The evolution of the Sprite detector from its invention in 1974 to the present is reviewed. The use of anamorphic optics to reduce the effects of carrier diffusion, together with changes to the device shape has produced very high spatial resolution. Improved material and two-dimensional structures have further increased the thermal sensitivity. A brief description of the technology for two-dimensional, electronically addressed diode arrays is also given, together with a discussion of the mechanism of p to n type conversion.

The first laboratory images from an active multispectral 3D imaging sensor under development for cross-country navigation of an autonomous vehicle are presented. Images of the reflected return from three spectral channels, 0.532 μm, 0.802 μm, and 1.064 μm, and a range channel were acquired simultaneously in performance tests during the first stage of development of the sensor. An analysis of the image data showed strong agreement between the actual sensor performance and calculated estimates. Position, shape, and spectral discrimination of objects are evident in the images and will function as input to the autonomous vehicle for locating and classifying obstacles and terrain.

This paper presents the design of a VHSIC multimode tracker being developed for the US Army Missile Command that will interface with imaging infrared (IIR) sensors implementing DOD-developed Phase I VHSIC devices. The system is being initially configured for demonstration testing with MDAC's Mast-Mounted Sight system which uses a common module FLIR and with a second generation IIR seeker for the Hellfire Fire and Forget Missile. Requirements for the VHSIC processor are discussed and the architecture presented as derived from the requirements. This design incorporates array processors, 1750A processor, and a three-level processor-to-processor interconnection network into a modular design that can be scaled to match performance and size/form factors to meet acquisition sight and seeker implementations. A multimode algorithm tracker includes parallel calculations of spatial correlation, centroid, and conformal gate for target tracking. The control software provides an environment for multitarget tracking. A field test program will evaluate the VHSIC tracker interfaced with the Mast Mounted Sight thermal imaging sensor in 1988. A captive flight testing of an advanced IIR seeker is planned at MICOM in the fall of 1988.

Staring infrared Focal Plane Arrays (FPA's) utilizing integrating-mode detectors suffer from insufficient charge storage capacity to stare at a room-temperature scene for standard television frame times. This paper presents a high-speed readout technique, with subsequent frame averaging, that extends the charge storage capacity of monolithic Metal Insulator Semiconductor (MIS) FPA's by a factor of 32. A 128 by 128-element indium antimonide FPA is combined with a silicon row driver and preamplifier/multiplexer chips and packaged in a Common Module Dewar. The FPA assembly can be read at 960 frames/s and up to 32 frames averaged off focal plane for an improved signal-to-noise ratio. This work has been supported by Northrop Corporation's Independent Research and Development (IR&D) funds.

The effects of background temperature changes on the effective, or apparent, blackbody radiant output are analyzed as a function of thermal imaging system (TIS) spectral bandpass characteristics. The error sources are quantified and methods of correction are presented.

Imaging spectrometers operating in the visible through midwavelength infrared (MWIR) spectral regions have been used or planned for several missions in the exploration of the solar system. These instruments generally require cryogenic cooling for their photo-detector arrays. A current concept of the instrument system is described herein. The feasibility of an uncooled imaging spectrometer is assessed through evaluation of the tempera-ture environments of the instrument in various spacecraft orbits, and by analysis of available photo-detector materials and sensor focal plane array configurations. It is concluded that the uncooled approach appears practical for many solar system missions with current technology.

The requirements and technology solutions for present day CO2 Doppler laser radar receivers are explained and heterodyne sensitivity for state-of-the-art Hg1-xCdxTe high speed detectors quoted. Future system needs and the technologies which will be applied to meet those needs are discussed.

Requirements placed on surveillance infrared sensors have steadily increased and technology developments to meet these requirements are continuing. These advancements in sensor technology place ever increasing demands on the test and calibration equipment utilized to verify performance and provide preflight calibration. The test chamber collimator optical system described herein meets these stringent requirements.

This paper will discuss the use of aluminum alloy castings as diamond machining substrates for optical surfaces. Surface texture, infrared reflectivity, and temporal stability results from cast diamond machined samples will be presented. Physical data on these alloys will be given along with casting considerations.

This paper describes in overview the current status of analytical models used to design, evaluate, and predict the performance of passive thermal imaging systems (TISs). The scope is limited to real-time imaging systems and does not consider such equipments as infrared search-track sets, line scanners, or narrow band laser augmented sensors. The purpose is to establish the current limits of our analytical capability, problems in our ability to analyze TIS equipment, and improvements that are required at this time.

Methods are discussed for analytic modeling of staring autonomous infrared systems for use in point-source detection. Our analysis generally reviews frequency domain methods that can be used from a system engineering perspective to conceptually design an infrared sensor and predict its performance in conjunction with a multidimensional matched filter to perform either temporal, spatial or spatio-temporal (3-D) image processing. Specifically we will be concerned with point-source targets that have a unique temporal behavior and are embedded in background clutter. Both the signal and the clutter can have some level of apparent motion associated with them in terms of a fixed velocity. The modeling process is fully parametric and can include all salient features of a time-dependent target, stochastic back-ground, atmospheric propagation, optical system and staring infrared detector array. We show that this modeling technique lends itself very well to parametric analysis that is often very useful in design trade-offs and optimization studies. An example is given that deals with moving target detection, temporal filtering, spatial filtering, and spatio-temporal filtering.

