The Multispectral Thermal Imager (MTI) is designed to demonstrate the utility of multispectral remote sensing from a satellite platform for a variety of applications of interest to the U.S. Department of Energy. These applications include characterization of industrial facilities, environmental impacts of effluents, global change, hazardous waste sites, resource exploitation, crop health, and others. The MTI was designed using a procedure which we call `End-to-end modeling and analysis (EEM).' We began with target attributes, translated to observable signatures and then propagated the signatures through the atmosphere to the sensor location. We modeled the sensor attributes to yield a simulated data stream, which was then analyzed to retrieve information about the original target. The retrieved signature was then compared to the original to obtain a figure of merit: hence the term `end-to-end modeling and analysis.' We based the EEM in physics to ensure high fidelity and to permit scaling. As the actual design of the payload evolved, and as real hardware was tested, we updated the EEM to facilitate trade studies, and to judge, for example, whether components that deviated from specifications were acceptable. During detailed calibration at the Los Alamos Radiometric Calibration Facility we used our models to explain certain observations, and to extend limited measurements to larger domains of applicability. Data analysis programs have been developed to generate a comprehensive set of data products through our Data Processing and Analysis Center. The satellite was due for launch on 8 February 2000: the actual launch data was 12 March, 2000. At the conference we anticipate sharing some preliminary observations from on-orbit.

MEMS technology now makes possible to produce active microdevices combining detection, signal processing, and data storage with accuracy and compactness. In view of their characteristics, it can be expected that such microsensors will be used extensively in space applications dedicated to micro and nano satellites. For this purpose, a specific investigation dealing with the complete development of a micro-earth sensor used for attitude control of Low Earth Orbit satellites is under realization and test.

Clouds and the Earth's Radiant Energy System (CERES) instrument, with it's three scanning thermistor bolometers, was designed to provide accurate measurements for the long- term monitoring of Earth's radiation energy budget. The sensors measure broadband radiances in the shortwave (0.3 - 5.0 micrometers), total (0.3 - >100 micrometers) and 8 - 12 micrometer water vapor window regions. Two of the CERES instruments, Flight models 1 and 2 (FM1 and FM2) are scheduled for launch aboard the Earth Science Enterprise Terra Spacecraft in November 1999.

The Airborne Hyperspectral Imager (AHI) system is a long- wave infrared imaging spectrometer originally designed to detect the presence of buried land mines. Subsequent work with AHI has shown the utility of the long-wave infrared for other applications. The AHI system has been used successfully in the detection of buried land mines using infrared absorption features of disturbed soil. Gas detection was also shown to be feasible, with gas absorption being clearly visible in the thermal IR. This allowed the mapping of a gas release using a matched filter. Geological mapping using AHI can be performed using the thermal band absorption features of different minerals. A large-scale geological map was obtained over a dry lake area in California using a mosaic of AHI flightlines, including mineral spectra and relative abundance maps.

An improved acceptance test procedure for imaging infrared (IIR) equipment has been developed to remove subjectivity, improve repeatability, and decrease the time and expense over existing methods. Traditional procedures for acceptance testing of IIR equipment require manual minimum resolvable temperature difference (MRTD) measurements at multiple spatial frequencies with both positive and negative contrast targets. Although the manual measurement of MRTD is the standard evaluation technique used in developmental testing, it is deficient for use as an acceptance test. The main limitations are subjectivity, length of time required, and manpower (i.e., cost). The Missile Guidance Directorate of the U.S. Army Aviation and Missile Research, Development, and Engineering Center has developed the automated infrared sensor test facility to perform automated acceptance tests and remove the subjectivity of the test procedure.

