Image Processing Industrial Image Processing—Easy and Profitable to Use

Related Vendors

MVTec Software GmbH

FAULHABER Drive Systems

Wuxi InfiMotion Technology Co., Ltd.

Automation and digitization are making their way into more and more industries. Restructured supply chains, skilled labor shortages, and shorter product lifecycles are driving this development. In this context, industrial image processing is becoming a key technology, as it enables the automation and optimization of processes along the entire value chain.

Industrial image processing (machine vision) has become indispensable over the past decades. As the "eye of production," it constantly monitors manufacturing and logistics. But how does machine vision work in practice? What components are needed, what advantages does the technology offer, and where can it best demonstrate these benefits?

First, the basics: A machine vision application requires hardware and software components. Hardware includes industrial PCs, appropriate lighting, and, of course, image acquisition devices such as cameras or sensors. These capture production processes and generate image data. Comparatively, they function like the lens of an eye.

Gallery

Machine Vision—The Eye of Production

As in the brain, the captured information must be processed to achieve the desired added value. This is handled by integrated machine vision software. It analyzes image data and provides the results. A range of methods and technologies is available for this purpose, including artificial intelligence.

Essentially, machine vision recognizes objects and contexts in the production environment. This enables automated identification and assignment of objects in the supply chain. It works based on external features such as color, shape, or texture, or via printed codes.

Moreover, industrial image processing is a key technology of Industry 4.0. For example, in quality assurance: defects in objects are detected more reliably and accurately around the clock than the human eye ever could. In the context of connected production, the defective products can be automatically classified as rejects and sorted out in time.

Machine Vision And Robots—A Dream Team

The technology is also indispensable in robotics. Traditional and collaborative robots use it to precisely grasp, process, and place objects. This enables them to independently perform additional tasks, and processes can be fully automated.

One example is logistics: during the loading and unloading of pallets, an industrial camera captures the objects, the image processing software determines their position, and sends it to the robot controller. This allows the robot to recognize, precisely grasp, and place objects. In other words, it can "see."

But even without robots, machine vision is valuable. In intralogistics, objects are usually equipped with barcodes. These are reliably detected and tracked—from goods receipt through quality and completeness checks to the detection of free shelf spaces. Picking steps and removal processes can also be automated in this way.

Where Can Image Processing Be Used? The Question is Rather: Where Not?

Whether in the automotive industry, mechanical engineering, or food production—machine vision supports every industrial sector. The electronics industry, such as semiconductor manufacturing, particularly benefits from it. This is because the complexity is very high due to the many production steps and the small size of the components. Naturally, machine vision plays a key role in tackling the demanding inspection and testing tasks of quality control.

Time is Money—Machine Vision in High-Speed Applications

Machine inspection using industrial image processing is significantly faster than the human eye, with results that are objective and reproducible, and the quality of the inspection does not deteriorate, for example, due to fatigue or the monotony of the task. Therefore, machine vision is ideal for high-speed applications. For instance, the highly accurate localization of any object takes less than 10 milliseconds, while a blink lasts 300 to 400 milliseconds.

About the author

Ulf Schulmeyer is the Product Manager for Merlic at MVTec Software GmbH, where he has been working since 2023. His role is to further refine the profile of the no-code software Merlic, which is primarily aimed at companies with limited resources for software programming, making advanced image processing technology more accessible. Previously, he worked as a product manager and sales manager for Data Becker GmbH and Franzis GmbH, where he was primarily responsible for the development and management of numerous and diverse software products.

Rule-Based or With AI Methods? Both!

On the technical side, it also applies: Rule-based methods such as matching, code reading, blob analysis, or measurement can be combined with AI-based approaches like deep learning. This allows many applications to run even more robustly and accurately.

Since deep learning independently "learns" through the evaluation and analysis of digital image data, it also enables applications that would not be achievable with rule-based methods alone. This is the case, for example, when data exhibits greater variations due to natural changes—such as with fruits and vegetables or during scratch inspections. In such instances, rule-based algorithms cannot be clearly defined: when is a surface acceptable, and when is it no longer?

Despite these new optimization possibilities, deep learning also has its limitations. For certain applications, the technology is simply too complex, i.e., too resource- and cost-intensive. Therefore, it is advisable to always combine both worlds of image processing—deep learning and rule-based methods—for maximum flexibility.

Combine Ease of Use With Powerful Technologies

For image processing beginners, it is also crucial that powerful machine vision applications can be created and operated with ease. The technologies must be reliable for a wide range of applications—while still being easy to use. These requirements are met, for example, by the intuitive no-code software Merlic from MVTec. Without programming knowledge, even non-experts can use it to create and seamlessly integrate all common machine vision applications.

Smart-Factory

Access solutions for the Smart Factory - Industry 4.0