Machine vision-based quality control

With these systems, waste can be cut down by detecting errors in production lines in advance, thus improving the finished product to speed up the production process.

Machine vision systems are often used in harsh environments, in fact thanks to them it is possible to inspect very small objects that would be difficult to access, thus ensuring greater uniformity of treatment and accuracy, compared to an inspection done with the human eye. The machine vision system thus detects product defects accurately and precisely.

Deep learning is the most current expression of artificial intelligence, thanks to special technological learning it has the ability to find solutions to problems caused by varying conditions. A system based on deep learning, can learn progressively, when the artificial neural network is sufficiently educated, so it is able to reason in a similar way to the human brain, increasing the capabilities of the neural network, without going to deteriorate the balances that have already been achieved in terms of selective abilities and judgment.

Therefore, systems that exploit deep learning tend to train themselves by adapting a behavior according to a pattern that goes to simulate human attitude-a “human-like” operation.

When computer vision is not used for quality control processes we fall back on irregularities in fabric texture, which can be slight color variations even on a plastic object. There can be small defects, which is why we could range far and wide among a wide variety of industries, talking about countless products where machine vision is not implemented in production processes where aesthetic checks are performed by dedicated personnel. However, the defect sought represents only an anomaly that is purely aesthetic, something the eye would not want to see, or it could be a functional anomaly given by a variation that compromises the physical properties of the object itself.

Aesthetic defects are nothing more than irregularities that are mistaken for natural variations present in the appearance of an object, they may be the thickness of threads or small shifts of threads within the fabric weave, aesthetic defects are difficult to recognize through formal description. Any adequate rule to intercept anomalies would end up picking up other details that in the end are not even anomalies.

The answer to these needs is machine vision based on deep learning, the latest expression of artificial intelligence. Our human-like aesthetic control solutions are characterized by: objectivity, repeatability, no rules (the system learns from experience), with the ability to adapt to changing situations and the integration of any workflow.

What happens when technology advances, entering the work spaces occupied by people?

It creates a human-machine marriage that will be able to deliver new levels of performance. The two sides complement and enhance each other, the weaknesses of each being the strengths of the other.

When quality is not translatable into formal language, technology that is based on parameters or rules cannot be used to do work that is normally entrusted to people. So artificial intelligence that learns from experience and can understand much more than just numbers comes into play.

A machine vision system based on deep learning is able to acquire much from the capabilities of its “human colleagues”-our systems are designed to work alongside people to form a winning team.

Our machine vision system for quality control processes does the job of observing the product to be inspected, using its strengths of speed, objectivity and repeatability. If a vision system encounters a “difficult” situation, where the human detail that has been intercepted is suspended between two possible judgments, there is a need for the human colleague to provide a final verdict.

MACHINE VISION SYSTEMS FOR PRODUCTION CONTROL

Machine vision systems for production control are less intrusive solutions for integrating measurements and control on manufactured goods into systems. These devices, like the human vision system, perform all their control operations “remotely” and have no contact with the object being analyzed.

Machines must be able to adapt and performance and quality must not be sacrificed, so it is important to understand what is happening on the line. To see what is happening, there is nothing more useful than machine vision systems creating a direct and integrated connection, like every other part of automation. When the light is integrated and the optics are synchronized, you get a perfect image even at full speed. The image that has been captured, allows maximum information to be extracted, so it ensures the quality of each individual product without slowing down any process.

This type of vision is contributing to the implementation of the fourth industrial revolution, where machine vision systems for control over production are linked to the control of manufactured products, with use being extended to the point throughout the industrial process chain.

A machine vision system consists of optical, electronic, and mechanical components that help us acquire and process images in the visible light spectrum and even outside of it. On the industrial side, on the other hand, automated systems are used, which provide more control at the level of dimensional measurements, typically they are installed as we said on a production line.

This technology offers so many advantages and allows for the objectification of quality control with the ability to determine the optimal quality/scrap ratio, thus ultimately reducing production costs for a technological increase in the product.

We tend to want quality more and more, moving toward production that has zero defects, with quality control that is reliable for every single part, so machine vision systems can provide objective data on product quality in a repetitive and automatic way.

The applications of machine vision systems are the verification of the dimensions and contour of objects with some accuracies down to the micron in some cases; then the reading and decoding of bar codes and “data matrix codes” takes place, with the verification of the various components from which a part is composed and its correct assembly, the control of the workmanship and the quality of surfaces with the identification and verification of colors.

Digital processing is applied not only to issues related to production quality, but it is a valuable tool for acquiring and analyzing information of what has been produced up to that point, to automatically track production, even to backtrack the production process, understanding the characteristics of the objects that have been analyzed to properly set the production line.

The components that make up a vision system are linked to some important evolutionary trends, these will enable an increasing number of applications to be solved easily and reliably. Ultimately they facilitate quality control in production, which is an increasingly important factor for those who wish to maintain a high standard of product quality.