The Automated Objective Minimum Resolvable Temperature Difference (AO-MRTD) is an approach to provide an automated and objective measure of a thermal imager's performance. The algorithm is intended to provide a more accurate, reliable and cost effective figure of merit than traditional subjective MRTD measurements. MRTD values are calculated using signal-to-noise ratios obtained by match filtering digitized images of low contrast four bar patterns. Match filters are constructed from a high contrast image of the four bar patterns and the algorithm can assess sensor performance beyond Nyquist (similar to the minimum temperature difference perceived figure of merit). The MRTD values derived by the algorithm do not represent the minimum temperature differences perceived by human observers; however, simple modifications to the threshold signal-to- noise ratio result in human-like MRTD values. The algorithm has the flexibility to construct multiple types of match filters derived from a single high contrast image. Currently, the algorithm derives separate MRTD values for match filters corresponding to the bar pattern and it's derivative. Preliminary results suggest that a combination of the two match filters may yield a match to the human subjective MRTD results. The basic construction and operation of the algorithm is outlined.

Three perception experiments were conducted to quantify the relationship between tactical military vehicle identification (ID) performance when using an imager and the Modulation Transfer Function and noise characteristics of that imager. The results of these experiments show that the limiting resolution metric provides a reasonably accurate prediction of target ID performance. For example, limiting resolution is a better predictor of performance than Modulation Transfer Function Area or Integrated Contrast Sensitivity. However, a metric consisting of integrating the square root of the product of Contrast Sensitivity and spatial frequency provides a better fit to data than limiting resolution. This paper describes the perception experiments and test results. The predictive capability of a selected group of image quality metrics is evaluated. This paper also discusses possible improvements to target acquisition performance models.

Dynamic measurement of minimum resolvable temperature difference (MRTD) has been shown to avoid the problems of phase optimization and beat frequency disruption associated with static MRT testing of under sampled systems. In order to predict field performance, the relationship between static and dynamic MRTD (DMRTD) must be quantified. In this paper, the dynamic MRTD of a sampled system is performed using both laboratory measurements and a simulation. After reviewing, the principles of static and dynamic MRTD, the design of a sensor simulator is described. A comparison between real and simulated DMRTD is shown. Measurement procedures are documented for both the static and dynamic MRTD. Conclusions are given regarding the utility of the simulator for performing comparative experiments between static and dynamic MRTD.

The advent of high resolution infrared resistor arrays, has greatly increased the level of fidelity of infrared sensor testing that can be accomplished in the cost effective laboratory environment. However, the sensor output image quality depends on the uniformity of the projector array. In addition to the advanced proprietary design and fabrication process used to create a highly uniform emitter array, Santa Barbara Infrared, Inc. (SBIR) applies a high speed correction algorithm to the incoming data stream that improves the uniformity of the final infrared image. The key to this algorithm is a set of calibrated tables that are measured for each emitter element in the array. SBIR has developed a Calibration Radiometry System (CRS) which is used to quickly perform these high precision measurements for each emitter element. This paper looks at the CRS system, reviews the algorithms used for applying the correction and for making the calibration measurements. It concludes with some initial results showing the effect of the calibration tables derived using the CRS.

Infrared staring sensors used in a large field of view (panoramic) applications such as IRST and MWS are still in need for specialized figures of merit to bridge the gap between feasible laboratory measurements and specification and actual performance. Imaging applications has so far dominated the industry attention and so the need to examine the applicability of conventional analyzing concepts and testing procedures for the new applications was overlooked. In this paper we present a universal test station for panoramic MWS/IRST sensors, designed by Elisra and built by CI-Systems Inc. Following the description of the test station configuration, a set of measurable figures of merit and corresponding test procedures that were devised by the authors to support a panoramic sensor specification are introduced. The figures of merit, replacing conventional resolution, sensitivity and pointing accuracy mapping concepts are suggested and explained as viable alternative to the analogous imaging sensors measurement concepts.

An algorithm is under development, which is based on the Triangle Orientation Discrimination (TOD)-method and predicts the characterization by human observers of camera- system performances. The algorithm combines the TOD-method, an early-vision model, and an orientation discriminator. The algorithm uses the same images as used in human-observer experiments. After correction for the physical properties of the display and the human eye, the algorithm tries to find the orientation of the stimulus. The algorithm can also predict the performance of only image processing using a simple scene-generator instead of a camera setup.

