| Popis: |
Many microbolometer arrays being used for high-accuracy applications, such as body temperature measurement, are observed to exhibit a pixel crosstalk-like artifact which is not reflected in the system specifications. The ISO/IEC standard for thermographic screening specifies a total system uncertainty incorporating a calibrated image sensor and a paired IR calibration target to linearize the scene to high accuracy. Despite this use of an IR calibration target and calibrated image sensor meeting the required specifications, this crosstalk-like artifact is not addressed and can cause readings of facial temperatures to vary by as much as 2C in some systems, due to the artifact’s dependence on the thermal structure of an image. This artifact has been described previously as the “size-of-source” artifact, and is well-known within metrology laboratories but most users are unaware of it. To our knowledge, this thermal structure-dependent artifact has not been characterized nor corrected for human body temperature applications. We discuss test methods for evaluating the artifact and its impact on any resulting application, calibration methods for determining correction parameters and an example implementation of such a correction in real-time thermographic imaging. Finally, we propose a modification to the total system uncertainty equation in the IEC performance standard to account for the effect of this artifact. |
