From the world of artistic photography to the precision-driven realm of imaging system engineering, Uwe Artmann has built a career at the intersection of creativity and technology. Now, Uwe shares his expertise as part of our AutoSens Advisory board.
In this exclusive AutoSens interview, Artmann shares his perspective on the industry’s biggest challenges and emerging trends.
Your career has spanned both artistic photography and advanced engineering. How has this unique background shaped your approach to testing and developing imaging systems?
I always like to say that I cannot design a lens or a sensor, and I do not program signal processors, but what I can do is help the lens designers, sensor architects, and programmers see the entire system. That is important as cameras, particularly camera systems combined with machine vision tasks, are very complex. I often see experts in machine vision trying to fix a lens issue on the software side when it’s most likely a quick fix in the lens itself. When designing test solutions, it is crucial to understand the complete imaging chain. This understanding is what my career has so far allowed me to achieve.
You’ve highlighted a special interest in noise reduction and MTF measurement. How do you see these factors influencing the development of high-performance imaging systems for ADAS and autonomous vehicles?
When I started working with digital imaging systems in 2003, you could describe the performance with a few key performance indicators, for example, noise and MTF. With continued advancements in signal processors and their capabilities to reduce noise in the image, achieving higher performance for noise and MTF became much more manageable. However, the images looked even worse because of the resulting texture loss (the loss of low-contrast fine details) due to the noise reduction. Over the years, I’ve developed and adopted different test methods to reliably describe a camera system’s performance without blind spots, which results in good measurement results but still shows poor performance in the final image.
While this is vital for all human vision applications, machine vision requirements differ. The interesting part, however, is that there are still no precise models relating to a machine vision observer. So, the decision on which metric is essential and how well the signal processor tuning performs has still not been settled for this use case.
What excites you most about the advancements in imaging devices and the impact they’re having on the automotive industry? What emerging trends do you believe will have the most significant impact on the automotive industry in the coming years?
I have seen a pixel race in different products in my professional career. For photography cameras and mobile phones, there is a claim that more and mainly smaller pixels will improve performance and benefit users. After a while, however, the camera engineering community recognized that the highest technical possible maximum was not the most beneficial for the end user, and instead, a sweet spot of best performance is more desirable. In the automotive sector, we also see a clear trend to smaller pixels, but we are still searching for the exact size and number of pixels per frame that provides the best performance regarding our use cases in this industry.
Another growing trend is validation through simulation and lab setups. It has long been accepted that systems tested on the road are validated after a certain number of kilometers. While there is a place for road testing, we also need to enhance testing using structured system checks in a test lab against all possible lighting scenarios and, of course, with digital twins and simulations. These lab tests will help ensure that no technical flaws will put people at risk in the real world.
What are the biggest challenges and opportunities you foresee for the automotive imaging sector in the next five years, and how do you think the AutoSens platform can contribute to addressing them?
Cameras in a car are no longer a convenience but a critical component of a safety-relevant system. So, the performance and safety of the complete system, for which cameras are only one part, has to be tested and validated. We need to be able to check the entire system, not only on the road but also in the lab and within simulations.
We cannot only test the performance of the sensor and lens; other factors also need to be considered when determining their impact on the overall performance. These factors include windshields or other covers, mounting, data flow, compression, signal processors, the machine vision stack, and any other components that contribute to performance and safety. Achieving this will require that different experts in their respective fields talk to each other and exchange their ideas and experiences. This understanding is what I see as the most significant benefit of a conference like AutoSens – bringing people together to connect experts so that they can work together to solve their problems.
As an integral part of the AutoSens community, what do you think makes AutoSens unique compared to other conferences, and what do you hope to contribute as a member of the Advisory Board?
Typically, there are two types of conferences: Academia-driven conferences that purely focus on research but lack industry insight and real-world application, and industry-driven conferences where companies pay a lot of money to present their latest products on a stage, which can lack integrity. AutoSens brings the best of both worlds together. With a combination of invited keynote speakers – key industry stakeholders and renowned academica – as well as peer-reviwed abstract submissons, it attracts a lot of technical and academic professionals; but with the exhibition floor showing real-wolrd demos you can still gain great insight into the latest products and ideas from the industry side.