Identification and recognition performance for four staring and two scanning thermal imagers, were measured in an observer experiment using images that were collected during a NATO field trial in Nettuno, Italy, in 1998. The dataset allows validation of the MRTD and alternative sensor performance measures such as the TOD (Triangle Orientation Discrimination threshold). The 75% correct Target Acquisition ranges were compared with the TOD sensor acuity and the MRTD spatial frequency at (Delta) T equals 2 K. The results show that the ratio between the 75% correct TA range and TOD sensor acuity (which is the equivalent to the cycle criterion in MRTD-based models), is independent of the sensor type used, which means that the TOD is a good predictor of sensor performance.

This paper discusses the different detector and systems technologies available today and in the near future, and how they measure up to these requirements. It is concluded that the only affordable alternative today and within the next ten years is QWIP technology. A system based on a Swedish developed and manufactured 320 X 240 QWIP has been demonstrated in the field. This and examples of other systems under development are presented.

Traditional FLIR performance analysis uses analytic models to predict sensor performance characteristics such as modulation transfer function (MTF) and minimum resolvable temperature (MRT). These characteristics are then used in conjunction with empirical criterion such as the Johnson cycle criteria to predict the performance of observers using the modeled sensor. In general, such an analysis suffers from inadequate descriptions of the effects of the background and incomplete descriptions of the observer detection mechanism. Accurate predictions of field performance in a particular setting require the expensive collection of imagery for metric analysis or perception tests. In this paper, an image-based approach is investigated. Using an advanced FLIR simulation, synthetic image sets are generated under controlled conditions. Using these image sets, image metrics are calculated and predictions of target detectability are made using a contrast-to-clutter model and a computational vision model. These predictions are compared to results obtained using a traditional range performance analysis from the MRT based ACQUIRE model. An assessment of the advantages and disadvantages of the image-based approach is given.

FLIR92 and ACQUIRE have become the standard simulation models used in virtually all Forward Looking Infrared (FLIR) system design. Recently, a software program called STADIUM FLIR has been written for use with the U.S. Army's FLIR92 and ACQUIRE models. This software provides many performance and ease of use enhancements for the models. Some of these enhancements include graphical user interfaces for all model parameter entry, data extraction between FLIR92 and ACQUIRE as well as comprehensive plotting of output curves. All data extraction and plotting is automatic and seamless. STADIUM FLIR is based on AET's STADIUM technology which adds powerful Design of Experiments and statistical analysis capabilities to simulation environments. The results are presented both quantitatively and graphically. STADIUM FLIR provides comprehensive plotting capabilities for both raw data as well as `overlayed' statistical variability data. STADIUM FLIR provides the power to perform multiple FLIR92 and ACQUIRE simulations with inputs (even multiple targets) varying over user specified ranges. This paper will describe the software and how it enhances the power of FLIR92 and ACQUIRE.

With the increased interest and use of staring infrared focal plane arrays, the characterization of fixed pattern noise in task performance is becoming more important. Past work includes theoretical treatments and laboratory measurements to describe the characteristics of fixed pattern noise on target acquisition performance. This is the first target acquisition experiment that describes the relative effects of fixed pattern noise and temporal noise on target identification. Static infrared tank images were processed with six different levels of fixed pattern noise and six different levels of temporal noise. A perception experiment was perform where 10 US Army soldiers were tasked to identify the tanks through the combinations of noise. Additive noise was applied in both Gaussian and uniform distributions. The results allow a direct comparison between the effects of fixed pattern noise and temporal noise on target identification.

Staring array imagers can exhibit sampling artifacts. Dither is a mechanical means of raising the spatial sampling rate without increasing the number of detectors on the focal plane array. Diagonal (two-point or slant-path) dither is easier to implement mechanically than rectangular (four- point orbow-tie) dither. Also, diagonal dither generates half the data rate of rectangular dither. However, diagonal dither does not sample the image as effectively as rectangular dither. The cost and complexity advantages of diagonal dither must be traded against the expectation of reduced performance. This paper discusses analytical and empirical predictions of the performance difference between diagonal and rectangular dither.

