The capabilities of vision systems used in meat processing

Vision systems are being used in an increasing number of applications within the meat industry. This is due not only to improvements in camera technology and hyperspectral technology, but also to the rapid development of artificial intelligence. Both technologies contribute to higher quality in production and in the final product that reaches the consumer.

What is machine vision used for in the meat industry?

The potential applications of machine vision in meat production are vast and continue to be developed. They can be divided into several categories:

1. Grading and sorting (detection and rejection of damaged cuts of meat; grading of carcass parts).
2. Detection of foreign objects (packaging residues, bone fragments, metal, plastic and glass fragments).
3. Detection of discolouration (in both raw meat and finished products).
4. Quality assessment (including meat freshness, colour, texture and sheen).
5. Measurements (volume, mass, shape, degree of oxidation, fat content, protein content).
6. Automation of slaughter lines (automatic cutting and portioning, carcass orientation recognition, livestock counting).
7. Packaging inspection (detection of leaks and contamination; checking the integrity of seams and seals, as well as labels, codes and dates).
8. Hygiene and cleanliness checks (detection of meat and fat residues on conveyor belts, knives and machinery; verification of cleaning effectiveness).
9. Streamlining intralogistics processes (identifying and classifying product batches by content).

One of the most interesting applications of computer vision in meat quality control is a study of beef conducted in Australia by Liao et al. Nearly 40,000 photographs of steaks were used to train the AI model. The study involved samples from local abattoirs. It turned out that the model was not only able to assess product quality better than a human, but also to determine the animal’s diet and breed [1]. The aim of the study was to create an effective tool to prevent fraud in the industry.

Do vision systems correctly classify every type of meat?

Whether it is meat from mammals, birds or fish, muscle meat or offal, machine vision is capable of accurately classifying all types of meat and detecting any anomalies. It is worth noting that the degree of processing (raw, minced, frozen, smoked meat, etc.) does not play a significant role here. This is made possible by:

• the use of artificial intelligence algorithms trained on datasets specifically designed for this implementation;
• creating suitable, stable conditions on the production line (particularly in terms of lighting);
• użyciu dostosowanych do aplikacji kamer w odpowiedniej liczbie (proste zastosowania wymagają zaledwie jednej kamery; te bardziej wymagające nawet kilku)Przykładowe aplikacje Scanway dla branży mięsnej i ich efekty

Examples of Scanway applications for the meat industry and their benefits

1. Quality control of chicken fillets

Using hyperspectral cameras, an AI model and specialised lighting, the system can identify haematomas and bruises, assess the colour of the meat, and detect fraying or the presence of bones in the meat. Thanks to these meat defect detection systems, the producer is able to remove lower-quality products right on the production line.

2. Checking the integrity of the seam on the cans

The system detects seam defects, edge damage, and the presence and orientation of the key. As a result, waste has been reduced and the quality of the product reaching the market has been improved. Find out more about the implementation.

3. Detection of plastic contaminants in fresh and frozen meat

A linear hyperspectral camera, dedicated lighting and AI algorithms were used for this task. The vision system reliably detected foreign objects such as films, packaging components and pieces of plastic. The colour and shape of the material were irrelevant to detection. This will enable safe processing and ensure a high-quality end product.

4. Classification of different types of small game

The measuring station is equipped with a vision system capable of classifying twelve different types of meat, based on an RGB camera, a capacitive sensor and trained neural networks. The classification accuracy was over 90%, and for some meats it was as high as 100%.

5. Distinguishing between cuts of pork half-carcasses

The solution is based on a single RGB camera and AI-powered image processing algorithms. The system distinguishes between right and left half-carcasses as they move along the conveyor belt. It then sends a signal to a device that directs them onto the appropriate conveyors for further processing.

6. Counting livestock on the unloading ramp

An RGB camera, supported by a deep learning model, counts the pigs moving along the slaughterhouse unloading ramp. Thanks to this technology, it can distinguish animals from other objects. The result is an accurate count of livestock and the elimination of losses at the slaughterhouse (over PLN 200,000 per year) caused by errors.

Automation in the food industry, including the meat sector, is becoming possible in increasingly sophisticated processes. The combination of vision technology and artificial intelligence modernises and streamlines the production of meat products, whilst also improving its profitability.

Sources:

[1] Liao, Q., Gardner, B., Barlow, R., McMillan, K., Moore, S., Fitzgerald, A., Arzhaeva, Y., Botwright, N., Wang, D., Nelis, J. (2025). Improving traceability and quality control in the red‑meat supply chain. Food Chemistry, 480.