An infrared scene generator code has been developed for use in testing electro-optic systems. The scene generator combines sophisticated computer graphic techniques with first-principals background modeling. The code generates apparent radiance images of a scene composed of polygonal and/or quadric objects for a specified viewing geometry, spectral band, weather, time of day, location, season, and sensor field of view. Surfaces may be any one of 15 different materials and any surface can be given a texture to make the generated scene more realistic. The warming of surfaces caused by sunlight is also modeled. All parameters, including the effect of the intervening atmosphere, the weather, material type, material texture, and the system modulation transfer function, are contained in independent data bases. The use of data bases, combined with efficient programming, make the execution time short. This allows the user to quickly see the effects of different conditions on the scene of interest and how varying conditions will effect a sensor system.

The Human Resources Laboratory has developed a dynamic IR scene simulation system to support research on simulation requirements for Forward Looking Infrared (FLIR) sensors in fighter aircraft. A primary objective in the development of this system was that FLIR imagery be part of an overall simulation package including navigation and targeting FLIR, Synthetic Aperture Radar (SAR), and out-the-window visual imagery. FLIR imagery is derived from a computer-generated visual scene. To simulate FLIR, color codes for objects in the visual data base are replaced with gray-scale values based on predicted IR exitance. The gray-scale values are constructed using software developed by the Applications Research Corporation which combines feature and surface codes from the Defense Mapping Agency source data together with wavelength dependent absorptivity and run-time scenario specifications including gaming area latitude and longitude, time of day, season, and weather to estimate radiant exitance for each object. Exitance values are then converted into gray-scale values. Sensor effects such as gain level, polarity, field of view, and modulation transfer function are added to the FLIR imagery by a General Electric-developed post-processor. Sensor characteristics can be separately programmed for a navigation FLIR which is presented on a simulated Heads-Up Display (HUD) and for a targeting FLIR. The effects of user inputs to the IR prediction software on radiant exitance, functions of the post-processor, and selection of texture patterns will be discussed. Fidelity requirements for simulation of FLIR in flight training will also be addressed.

Modern test applications require rapid change of simulated scenes that need to be displayed in optical spectra ranging from the ultraviolet to the far infrared. This paper reports on dynamic scene projection in this wide spectral range which is based on cathode ray tube phosphor technology and is capable of displaying scenes that vary rapidly in shape, wavelength, and intensity. The principle is suitable for relatively low cost, high resolution projectors emitting in ultraviolet, visible, and infrared spectra. To optimize the high-speed performance, a computerized display control was developed.

A large new calibration facility, designated BIRS, has been recently completed and made operational at the Boeing Kent Space Center near Seattle. Initially configured for calibration of the Army's Airborne Optical Adjunct (AOA) sensor, the facility has the capability to perform radiometric and goniometric calibrations of large infrared sensors under simulated operational conditions. It also has the potential for growth into other applications such as hardware-in-the-loop and seeker testing. In its present configuration, the BIRS facility provides environmental simulation for endoatmospheric sensors. The facility is designed for easy adaptation of the sensor chamber and optics chamber to simulate exoatmospheric environments as well. This paper will describe the development of this unique capability and will present aspects of the facility useful to the potential user.

The performance of an infrared space surveillance system is ultimately limited by the celestial background. The impact which the various components of this background have on the system is presented, as is the requirement for an accurate scene description capability for system design trade-off and performance analyses. The Geophysics Laboratory, partially through contract to Mission Research Corp., is creating a Celestial Background Scene Descriptor to meet this need. The development path for this description capability and its current status is presented. A low resolution, image-based descriptor will be available shortly, as will models of the point source and diffuse zodiacal backgrounds. Procedures are discussed which will extend this preliminary descriptor to higher resolutions, lower limiting brightnesses, and a broader spectral capability.

Details will be given about the recently completed facility for the calibration of infrared sources in a low background environment. The basic components of the facility are a large (60cm diameter by 152cm long) stainless steel vacuum chamber housed in a soft-wall cleanroom. A low background environment inside the chamber is achieved by cooling internal cryoshields to temperatures less than 20K using a closed cycle helium refrigerator. Sources of up to 30cm on a side can be inserted into the chamber for calibration. Total radiant power from the blackbodies is measured with an Absolute Cryogenic Radiometer. Plans will be discussed for future enhancement of the system allowing for measurement of spectral and angular distribution of the emitted radiation and possible experiments which could utilize the full capabilities of this system.