This paper presents recent design improvements for the Gray- Level Co-occurrence Metric based Trackability Metric and the resulting performance enhancements. As in depth performance trade study for design modifications including a hot spot metric was performed. These enhancements to the original Trackability Metric should provide better state-to-the-art performance prediction and more accurate performance modeling for specific imaging autotracker design. The results of the study and the final implementation of the Trackability Metric are presented.

The forward looking infrared subsystem of the Northrop Grumman LITENING II targeting pod incorporates a plateau algorithm for automated gain control (AGC) to enhance the image display for visual detection and identification of targets. We developed an adaptive, automated algorithm to optimize the input parameters of the baseline plateau AGC in real time based on rudimentary features of the intensity histogram. The optimized plateau algorithm retains the desired contrast enhancement provided by the baseline algorithm while maintaining an overall balance of target and background intensities without unacceptable boosts in noise levels.

A Kalman filter is developed to estimate the temporal drift in the gain and the offset of detectors in focal-plane array sensors from scene data. The novelty of this approach is that the gain and the offset are modeled by random sequences (state variables) which must be estimated from the current and past noisy scene data. The gain and the offset are assumed constant over fixed-length blocks of frames; however, these parameters may slowly drift from block to block according to a temporal discrete-time Gauss-Markov process. The input to the Kalman filter consists of a sequence of blocks of frames and the output at any time is a vector containing current estimates of the bias and the offset for each detector. Once these estimates are generated, the true image is restored by means of a least- mean-square error temporal FIR filter. The efficacy of the reported technique is demonstrated by applying it to two sets of real infrared data and the advantage rendered by the Gauss-Markov model is shown.

Infrared Search and Track (IRST) systems are designed to automatically detect, locate, and track infrared objects and targets. A generic method for predicting the detection performance of an IRST is presented in which the effect of a 3D cluttered background can be considered. The performance predictor includes models of a target, background, atmospheric attenuation, IR sensor of the IRST, and it also includes algorithms for target detection.

The German Aerospace Center (DLR) and its industrial partners are working on two new spaceborne fire missions: (1) the Bi-spectral IR Detection small satellite mission (BIRD) to be launched in autumn 2000, and (2) the Innovative Infrared Sensor System FOCUS to be flown as an early external payload of the International Space Station. Both BIRD and FOCUS will use MIR/TIR solid state pushbroom imagers with real time digital signal processing providing an adaptive and very high dynamic range in radiometry. Promising results are obtained with the BIRD Airborne Simulator which has been flown at DLR in several airborne campaigns since 1997.

T-CAT (Thermal Camera Acuity Tester) is a thermal `eye chart' for measuring the spatial resolution (`sensor acuity') of thermal imaging systems. It is a small, portable system,, that is used in a similar way as the optometrists' visual acuity charts. The design is an implementation of the TOD (Triangle Orientation Discrimination) method for Electro-Optic sensor characterization that has recently been introduced. The purpose of the T-CAT is do quick and easy assessments of thermal imager performance, e.g. for routine testing such as go/nogo decisions. T-CAT measures system performance including the human observer, which is still the best way to assess actual field performance of an imaging device.

`KENIS', a complete, high performance, compact and lightweight thermal imager, is built around the `OSPREY' infrared detector from BAE systems Infrared Ltd. The `OSPREY' detector uses a 384 X 288 element CMT array with a 20 micrometers pixel size and cooled to 120 K. The relatively small pixel size results in very compact cryogenics and optics, and the relatively high operating temperature provides fast start-up time, low power consumption and long operating life. Requiring single input supply voltage and consuming less than 30 watts of power, the thermal imager generates both analogue and digital format outputs. The `KENIS' lens assembly features a near diffraction limited dual field-of-view optical system that has been designed to be athermalized and switches between fields in less than one second. The `OSPREY' detector produces near background limited performance with few defects and has special, pixel level circuitry to eliminate crosstalk and blooming effects. This, together with signal processing based on an effective two-point fixed pattern noise correction algorithm, results in high quality imagery and a thermal imager that is suitable for most traditional thermal imaging applications. This paper describes the rationale used in the development of the `KENIS' thermal imager, and highlights the potential performance benefits to the user's system, primarily gained by selecting the `OSPREY' infra-red detector within the core of the thermal imager.