Modeling MRT for well-characterized thermal imagers

Abstract

Predicting an accurate Minimum Resolvable Temperature Difference (MRTD) for a thermal imaging system is often hindered by inaccurate measurements of system gain and display characteristics. Variations in these terms are often blamed for poor agreement between model predictions and measured MRTD. By averaging over repeated human measurements, and carefully recording all system parameters affecting image quality, it should be possible to make an accurate prediction of MRTD performance for any resolvable frequency. Utilizing the latest NVESD performance models with updates for noise, apparent target angle, and human vision, predicted MRT are compared with measured curves. We present results for one well characterized mid-wave thermal staring system. *This document is approved for public release; distribution is unlimited. BACKGROUNDThe Minimum Resolvable Temperature Difference (MRTD or MRT) [1],[2] is a laboratory test that had performance implications for 1st and 2nd generation FLIRs (scanned systems). The MRT is considered to be a visual acuity evaluation of sensor performance, MRT is measurable in the laboratory, and can be related to field performance. Embedded in test results are a characterization of the thermal sensitivity and resolution of the sensor and the ability of a human observer to use the sensor to discriminate objects.It was once extremely important to model and measure the MRT performance for a given sensor. The utility of this concept was that the detection, recognition and identification criteria for a thermal task were related to the spatial frequency of a just visible four bar pattern (comparable to Johnson's criteria of cycles on target) as predicted either by theory or measured in the laboratory. The MRT equation, in the 1975 NVL Static Performance Model and its successors FLIR90/92 and NVTHERM, predicts ∆T for the recognition of 4-bar patterns as a function of spatial frequency. The modeled MRT equation contains sensor, target, and observer characteristics. These include detector noise, sensor component Modulation Transfer Functions (MTFs), processing electronics, display characteristics, and observer eye/brain models.Due to the sampling limitations of undersampled imaging systems, along with a myriad of other sensor effects that were not encapsulated in the MRT equation (e.g., digital effects, sensor boost, LACE, LAP, super-resolution), a new model needed to be developed that was no longer tied to a MRT equation. [3], [4] The new models NVThermIP and NV-IPM can utilize laboratory measurements as inputs, but they rely only on objective measurements such as Modulation Transfer Function and 3D Noise to determine thermal performance predictions.Still, the laboratory MRT measurement has utility in determining possible system level degradations, not easily revealed through objective laboratory measurements. These could include platform vibration, non-optimized displays, eyepiece effects, data resampling, human viewing factors, and other degradations not included in the performance models. If all ...