Real-time red meat freshness assessment using multi-mode spectroscopy and fusion-based machine learning

There is significant waste and loss in every stage of the food supply chain. Food loss, spoilage, and fraud in preparation and storage is the target of this proposal. We propose a hand-held Quality, Adulteration, and Traceability (QAT) device that can establish shelf-life for red meat, allowing management of storage, transportation and dynamic pricing of food close to expiration. We have previously demonstrated this technology’s performance for seafood. The device, combined with AI software, can identity 44 fish species to 95% accuracy and determine freshness within one day. In this project, we will expand its performance to red meat. Current methods for assessing freshness/shelf-life rely on either highly trained people or chemical analyses that involve sampling, one-time kits and readers. None of these methods provide an integrated solution that is accurate, rapid, and designed for field use. Spectroscopy and machine learning attracted attention recently as a method that is rapid and non-destructive. SafetySpect, in collaboration with University of North Dakota is developing an affordable multi-mode hand-held spectroscopy device, usable by unskilled workers, and equipped with fusion AI, to revolutionize an industry long due for a reliable, transparent and accurate measure of quality. The innovativeness of the proposed technology is three-fold. First, in hardware design and manufacturing, SafetySpect has made headway in manufacturing a miniaturized prototype that fits in the hand of an unskilled worker and is equipped with custom designed illumination. Second, it utilizes fusion AI to integrate three modes of spectroscopy that combined perform better than any one individually. Thirdly, an AI architecture based on a system of dispute models capable of improving the accuracy of the model was developed. Technical feasibility will take place in three stages. 1) Software: develop robust models using data collected throughout this project, 2) Multi-mode spectroscopy hardware: continue testing of the previously designed prototype, 3) Integration of hardware and software: SafetySpect has a successful track record of implementing AI algorithms on edge devices. Commercialization potential of this device due to the above-mentioned features has been determined through interviews with fish processing facilities as well as distribution, retail stores and restaurants. We intend to prepare the device for commercialization in the red meat industry as well. SafetySpect’s QAT device, by determining the remaining shelf life, can help reduce unnecessary meat waste by determining which lots to mark down for quicker consumption or send to processing facilities in more remote areas.