Video quality control of poultry fillets as part of production safety

Poultry meat production is currently taking place under increasing pressure: increasing processing scale, high quality requirements and limited availability of skilled workers. At the same time, industry data show that 3-4% of all meat production is lost already at the processing stage due to technological errors, human errors and process problems. In high-volume conditions, this means real losses of raw material and the risk of reduced safety of the final product.

The scale of production and exports further reinforces the importance of the problem. According to preliminary data from the Ministry of Finance, in January-August 2025, the value of Poland’s meat product exports increased by 19% year-on-year and reached €8.4 billion, equivalent to about 22% of total agri-food export revenues. On such a scale, even single quality errors can have not only financial, but also image and regulatory consequences.

Vision system for inspection of poultry fillets

The vision system from the CUMULUS group is used for automatic inspection of raw material in the form of chicken fillets (breasts), transported on a two-track conveyor belt. The system works with a capacity of up to 130 pieces per minute, which allows integration into high throughput lines without limiting the production rate.

The main task of the system is:

• Detection of foreign bodies on the surface of meat,

• Identification of quality defects in the raw material,

• Unambiguous classification of the product into OK / NOK categories in real time.

Functional scope of the system

The system is capable of detecting a wide range of contaminants and quality defects that are important for food safety. As part of the foreign body inspection, the system identifies the presence of plastics such as PET, PE and PP, as well as metal components made of stainless steel and aluminum. Fragments of protective gloves, rubber and textile elements, as well as films – both colored and transparent – are also detected. In addition, the system allows detection of foreign bodies of natural and technological origin, such as wood, glass, paper, insects, and fragments of bone, feathers and cartilage. In parallel, quality defects in poultry fillets are inspected. The system detects blood stains and meat discoloration, fraying, and detects bone and skin residues that can affect the quality and safety of the final product.

Detection range of vision system for quality control of poultry fillets

The system operates continuously and each piece of meat is inspected. The entire vision system, including cameras and illuminators, is housed in enclosures that meet the requirements of the food industry, made of stainless steel and rated IP69K.

Foreign body detection – hyperspectral imaging

Plastic contamination detection is implemented using a hyperspectral camera. This technology captures the spectral signatures of materials, making it possible to clearly distinguish plastic from raw material – regardless of color, transparency or degree of dirtiness.

Analysis of spectral data and classification of detected objects is carried out using artificial intelligence algorithms (Machine Learning / Deep Learning), which allows adaptive adjustment of the system to the variability of raw material and production conditions.

Detection of quality defects

Identification of fillet defects is carried out using line cameras, working with dedicated industrial lighting. The system achieves defect detection efficiency of more than 98%, ensuring high repeatability of raw material quality assessment.

Operating principle and integration into the production line

Poultry fillets are transported on a conveyor in a linear fashion, with constant spacing between products. The passage of a single fillet through the inspection zone activates a presence sensor, which initiates the image acquisition process.

Once the data recording is complete, the system performs real-time analysis and classification. Each fillet is assigned to one of two classes:

• OK – the product meets the specifications,

• NOK – quality defect or contamination detected.

In the case of NOK classification, the system generates a signal indicating that a quality defect or contamination has been detected. This signal can be used for optional integration with an external rejector responsible for removing nonconforming product from a further stage of the process. At the same time, information about the detection of NOK classification is recorded in the system, allowing reporting on the number and type of nonconformities within a production batch. NOK events are also signaled on the operator’s panel, providing the operator with real-time insight into the status of the process and a quick response if established quality thresholds are exceeded.

Limitations of manual control in poultry processing

Work in meat processing plants is physically demanding and fraught with many environmental inconveniences. High staff turnover and labor shortages make it increasingly difficult to maintain stable, repeatable quality control based on manual inspection alone. Manual inspection of poultry fillets is prone to operator fatigue, subjectivity of evaluation, and limited ability to detect minor defects and contaminants at high line speeds.

In response to these challenges, automated vision-based quality control systems capable of continuous operation with repeatable accuracy and without human influence are playing an increasingly important role.