Anything is possible – with machine vision this is actually (almost always) true

Innovative machine vision applications are only developed in large companies? Not at all! Every day, the ingenious engineers at MSTVision show how companies can make profitable use of state-of-the-art technologies – even with a relatively small workforce, but lots of passion and creativity.

Michael Stelzl General Manager MSTVision

Small but mighty. Although perhaps not as hip as it used to be, this saying accurately reflects what has characterized MSTVision since its founding in 2016. As a small company, we are committed to using powerful tools and a wealth of ideas to develop efficient solutions for the broadest range of applications by customers, whether large or small, national or international. Machine vision provides lots of different tools, hardware, and software to this end. The challenge for us as a small company lies in finding the right tool and putting it to use in a way that ultimately achieves the best solution for customers. Creativity and passion are therefore very important to us. Our strength lies in our ability to diverge from the beaten track and always work with the latest technology. CPU solutions and programmable chips (FPGA) that enable us to push to the limits with our work are key factors.

As part of most projects, we perform feasibility studies for our customers. These then enable us to optomechanically implement even complex optical structures as integrated solutions. A key part of our service is the transfer of knowledge to our customers. For example, we regularly offer training on MVTec HALCON, line scan cameras, optics design, VDI guideline 2632 and more. The examples below illustrate the kinds of projects that we implement in this way.

Smart machine vision supports the development of advanced gene therapies

DiNAQOR is a life science company that develops advanced gene therapies and enables in vivo gene editing solutions. One application example is the development of individual gene therapies to give patients with inherited diseases (e.g., cardiomyopathies) a chance to be cured.

To test the effect of such gene therapies, DiNAQOR relies on its (patient-specific) artificial heart tissues. MSTVision supported DiNAQOR in developing a new technology platform for evaluating the contraction of three-dimensional engineered heart tissues (EHT) generated from human induced pluripotent stem cells (hiPSC). Thanks to MSTVision‘s improved approach, it is now possible to acquire 24 EHTs simultaneously and fully automatically without sacrificing accuracy. The new approach has eliminated the need for user intervention during measurement and the movement of the camera by an axis system. After the evaluation of the optical setup by the project team and the joint setup of the system, the HALCON scripts developed by MSTVision were integrated into a fully cloudbased data management system.

MultiChannel capture of a metallic surface including calculated photometric stereo image

Advancing e-mobility – also thanks to machine vision

Many sectors and companies are working to advance e-mobility as there are multiple aspects that need to be tweaked. One of these is the production of batteries for e-vehicles. In particular, the coating processes within the scope of electrode production need to be closely monitored as these steps are key to the reliability of the batteries produced as end products. The extremely high resolutions required for this result in the increasing use of line scan cameras. MSTVision can support companies with many years of experience in this area.

Our established MultiChannel technology makes it possible to use one or more line scan cameras to capture various surface properties in a single relative movement, e.g., through the combination of light and dark fields, different wavelengths, or different lighting angles. The images captured are prepared in the FPGA without any additional CPU load and can then be simply and efficiently processed further with MVTec HALCON. Information from different image channels can be compared and contrasted without calibration. This technology makes it possible to perform the photometric stereo procedure, also known as Shape from Shading, with high-speed line scan cameras. This in turn enables the separate evaluation of the brightness (albedo) and the topography (inclination or curvature) of each pixel. Continuous surface inspection is a highly complex technical process. The underlying algorithm was therefore implemented in the FPGA of the framegrabber, an electronic circuit for digitizing analog image signals. As the procedure is suitable for both very light and very dark surfaces, it is often used to inspect lithium-ion battery foils, bipolar plates for fuel cells, and circuit boards for power electronics.

Application-specific setup for the inspection of high-performance electronics

How high-speed machine vision helps reduce plastic waste

Another topical issue right now is the sorting of bulk materials. A few years ago, one of our customers commissioned us to modernize an outdated sorting system designed to detect black specks in white and transparent plastic granules. The difficulty lay in achieving the necessary performance data with regard to the optical resolution, high bandwidth, and low latency. To that end, MSTVision developed a proprietary technology platform, comprising line scan cameras, FPGA code, and its own electronic boards for precisely controlling the timing of quick-release valves. The performance data achieved are impressive: The FPGA-based solution can process 2.4 GBytes and more than a million objects per second, while also enabling reaction times of just one millisecond. The combination of the improved sorting accuracy and the higher throughput makes it far easier to waste less plastic. There are multiple potential uses for this technology. For example, the new system can be used in the recycling or food sectors by being easily expanded to include a multispectral imaging device (VIS and SWIR). Even the photometric stereo procedure in free fall is possible. In the future, it will also be possible to integrate customer-developed neural networks (deep learning).