Application of 3D printing technology for the construction of a prototype 2D camera equipped with a thermal image drift compensation model

How to Mitigate Temperature-Induced Image Drift in Cameras

Conventional 2D cameras are widely used today across numerous applications: microscopy, industrial automation, manufacturing, medicine, machine vision, and 3D scanning. However, image analysis frequently overlooks the fact that a camera exposed to temperature variations undergoes structural deformation, directly affecting the recorded image. Put simply, the image itself distorts as temperatures change - a phenomenon known as thermal image drift. Marcin Adamczyk, PhD Eng., from the Institute of Micromechanics and Photonics at Warsaw University of Technology, has devised a simple and relatively low-cost solution to this issue. He will validate his approach under the Proof of Concept action funded by the European Funds for a Modern Economy 2021–2027 (FENG) program and implemented by the Foundation for Polish Science (FNP).

"Thermal image drift is observed in commonly used cameras because their optomechanical design is not optimized to minimize this effect. In virtually every measurement application relying on a camera, image deformation caused by temperature fluctuations significantly degrades system performance. Temperature changes within the camera can shift the recorded image position by several pixels. For many measurement applications using vision sensors, image shifts or deformations of even a single pixel can result in substantial errors. In sectors like automotive, these errors can be severe. A camera in an autonomous vehicle governs spatial orientation relative to other cars, meaning a discrepancy of a few inches could lead to hazardous situations. Remarkably, while thermal image drift is widespread, it remains poorly understood and is frequently ignored in measurement analysis," explains Marcin Adamczyk, PhD Eng.

His research demonstrates that thermal image drift is inherently stochastic; its magnitude and behavior are non-repeatable across consecutive heating and cooling cycles. This lack of repeatability explains why current software-based drift compensation methods fail to correctly mitigate the phenomenon.

"To resolve this issue, one must first eliminate the random nature of thermal image drift. Studies clearly indicate that intervening in the camera's optomechanical design is essential, as the root cause of non-repeatability lies in how the image sensor is mounted. It is necessary to develop a structure that allows the camera sensor to deform freely during temperature changes. In my research, I solved this by designing a compliant mechanism (flexure suspension) for the sensor using wire EDM technology. This eliminated the stochastic nature of thermal image drift and enabled the application of a mathematical compensation model. As a result, drift was reduced by over 80%, substantially improving measurement image quality," says Marcin Adamczyk, PhD Eng.

However, wire EDM technology is costly, limited in accessibility, and labor-intensive. Therefore, the researcher currently aims to verify whether metal powder 3D printing can be used to manufacture the precision structure that allows the camera sensor to deform freely with temperature changes. This concept will be evaluated throughout the Proof of Concept project.

"To test this research hypothesis, I will use metal powder 3D printing to produce a flexure suspension design, build a prototype camera, and analyze the behavior of thermal image drift in the assembled unit. I will conduct a series of measurements, calculate drift parameters, and evaluate the degree to which its behavior becomes repeatable," explains Dr. Adamczyk.

The technology has been filed for patent protection in Poland, with international PCT extensions underway for the USA, Canada, India, Germany, Japan, and the Netherlands. Potential applications are broad, spanning any domain using machine vision for metrological purposes. Key implementation fields highlighted by the researcher include automotive systems (e.g., autonomous vehicles using cameras to interact with their environment), 2D camera metrology (e.g., microscopy as well as 2D, 3D, and 4D scanning), and industrial machine vision (e.g., collaborative robots - cobots).

The advantage of this solution lies not only in significantly improving image quality through drift reduction, but also in expanding the operational temperature range for measurement cameras. "The developed method is passive, requires no external power, leaves camera dimensions and appearance unchanged, and is relatively easy to automate for large-scale production. It is also cost-effective and can be integrated straightforwardly into existing commercial cameras," concludes Marcin Adamczyk, PhD Eng.

Marcin Adamczyk, PhD Eng., is a mechanical design engineer with extensive project and leadership experience. A graduate of the Faculty of Mechatronics at Warsaw University of Technology with a specialization in Precision Engineering, he defended his PhD with honors at the Institute of Micromechanics and Photonics in 2019. He gained expertise by participating in over 20 scientific, R&D, and industrial implementation projects, collaborating with companies such as Barlinek, Mitsubishi Electric, KSM Vision, PhiBox, SmartTracking, Mnemosis, OVE, CLKP, Sygnis, and MACRO-SYSTEM. He is also a graduate of the Executive Leadership Principles course at MIT, lead author of 12 JCR-indexed publications, and a reviewer for journals including Applied Optics, Optics Express, Sustainability MDPI, Sensors MDPI, and Processes MDPI.