Development of a calibration for geometry-independent temperature measurement with thermal imaging cameras
Brief description
Infrared (IR) thermography is a non-invasive, non-contact measurement method that can be used to determine the temperature distribution of a body. For this purpose, IR cameras are used to capture infrared radiation emitted by the imaged objects. The measurement signal is then converted pixel by pixel by means of a camera-specific calibration function into a radiation intensity and, taking into account the emissivity, into a corresponding temperature. As in visual imaging, errors occur in the optical imaging of objects, which depend in particular on the object geometry (e.g. size). In the case of thermal imaging cameras, the contribution of these imaging errors to measurement uncertainty has not yet been investigated in detail in an academic context, nor is it known that any manufacturer considers this phenomenon in its calibration procedures.
The correction of these imaging errors by improving optical components has been practically exhausted. The idea behind the project is to compensate for the aberrations by means of smart, digital image processing. However, the systematic aberrations can neither be determined exactly, nor is it mathematically possible to invert them directly for compensation. Therefore, the solution approach of the compensation is based on a physically based meta-model structure, which approximates the inverse via decomposition into a chain of partial inversions and these partial models are then parameterized in a data-driven manner. Compensation is very challenging because the meta-model structure must represent the superposition of the different optical effects and these should be separated for calibration data collection and compensation. Furthermore, due to the large variety of combinations of optical components in a high-end camera and due to manufacturing tolerances, a camera-specific calibration of the compensation is most likely unavoidable. This requires a sophisticated experiment design and an automation of the calibration process to achieve the required reproducibility and a short measurement time. The implementation involves significant technical risks with the assumed separability of the effects, the unclear transferability to arbitrary, real temperature distributions and the controllability of the variety of optical components.
Person in charge
Funding
Project duration
October 2022 - February 2025