Radiometric calibration for the MWIR and LWIR is time consuming and difficult. Recent primary radiometric standards are not directly useful in the detector vendor's test equipment. It is proposed that InSb PV detectors can be used as primary standards, using the self-calibration methods similar to those recently described for Si diodes.

This paper presents the process by which calibrations of radiometric detectors (either alone or as a part of electro-optical sensor systems) are conducted and applied to meaningful system applications. It is assumed that if a calibration is to be performed one desires that the calibration be "accurate". Accuracy always implies a comparison with a recognized standard. When one is dealing with a common property (say length) this may be straightforward. With a derived quantity (such as detector responsivity as an example) it becomes more difficult. There are two definitions for accuracy which are commonly used. One states that accuracy is "the absence of bias". Another view relates accuracy to the overall uncertainty of the measurement including a bounded but unknown bias. This paper is predicated upon the latter definition. Fig. 1 is used to explain why this point of view is taken.

A unique apparatus for measuring the specular reflectance of bare and anti-reflection coated optical components, at cryogenic temperatures, has been developed through a joint effort between the Metrology Department and the Electro-Optical Center of Rockwell International Corporation's Autonetics Electronics Systems Division, in Anaheim, California. An invention disclosure for the apparatus has been filed with Rockwell.

Rockwell International Corporation Metrology, in Anaheim, California, has implemented a unique capability for measuring the transmission of plane optical components at cryogenic temperatures. This capability supports Rockwell's Electro-Optical Center in the manufacturing and testing of short- and long-wave IR detectors designed to operate at low thermal background levels.

Light from a small aperture at the focal plane of a collimator appears as a point source located at infinity. Light from the aperture of a blackbody is imaged onto the small focal plane aperture. This defines the input cone of light. The light exiting from the focal plane aperture into the light cone collected by the collimator is calculated in this article using a diffraction model. The input cone angle is shown to be the only system parameter affecting the spectral transmission of a given aperture. Data from the POST test chamber is presented to demonstrate the model. The data is analyzed at longwavelength (LWL) and at shortwavelength (SWL) to yield a measured value of the input cone angle. The SWL data, diffraction losses less than 20%, and the LWL data, diffraction losses from 20 to 60%, give nearly identical values for the cone angle. The measured cone angle is near to the design value of the cone angle.

This report presents a closed-loop automatic tracker for imaging IR-based fire control. A tank target state estimation filter is discussed and major tradeoffs in the implementation of Kalman filters and tracker rate-aiding are given. Tank target maneuver capabilities are surveyed and results are incorporated in the filter. The developed augmented Kalman filter is based upon nine filter states and three measurements. Simulation studies show acceptable target prediction accuracies for tank targets even for constant Kalman gains. Gain scheduling is performed during filter initialization or when a "target maneuver" is detected from testing the statistics of filter residuals. The augmented Kalman filter uses range and line-of-sight (LOS) angular rate measurements obtained from platform-mounted range and angular rate sensors. Filtered inertial angular rates are used for tracker rate-aiding before and after "breaklock"-if this occurs (e.g., if target gets occluded). During the tracker "coast-until-reacquisition" phase the propagated filter rates are used to command the gimbals.

Pixel or image based tracking systems, which rely on correlation or contrast algorithms, have been shown to be unreliable for tracking of extended objects. We present an object based tracking algorithm which first segments the target from its background and then performs silhouette matching to perform tracking. Isssues of segmentation are discussed and a segmentation algorithm is presented which has been demonstrated in simulation and is presently being implemented in hardware. The segmentor selects an optimal threshold which maximizes the edge strength between target and background. A silhouette is than extracted for each segmented object and matched to the reference. Silhouette matching techniques are discussed and real-time implementation trade-offs are presented. A silhouette matcher that applies to a wide range of tactical scenarios is presented. The silhouette matcher finds the best object match despite partial occlusion and/ or the so-called "segmentation wobble" phenomenon. The resulting silhouette tracker is examined through extensive testing in both lab and field situations.

A real-time electro-optic tracker is described, which uses knowledge about scene processes to resolve tracking problems involving obscuration, partial obscuration, target-like clutter or other targets in the field of view that may cause the tracker to break track. The tracker incorporates a scene model which combines extracted information from the sensor imagery with prior knowledge of scene formation processes and a dynamic model of object movement. The tracker maintains a tracking window large enough to allow measurement of (clutter) objects near the target so that if the target does indeed occlude or becomes occluded by the object, the scene model has the capability to resolve the conflict and maintain track on the designated target.

