Minimum resolvable temperature difference (MRT): procedure improvements and dynamic MRT

Minimum resolvable temperature difference (MRT or MRTD) is the primary perf ormance test for tactical military infrared (IR) sensors. It is a lab measu rement that may be related to discrimination task performance in the field. Also, a theoretical model exists for deriving the MRT of a particular sens or, where the model is used in sensor design evaluation and trades. The mod el includes both the human and the sensor/display in the overall system per formance and can be considered a visual acuity test of a human looking thro ugh the IR imager. Both the model and test are validated for first and seco nd generation forward looking IR sensors. The test procedure has been incor rectly applied to undersampled staring array imagers that are becoming more common today. Additionally, it lacks stringent controllability in the proc edures, which hampers test repeatability. As a result, multiple measurement s are typically performed and the results averaged. This is usually without regard to statistical requirements for estimating values from distributed data. Improvements to the MRT testing procedure are investigated in this re search. Data is reported that supports the use of improved techniques. The first improvement is to standardize the sensor and display configuration pr ocedures. A process for standardization is presented and applied to a mid-w ave IR imaging sensor. The currently accepted procedure for the static MRT test is accompanied by errors for undersampled staring array imagers. Thus, the second improvement suggested is to utilize dynamic MRT (DMRT) testing for undersampled IR imaging systems. Two motivations for the study presente d here are to minimize the uncertainty associated with MRT testing and to i mprove the characterization of undersampled imagers. These two goals are ac hieved by the suggested improvements. The resulting DMRTs are compared with modeled MRTs and static MRTs. (C) 2002 Published by Elsevier Science B.V.