Computer Vision for Quality Control

Computer vision is an AI-related technology that finds extensive applications in the manufacturing industry, especially in the industry 4.0 environment. Today, computer vision for quality is a crucial element of quality control, and more and more companies use this technology. In this article, we are going to take a closer look at quality control and see how you can use computer vision for quality.

When it comes to quality control, manufacturing companies use diverse methods and technologies. The most popular ones comprise:

• Computer Vision services
• Machine learning
• IoT
• Digital twins

Today, though, we want to focus strictly on computer vision for quality. We are going to examine how this technology works, where it can be useful, and what companies use it in their everyday work.

Quality control with computer vision

Generally speaking, computer vision is a subfield of AI that works primarily based on a set of cameras (hence the name) that continually monitor the environment or surroundings. In the manufacturing world, these cameras monitor the production line. Moreover, there has been created a new subfield of computer vision, strictly with the manufacturing sector in mind. In manufacturing, we talk about machine vision. It all works pretty much in the same way, but every single element is tailored to the needs of manufacturing companies.

Therefore, machine vision algorithms are designed especially to detect deviations, glitches, and defects in newly made products. Machine vision systems are capable of processing a lot of visual information, looking for every possible defect at the same time.

There are many benefits related to these solutions. Computer vision systems can work 24/7, all year round. They are never tired, never on vacation or sick leave. Therefore, you can achieve improved thoroughness without the need to hire more employees. When you think about it, computer vision (combined with machine learning) can be simply cheaper than a human quality control team. The benefits don’t end here. Computer vision for quality can operate at high speeds, meaning there are no delays on your production line due to quality control–intelligent algorithms can conduct many inspections simultaneously.

Now, it is important to say that computer vision systems are frequently backed up by machine learning algorithms. This way, these technologies create synergy, which makes both of them more effective.

Quality control: humans still needed

At this point, you might think that it all means that you can dismiss your entire quality control team and just stick with AI. Although this idea might be tempting, we don’t recommend firing all your QA staff. It would be best if you kept at least a small team that can oversee the work of computer vision applications, manage their work, and implement changes when necessary. In some instances, AI-based technologies need additional support, even though they are designed to work without human assistance.

Image analysis

It’s yet another indispensable part of computer vision for quality. In general, computer vision apps can simply “look” at other things. You need image analysis to spot potential irregularities and alert your quality control team to take necessary corrective actions. Image analysis algorithms analyze every picture they get from CV and analyze it in order to pinpoint defects in products or the production process itself.

Automated visual confirmation

IBM, the company behind the Watson supercomputer, has gone even further and provided manufacturing companies with means to streamline visual confirmation. When it comes to VC, it’s all about making sure everything is as agreed with the client. In other words, to ensure the final product is error-free.

How can Watson be useful here? IBM’s product compares high definition images from the production line and uses a library of images displaying known defects in final products in order to detect faults almost in real-time. Watson can quickly analyze your parts and other components, the assembly process, and, obviously, final products. With Watson, even visual confirmation can be automated and accelerated, at least to some extent. It goes without saying that IBM’s computer also uses machine learning.

With Watson, all you need to do is to create a comprehensive dataset of known product defects from the past. You need a library of pictures showing broken or defective elements and parts and feed Watson with this information. As a result, it will analyze images captured from the manufacturing floor and compare them with its database.

Companies using computer vision for quality

As we mentioned earlier in the text, more and more companies use computer vision for quality control purposes. Let’s take a look at some examples:

Oal connected: APRIL™ Eye

APRIL Eye combines machine learning and computer vision to streamline the traditional date code verification process. This solution is extremely useful in the FMCG sector, where every product has a specific expiration date. OAL’s system takes a photo of each date code and then reads each one back using scanners to ensure they match the programmed date code for the given batch.

And what happens if the code seems to be incorrect? The production line comes to a complete stop, ensuring that no incorrect labels are released into the supply chain. As a result, the verification process is fully automated and enables FMCG producers to achieve full traceability and significant time savings. Take a look at how APRIL Eye works in practice:

Ivisys shapeinspector

And here’s another interesting example. Ivisys is a company building 2D and 3D inspection solutions. Their ShapeInspector moves the product that needs to be verified into the system and then runs the inspection sequence with optimized light and camera position. The results of these inspections can be quickly verified and stored in the system’s internal database.

When it’s all done, the product can be replaced with the next one. All you have to do is select a product recipe for the inspection. The product can be loaded manually or by the robot.

Take a look at ShapeInspector in action:

Keyence CV-X

Keyence is a company based in Belgium. They offer a wide range of sensors, safety, and measurement solutions. One of their main products is CV-X–a machine vision system that enables automated inspections in all industries, including automotive, semiconductor, medical, food, and packaging.

Their computer vision systems can be used to inspect the presence/absence of specific elements, but also flaws, colors, character recognition, and more. Keyence’s system is based on a combination of industrial CCD cameras, high-intensity ring light, and advanced processing tools.

As you can see, at least in the manufacturing world, computer vision for quality is an effective and profitable solution. Because everything is automated, and computer vision systems can conduct many different inspections in real time, you save a lot of time and money. Additionally, you don’t have to maintain the whole quality control team. The benefits are apparent!

If you’d like to find out more about computer vision in quality control, or you are simply interested in how technologies such as computer vision and machine learning can be used in your company – feel free to drop us a line. Addepto is an experienced AI consulting company based in the United States and Europe. We will gladly help you find and implement the best solution for your business.