Machine learning and optics in quality control

Modern quality control has more and more advanced tools at its disposal. One of them are vision systems based on the so-called deep neural networks.

Machine learning based solutions are more flexible and universal because:

  • the client can "teach" the system to recognize a possible defect by adding new reference images as needed.
  • as the dataset is trained and analyzed, the system itself recognizes potential deviations from the norm.
  • make it easier and more efficient to change products and formats during production (without additional downtime and unnecessary costs).
machine learning
Application of vision systems
Application of vision systems


Vision systems based on neural networks and the so-called machine learning is used in practically every industry. They are an ideal solution wherever defects / deviations from the norm cannot be defined "rigidly". This is the case, for example, with

  • a large number of variables,
  • a diverse range of products,
  • frequent changes in production - e.g. in the cosmetics or food industry,
  • the need to detect defects in heterogeneous (unique) products, such as concrete or wood (e.g. construction industry).

Why is it worth investing in intelligent quality control systems?

Building a competitive advantage.
Detection of defects not foreseen at the time of system installation.
The operator can easily add new reference images and "re-train" the system.
Can be used on high-speed production lines (capacity of up to 70,000 pcs/h).
Easy operation and shorter operator deployment time.
Classification of defects, and thus optimization of the technological process.
Easy integration of the system with the existing production lines.
Can be installed in almost any industry.