The power of AI in vision solutions

Artificial (artificial) intelligence - or AI for short - is increasingly being used for a variety of purposes. As a Vision System Integrator, EKB is therefore increasingly being asked to integrate AI into its (already) advanced vision solutions. For example, in the context of inspection issues and for quality purposes. With all the knowledge and experience the company has accumulated over the years, the 20 (!) specialists are now able to assess for each issue whether AI has added value and how best to translate this into a technical solution.

Whereas AI used to be an elusive phenomenon, now no one is surprised by applications such as ChatGPT or the way smartphones recognize images. AI technology has also made its appearance within the world of vision. Michiel de Lange has been a director at EKB for 17 years and is also responsible for the vision department. In recent years, he has seen and experienced what AI can do for vision. "It is pre-eminently a technology that can 'learn' quickly based on a limited number of images," he says. "Whereas traditional vision systems used to require thousands of images to distinguish 'right' from 'wrong,' AI can often get by with a few dozen images. So you get to a working and reliable application faster. In addition, with AI you can more easily adapt the application when requirements or situations change."

AI is only a particle

Yet an expert like EKB, in particular, warns that implementing AI in a vision system is not a standard action. On the contrary. Sébastien Negrijn, manager of engineering vision knows: "AI is ultimately only a small part of the solution. Its success is determined by a multitude of components and factors including the input (image) material, sensors, communication and algorithms. Because we at EKB control all these aspects in detail, we are able to create a meaningful, efficient and sustainable solution. With or without AI."

Negrijn is referring to the fact that AI does not 'just add value' everywhere. A project in which the combination of vision and AI is envisaged therefore always starts with accurately determining the customer's needs. This is followed by a quick scan to determine whether AI indeed 'fits', after which a feasibility study and concept design validate the solution. Negrijn: "This is the only way to arrive at the system with the desired speed and accuracy that a 100% control could possibly contribute to an output of high quality and value."

Applications

Meanwhile, EKB has a multitude of applications on its reference list. Project manager and experience expert Robbert Niemans cites as an example a company that packs cotton pads in cylindrical packaging. "The material of these discs is not always homogeneous, so there is a chance that some discs do not end up neatly in the packaging. This does not matter for the functionality of the product, but it simply does not look pretty and the manufacturer only wants cosmetically perfect packaging. Our AI solution - which required a lot of research and testing - can now very reliably assess whether a package meets the quality requirements. Also in agri with its diversity of products that never look the same, the combination vision/AI is widely used."

Future

EKB has been operating as a Vision System Integrator since 2018. In that time, the associated department has grown from three pioneers to more than 20 people working on vision and AI on a daily basis, often as part of a larger project. "Requirements are getting higher and claims are increasingly lurking," De Lange knows. "However, the technology is getting cheaper and cheaper, and the knowledge and experience to apply it optimally is growing. The flight that AI is currently taking is therefore not over yet. AI will be able to support the market better and better."