This paper describes a knowledge-based approach to the detection and classification of ground-based military target, formations in infrared (IR) image sequences at long range from an airborne platform. The motivation for this ap-proach is that at long range military vehicle targets in IR imagery typically consist of only a few pixels, have low contrast and exist in a high clutter environment making reliable detection, classification or identification of indivi-dual targets extremely difficult. Thus, the system detailed here detects targets based on their membership in a military target formation or convoy, relying on the structure or spatial organization of the formation and not individual target signatures. The system consists of two major components: Point Target Detection and Tracking (PTD); and Knowledge-Based Constraint Hierarchical Target Formation Detection and Classification (TFD). The first component detects and tracks both stationary and moving point targets which are passed on as candidates for clustering into target formations by a rule-based matching process in the second component. The PTD module makes use of image pre-processing, stationary and moving target detectors, image registration for sensor motion compensation, temporal processing for detection consistency over a large number of frames, in addition to inter-frame target and formation tracking. The TFD module consists of an expandable knowledge base containing characteristics of typical target formation types and a flexible set of rules that are used to match reference formations to actual formations in an image. The paper contains a complete description of the system methodology and provides experimental performance results of the system operating on real IR data.

Evaluation of military ground combat vehicle system's IR signature has in the past been a time consuming measurements using a thermal camera. Many limitations exists using modify and measure techniques for thermal signature design and analysis including cost, logistics and limited data sets. One further consideration occurs if the thermal signature of a concept system is to be evaluated at the CAD/CAM stage of development. To help solve these problems, the background and, target are simulated using high fidelity thermal prediction models. This paper will present recent advances in background and target thermal models. Several examples and image sequences will be presented showing detection performance through dynamic scenes.

Many air target detection scenarios involve a sky background. In most cases, sky backgrounds include clouds, which represent a major source of clutter. Development and testing of algorithms to deal with these scenarios is aided by synthetic scenes which are readily controllable and into which targets with a range of characteristics can be inserted. These scenes should be as physically accurate as possible. If motion is to be incorporated, it is necessary to generate sequences of scenes and to utilize three-dimensional clouds and cloud fields. At the same time, it is desirable to be able to generate the scenes and scene sequences rapidly, while varying the physical parameters.

Misunderstandings are described concerning blackbodies, simulators, cavities and apertures. Blackbodies are identified as ideal (not real) items, while the expression blackbody simulator is recommended for real (not ideal) approximations to blackbodies. Surface and volume considerations for blackbodies and blackbody simulators are addressed. Blackbody and blackbody simulator history are traced. The important roles of blackbody simulator cavities and blackbody simulator apertures are emphasized.

Three types of cavity emissivities are described based on wavelength, two types based on temperature, and five types based on geometry. Nonisothermal cavities are shown to be best represented by a two-temperature emissivity while most other items, including isothermal cavities, are customarily represented by a one-temperature emissivity. Radiance based emissivity and local hemispherical emissivity are recommended for making cavity evaluations and comparisons, while the hemispherical emissivity of the cavity aperture and a cavity emissivity that is based on detector considerations are shown to be faulty.

An Infra-Red Target Generator (IRTG) has been developed to generate a scene whereby the motion and IR signature of an aircraft and countermeasures may be simulated. This scene is presented to a missile seeker head in a surface-to-air or an air-to-air engagement. The opto-mechnical design required the mixing of two IR beams representing aircraft and countermeasures in the 1-5 um spectral band. Both sources may be modulated in intensity and time with a series of rotary neutral density filters and shutters. The focal plane components also allow for variation of source angular size.

The TACOM Thermal Image Model (TTIM) is an interactive computer code which simulates the performance of IR imaging sensors operating in realistic battlefield environments. Its ability to produce simulated imagery has proven useful in applications including sensor and tracking algorithm simulations, obscurant model development, and vehicle design evaluations. This paper reviews the development and current functionality of TTIM, concentrating on descriptions of the most recent distribution version (3.0) and two new modules which treat staring Focal Plane Array (FPA) and Image Intensifier sensors, respectively.

A thermal imager undergoes several different types of test in its final assembled form in order to ascertain that it performs to specification. The actual measurements made on the imager will depend to a certain extent on the use for which it is intended. Where imagers are to be used in applications such as surveillance and navigation, in which a human observer views the final image, the measurement of minimum resolvable temperature-difference (MRTD) is both relevant and customary. The MRTD provides data which can be used to predict performance in terms of the temperature-difference and range at which specific targets can be identified with a given probability. It is a subjective measurement and its value lies in the fact that it assesses the combination of imager and observer. Its main disadvantage is the variability in measured results which arises from its subjective nature.

