In the United States, more than three-quarters of pedestrian fatalities happen after dark.[1] Nighttime pedestrian collisions at intersections have an 83% higher chance of being fatal without street lighting.[2] The severity of pedestrian safeguarding has prompted the US Department of Transportation’s National Highway Traffic Safety Administration (NHTSA) to finalize a new Federal Motor Vehicle Safety Standard in May 2024, mandating higher detection and safety standards for the automotive sector in upcoming years. Optikos, as a leading optical systems and testing hub, has recognized this critical challenge. Beyond simply deploying new optical solutions to enhance vehicle and pedestrian safety, we’re anticipating the industry’s future needs by starting to look at the next generation of Advanced Driver Assistance Systems (ADAS) technology: comprehensive, high-performance thermal camera testing systems designed specifically for the automotive sector. And we are talking about it at this year’s edition of AutoSens Europe.
Ahead of his keynote “Why Ignoring Emissivity in Thermal Camera Testing Should Not Be a Grey Area” at 4:15 PM on October 8 in Room 6, we sat down with Kevin Sweeney, Senior Optical Engineer from Optikos, to discuss the growing importance of thermal imaging in automotive applications.
As autonomous systems become more sophisticated, the quality and reliability of sensor inputs becomes paramount for production. You can’t have an AI making split-second driving decisions based on inadequate thermal imaging data. This isn’t just about meeting regulatory requirements. It’s about ensuring these systems can truly keep people safe by having the right data at hand.”
Q: Kevin, your talk focuses on thermal camera testing for vehicles. What’s driving this sudden interest in thermal imaging for this specific sector?
Kevin Sweeney: There are really two key factors. First, there’s a regulatory push. The US passed legislation last year requiring cars to have the ability to sense pedestrians at nighttime. For example, we’re talking about detecting a stationary pedestrian from a distance while approaching at 35 mph, or spotting a parked car on the side of the road with someone crossing the street from behind it. These real-life scenarios are becoming high-level test requirements and we expect this legislation to have a global impact.
Second, as vehicles become increasingly autonomous, they can’t rely solely on ambient light or street lighting. Thermal imaging is becoming a solution by detecting the thermal emissions that people and objects naturally give off. It’s an old technology really, but it’s finding a powerful new application in ADAS.
Q: So, from what you’ve observed in the market, the automotive industry is racing to adopt this technology?
Kevin Sweeney: Exactly. Right now, the industry is focused on production capacity, as in making sure manufacturers can equip all new cars with working thermal cameras. This technology has never seen a use case requiring this volume before. But our talk at AutoSens Europe takes things a step further. Let’s say you can manufacture millions of thermal cameras. You still need a robust way to characterize their performance. That’s what I will be focusing on in the keynote.
Q: Why isn’t the current testing approach sufficient?
Kevin Sweeney: The fundamental issue with current thermal camera testing is that it only captures half the picture. Video footage from these thermal cameras feed directly into AI models for object detection, which then determines what happens with the vehicle. Those AI models are highly sensitive to the quality of input they receive. Traditionally, the industry measures MTF, the Modulation Transfer Function, which indicates how sharp the images are. But for thermal cameras, MTF alone is insufficient. To build a comprehensive thermal camera system that delivers reliable results, you need to measure additional critical metrics beyond just image sharpness.
For example, there’s also the sensitivity challenge, which is very important here. People are warmer than their environment, but not dramatically so. That means you need to detect very small temperature differences to identify objects in the first place, especially when accounting for varying environmental conditions that could intervene in your data.
Q: How do you solve this issue of measuring the right things?
Kevin Sweeney: We’re advocating for a more comprehensive metric, namely the Minimum Resolvable Temperature Difference, or MRTD. This metric combines both elements—the sensitivity problem of detecting the smallest temperature difference visible, and the MTF that describes image sharpness—into one fuller-picture measure of image quality. The defense sector has used this approach for years, but with much larger thermal cameras. Now we need to miniaturize this testing equipment for small automotive longwave cameras, which can be done with some of Optikos’ existing technology like the Meridian line and Starfield systems.
Q: What are the three key takeaways you hope attendees walk away with?
Kevin Sweeney: First, that MTF testing alone is not sufficient for thermal cameras. A camera test station that only measures MTF simply isn’t adequate when you’re dealing with low temperature differences.
Second, a truly comprehensive metric requires specific test system calibration, particularly measuring the emissivity of the source. That’s the “grey area” we’re addressing in our talk title.
Third, in the context of machine learning algorithms being trained on these images, having objective, comprehensive metrics to characterize the inputs going into those models will be critical. We need to understand what level of camera performance is actually needed to support those machine vision models effectively.
Q: This sounds like it could significantly impact autonomous vehicle development.
Kevin Sweeney: Absolutely. As autonomous systems become more sophisticated, the quality and reliability of sensor inputs becomes paramount for production. You can’t have an AI making split-second driving decisions based on inadequate thermal imaging data. This isn’t just about meeting regulatory requirements. It’s about ensuring these systems can truly keep people safe by having the right data at hand. I’m looking forward to having these discussions at AutoSens Europe.
Kevin Sweeney’s talk “Why Ignoring Emissivity in Thermal Camera Testing Should Not Be a Grey Area” takes place on October 8 at 4:15 PM in Room 6 at AutoSens Europe.
[1] https://www.ghsa.org/resource-hub/pedestrian-traffic-fatalities-state-2024-preliminary-data-january-december
[2] https://www.sciencedirect.com/science/article/abs/pii/S002243752100092X
About Kevin Sweeney
Kevin began his career at NMG Aerospace in 2012 as a Process Engineer, a role he held until 2014. In 2015, he joined Optikos Corporation as an Optical Engineer and later transitioned to the position of Senior Optical Engineer in February 2021. Kevin was promoted to Principal Optical Systems Engineer in 2023.
Kevin Sweeney’s educational history includes a Bachelor of Science degree in Astrophysics from Ohio University, which he completed between 2007 and 2011. Kevin then pursued a Master of Science degree in Optics/Optical Sciences at the University of Rochester between 2014 and 2015.
