manufacturers uses the conventional metal detector and X-ray imaging to filter out the contaminated food. However, there are limitations to the aforementioned machines, leading to complaints still being filed by the consumers regarding foreign materials being found even though measures already put in place to prevent such scenarios from happening. Not only does such experience cause psychological traumatism and danger to the consumer, business reputation would also be negatively affected. As such, terahertz imaging is proposed to aid the scanning of the finished food products, capturing foreign materials that the conventional scanning methods are not capable of. Application of passive terahertz imaging with continuous-wave to scan the food products is investigated in this project. The identification of foreign materials from the images obtained through passive terahertz imaging is automated through the application of deep learning. Results show acceptable accuracies for foreign materials detection. The limitations of the proposed method and further works that can be done for this method are also discussed in this report for future development to integrate in the industry.