Professor Eberhardt, your Steinbeis Transfer Center for Vision Systems has been part of the Steinbeis Network since 2014. What factors influenced your founding of this enterprise?
There are a number of unsolved problems facing the world of industrial image processing, despite major advances in the development of sensors, improvements in the power of modern IT architectures, and the trend toward simplified user interfaces. But it’s the sheer number of different technologies and trends that is overwhelming companies in their quest for a simple, pioneering, and – most importantly – functional automation solution. I’m often approached by classic small and medium- sized enterprises (SMEs) asking for help with their automation problems. The scope of questions I’m asked is extremely wide. The requests range from conventional consulting to the development of prototypes. The Steinbeis Transfer Center for Vision Systems gives me a framework within which to carry out these projects. It allows me to focus on the key issues and removes the burden of administration and routine tasks. I have major hopes for the future in terms of new ideas arising from collaboration within the Steinbeis Network and the possibility to take on more people.
The know-how at your Steinbeis Enterprises revolves around R&D and consulting in fields relating to industrial image processing. What areas of industry do your customers work in and what sort of problems do they approach you with?
The problems they approach me with at the transfer center can be completely different, and they come from really broad areas of industry. The latest client requests range from woodworking to pharmaceuticals. Of course there are also clients that I got to know when I was working in industry, with whom I have a close understanding. Lots of the questions I’m asked are about investment plans in the field of industrial image processing. An increasing number of queries come from SMEs in the region, which value the opportunity to work with a professional partner in the area. For them, too, it’s mainly about professional advice related to automation projects, although there are also feasibility studies and prototype developments.
You’re also involved in the field of machine vision, where irrelevant information is filtered out and only meaningful data is handed on for analysis. In times of too much information, that sounds extremely appealing. What are the challenges with this technology and what is the current status of developments?
Machine vision has made tremendous progress in recent years. On the one hand, the area is benefitting directly from the development of faster and faster computer architectures, while, at the same time, energy consumption is going down. Algorithms that were undergoing scientific testing years ago have now finally made their way into industrial projects under suitable conditions. Then, on the other hand, there has been major progress in camera technology. All industrial cameras are now digital with megapixel resolutions and quick image refresh rates. Also, advances made in sensor production make it possible to achieve ultimate sensitivity and cover a broad spectrum. Because cameras are getting faster and faster at producing bigger and bigger images, it’s now all the more important to be able to process the huge data volumes. This makes it necessary to distribute image data in parallel across several systems. This is where we’re reaching the existing borders of possibility. To process camera data simultaneously, it often takes an entire cluster of computers. The pixels have to be consolidated, filtered, segmented, transformed, and the put into categories using neural networks, support vector machines, or similar kinds of algorithms. The costs of such image processing systems are correspondingly high, often a 6-digit number.
In less complex areas, increasing use is being made of “intelligent” cameras made by traditional sensor manufacturers. These cameras already come with an integrated evaluation algorithm and they’re much easier to operate. Users don’t have to think about all the pixels, they just interact with the system and define what needs checking on the component.
The number of areas industrial image processing has entered into has risen sharply in recent years. Which trends do you think will dictate the future for us?
End customers have rising expectations in terms of quality and it’s already the norm in lots of areas to have 100% checking and traceability. Visual checks and quality controls are already key technologies and without them, the majority of the automation tasks that are carried out would be impossible – not just in classic quality assurance but also when it comes to robot vision systems.
On top of that, image processing is increasingly making its way into tasks related to optimization and manufacturing controls. This technology is profiting from the trend toward more flexibility in production, right down to batch sizes of one. The concepts underlying this general automation trend are often mentioned within the context of Industry 4.0 – something that wouldn’t even be possible in many cases since flexible production cells are dependent on the output produced by highly sophisticated sensors.
One major trend is the increasing use of 3D technology which not long ago was considered complicated and expensive. Now people are using a whole variety of 3D technologies, some of which are extremely economical, in areas ranging from stereo image processing to laser scanning.
Just as it’s always been, the automotive industry is still the biggest area using this technology, and it has played an instrumental role in industrial image processing becoming an irreplaceable part of automation technology in recent years. But more and more attention is now also being given to non-industrial applications, especially the fields of medicine, transportation, security, sports, and farming. The driverless cars of the future will need a variety of 2D and 3D sensors and cameras to take in the traffic environment. In sports, we need ultra-precise imaging devices to do things like spot the ball crossing the line. But we’re also increasingly seeing imaging technology used in farming to drive autonomous harvesters but also to monitor the optimum time to harvest crops.
Professor Dr. Jorg Eberhardt is director of the Steinbeis Transfer Center for Vision Systems at the Ravensburg-Weingarten University of Applied Sciences. The services offered at center range from the development of optical camera systems (2D, 3D, color measurement) to consulting in the fields of optical measurement technology, lighting development, 2D and 3D technology, applied research in the field of optical measurement technology, and 3D camera technology, as well as seminars and training.
Professor Dr. Jörg Eberhardt
Steinbeis Transfer Center Vision Systems (Meckenbeuren)
SU1840@stw.de