Chapter 8:
Laboratory Measurements of Sampled Infrared Imaging System Performance
Published: 2000
DOI: 10.1117/3.353257.ch8
An infrared imager (often referred to as a FLIR, from the term forward looking infrared) can be characterized in terms of sensitivity, resolution and human performance. Sensitivity, resolution and human performance have been classically described by the following measurable parameters: noise equivalent temperature difference (NETD), modulation transfer function (MTF) and minimum resolvable temperature difference (MRTD or MRT). These are laboratory measurable quantities that can be used to verify the performance of an infrared system. These laboratory measurements are used to evaluate expected design performance and to periodically test an imager during its life cycle. These quantities are predictable in sensor design through the use of analytical models. Both model estimates and laboratory measurements can be used in a target acquisition model to determine the field performance (probability of detection, recognition or identification) of the imager. Sensitivity, resolution, and human performance are influenced by sampling artifacts that must be characterized. First, sensitivity is no longer sufficiently described by a single-valued NETD. The 3-D noise methodology, inhomogeneity equivalent temperature difference (IETD), and correctability are noise figures that have been developed over the past decade to more adequately describe both temporal and fixed pattern noise associated with focal plane arrays. Undersampled imaging systems are not shift-invariant. The shift-invariance condition, in particular, is compromised by under-sampling so that sensor performance becomes a function of phase (relative position between the image and the detector array). Resolution depends on the target-to-imager phase, so the MTF measurement may reveal sampling artifacts that give large MTF variations with target to sensor position. Finally, the human-performance parameter is perhaps most affected by undersampling. MRT can be strongly dependent on phase, where dramatic differences in measured MRT are attributed to different phase shifts. MRT is normally measured at optimum phase, yet the authors don't feel that static MRT measured with either optimum or unknown phase relationships correlate well with field performance. In the same way, it is not clear how to write field acquisition calculations based on MRT values measured past the half-sample (Nyquist) rate. A dynamic MRT method has been developed and measured which uses motion to provide enhanced sampling.
Online access to SPIE eBooks is limited to subscribing institutions.

Back to Top