A thermal imaging computer simulation was developed for use in design studies of airborne forward-looking infrared (FLIR) systems. This simulation dynamically generates in-flight imagery while modeling flight scenarios, thereby facilitating the evaluation of end-to-end system performance. Program flexibility, which was a major concern, is realized through a modular code design as well as user specification of all critical imager parameters, the scenario geometry, and flight characteristics. A standard type of faceted target model and a generic platform model are used. The simulation includes a nonuniform background, atmospheric path transmission and radiance, and additive noise, as well as system image degradation. Realistic optical and electronic characteristics are modeled using an equivalent Gaussian point spread function (PSF) in the image domain convolved with the spatially oversampled image. The oversampled image implementation uses spatial overlap techniques in place of standard "z-buffering" to ensure proper image plane irradiance distributions. The detector responses are computed by performing spectral radiance integrals across the detector bandpass, weighted by a user-defined specific detectivity. The resulting two-dimensional detector image may then be used to test target detection, recognition, and tracking algorithms. To improve computational efficiency, the relative platform-target geometry is considered so that time-consuming rotations and projections of the target data bases are performed only when physically relevant changes occur.

To accurately measure the effectiveness of an infrared sensor for military or industrial purposes, the interaction between the human operator and the sensor must be considered. In particular, modeling the search process is crucial towards measuring the length of time and probability of success of target acquisition in a particular scenario. The CNVEO (Center for Night Vision and Electro-Optics) search model, used by the US armed forces and several allied coutries, is essentially a single-target, single-observer model. In this paper, we extend the model to the case where several observers are searching for the target under changing atmospheric conditions. The obscurant/atmosphere is characterized by an average transmission coefficient, a spatial variance, and a persistence time during which the obscurant/atmosphere is assumed to be fairly constant. Correlation between observers is determined by the degree that a particular level of obscurant for one observer implies the same level of obscurant for a physically separated second observer at the same time. Analytical solutions and computer simulations show that the degree of correlation between observers greatly influences the probability detection and the average time for detection; the greater the degree of correlation, the longer the average time for detection. In addition, radically different results are obtained depending on the persistence time of the obscurant. Methods to implement such a model in a battlefield simulation are described.

The assumed underlying probability of detection distribution is important when trying to estimate the mean of the population from a limited sample such as that obtained from minimum resolvable temperature (MRT) tests. If the underlying population is normally distributed, then the best estimate of the population mean is the arithmetic average of the observers' individual MRT values. On the other hand, nearly all visual psychophysical data is plotted on a logarithmic scale with log intensity increments. Thus it is reasonable to assume that MRT responses should also be treated as a logarithmic response. With a log-normal distribution, the population mean is estimated from the geometric average of the observers' responses. Choosing a log-normal underlying distribution over a linear distribution does not affect existing forward-looking infrared (FUR) detection theory, since the threshold value is identical for both distributions and the distributions are essentially the same from 20 to 80 percent cumulative probability. The linear normal distribution is not bounded and therefore, mathematically, can provide a finite probability of detection for unrealizable negative signal-to-noise ratios. Since the log-normal distribution is bounded to positive values it appears to adequately represent the real world. To determine the underlying distribution for the standard MRT four-bar target, 2700 detection responses were obtained for various signal-to-noise ratios and spatial frequencies. Although any individual observer may have a linear distribution, the resultant composite data easily fits a log-normal distribution. Furthermore, limited laboratory data on actual FLIRs also indicate that a log-normal distribution provides a better mathematical approximation. As a result, the geometric average of observer responses is recommended for determining the composite MRT value.

A method has been developed for comprehensive end-to-end tests of scanning optical sensor systems, including the determination of their line of sight relative to a basic inertial reference frame and of their ability to process data obtained from optical stimuli originating from tactical scenarios.

The basic modes of electrical operation of photoconductive optical radiation detectors are analyzed. The nonlinearity inherent in "voltage mode" measurements can be eliminated by using "current mode" measurements. A HgCdTe photoconductive radiometer has been designed, based on this analysis, which measures the current through a biased detector. A built-in calibrating capability in the circuits of the radiometer makes it possible to eliminate the effects of a long-term drift in the bias voltage, thereby achieving a higher precision.