AI, Imaging Methods Could Improve Meat Inspection Processes
The UK Food Standards Agency (FSA) recently published the “21st Century Meat Inspector” project report. The project investigated how new and existing inspection technologies can be combined with advanced data analytics to improve UK meat inspection practices, with a specific focus on poultry.
According to the project report, poultry is the most widely consumed meat in the UK. However, poultry inspection is performed manually and is challenging to execute due to time constraints conflicting with the thoroughness and concentration necessary to inspect each bird that is processed at a facility. For context, the report states that one facility slaughters 250,000 birds per day, with only two full-time meat inspectors responsible for the inspection of each bird.
The project focused on the post-mortem inspection of poultry. A benefits-realization approach was utilized to determine the requirements for implementing new technology, as well as to ensure that business benefits are delivered to all stakeholders within the poultry supply chain.
One main finding of the project relates to the requirements for any new digital technologies that are implemented to assist meat inspectors in poultry facilities. The identified requirements are: provides clear benefits to the business, is robust and reliable, and is easy to use and clean.
The project also found that deep learning—a type of machine learning and artificial intelligence (AI)—can successfully identify the presence of abnormal color on meat products. However, a sufficient number of carcass images must be presented to the program to train it. More efficient data-labeling methods are also required with the use of deep learning.
Additionally, the project found that hyperspectral optical and X-ray imaging methods can effectively identify quality issues, including common problems such as wooden breast and white stripe in chicken breasts. However, further study is necessary to fully understand the potential of hyperspectral and X-ray imaging for the detection of quality issues in chicken breasts. The report suggests future work that utilizes larger numbers of samples, explores multi-sensor fusion activities, and addresses the technical and economic barriers to developing efficient systems for production environments.
Overall, the project’s findings suggest that the use of AI, sensor, and data analytic technologies can provide business benefits to regulators, meat inspectors, food businesses, and farms. However, the report notes that there would need to be large organizational changes made to realize the benefits, such as process redesign, job redesign, and investments in the design and development of new technologies. The report also acknowledges that new technologies would need to be thoroughly tested prior to implementation, and that it would be challenging to replicate a facility’s production process for testing purposes. The multitude of stakeholders in the food supply chain may also make it difficult to achieve consensus on standards for data access, data governance, responsibility and liability for automated decisions, and committing financial resources.
The project report recommends several focus areas for continuing studies related to the improvement of the UK’s meat inspection practices through the use of technology and data analytics. Such focus areas include: image collection during post-mortem inspection, infrared imaging for birds that are dead on arrival, large-scale laboratory and pilot-scale studies with X-ray and hyperspectral systems for the detection of quality issues, the design of effective business strategies for human and AI-augmented meat inspection in complex stakeholder environments, and novel production line testing environments for AI-enabled meat inspection.
Looking for quick answers on food safety topics?
Try Ask FSM, our new smart AI search tool.
Ask FSM →