The two-dimensional recognition range model proposed six years ago is reviewed and its current criteria are listed. This model accounts for the vertical resolution of the sensor as well as the horizontal, whereas the standard one-dimensional NVL model does not. For the advanced sensors with improved vertical MRT, this model is required to predict improved range performance. Analysis of example sensors is used to extend the applicability of the model from current to advanced sensors. The design implications for advanced sensors are summarized.

A set of software and models for computer-generated infrared imagery of a complex synthetic European village scene has been developed at Georgia Tech Research Institute. The results of a study utilizing these software and data sets are presented, showing the effects of seasonal and diurnal variation on the output imagery. A description of the software is given, followed by details of the modeling and database construction process. Examples of the imagery are given, followed by a brief discussion of further applications and directions for future research.

The utility of a thermal model can only be realized after careful validation. At KRC a series of tests have been conducted to validate The Physically Reasonable Infrared Signature Model (PRISM). PRISM is a first principles semi empirical model developed for but not limited to thermally modeling vehicles. There are many factors which go into accurately modeling a vehicle's thermal characteristics, error can have many sources. The ultimate goal of validation should be to isolate the sources of error as much as possible to gain a better insight into model behavior. Tests were conducted at KRC using a simple facet structure consisting of four one foot square aluminum plates. This structure was used to minimize the errors that result from facet geometry. The tests were run in both mid summer and winter to observe any change in model accuracy that is seasonally related. Tests were then conducted using a M2 Bradley to validate the model for vehicles.

For infrared optical systems a point spread function pt (or Pd ) is calculated in terms of the spot diagram obtained through geometrical ray trace (or the theory of Kirchhoff diffraction). Also is calculated the ratio T of the energy passing through the system onto the exit window, to the total energy radiated from a plane target into the entrance pupil. Formulas are derived separately for the following different cases: a circular or rectangular ring diffraction pupil, a blur of equivalent intensity distribution or a Gaussian blur, and a circular or rectangular field stop.

The paper will discuss a simulation code developed at Aerodyne Research, Inc. to reflect the major sources of clutter that are of concern to staring and scanning sensors at resolutions where discrete objects, e.g. buildings are not resolved. The Aerodyne Infrared Earth (AERIE) code uses topographic features to model terrain, typically from DMA data, but with a statistical overlay of the critical underlying surface properties (reflectance, emittance, and thermal factors) to simulate the resulting texture in the scene. Reflectance and emittance from broken cloud decks are also included, as are effects of cloud and terrain shadows, and atmospheric transmittance and radiance. The AFGL LOWTRAN6 model is included as a subroutine. The code is based on first principles radiometric models that preserve the integrity of the resulting simulations as various scenario properties (sun angle, spectral band, etc.) are varied. The paper will discuss the basic models utilized for the simulation of the various scene components and the various "engineering level" approximations that are incorporated to reduce the computational complexity of the simulation.

High fidelity IR scenes are needed to support the development of target detection and tracking techniques. It is necessary to have flexible and cost efficient testing methods to test the effectiveness of sensors and detection algorithms. While directly measured data can be obtained and used to vary sensitivity, dynamic range, spectral bands, and spatial and temporal resolution for evaluation of future sensor systems, synthetic infrared scenes can provide both radiometrically accurate information and exhibit realistic spatial variations. Two elements that contribute significantly to sensor performance are target radiance and background clutter. The IR scenes generated by a terrain, cloud, and target computer model provide for both of these elements. Many aspects of the phenomenology incorporated into each of these computer models have been validated. The scenes are raster images from a composite of target images inserted into a background radiance map. Multiple scenes can be linked together by using a common background and variably positioned target, thereby simulating motion of a target through a background. A series of scenes can relate to each other by performing parametric studies on various radiation sources in the scene. The scenes can be viewed with a predetermined timing sequence. This can simulate real time inputs to sensor systems.

Along with the advent of the information era comes an unprecedented increase both in data volume for storage and in transmission rate for communication links in both terrestrial and space applications. This increase more often than not places a severe demand on the need for higher capability in both the communication link and data storage. This problem can be significantly reduced by utilizing data compression techniques. This paper presents a practical, efficient and noiseless data compression technique. This techniques is a simplified version of the BARC image compression system which will be used on the Galileo mission to Jupiter [1], [2]. A conceptual VLSI design capable of meeting real-time processing rates is provided. The VLSI approach yields low power, weight and volume requirements that make this form of BARC an attractive addition to future low budget space flight missions. Potential applications of this design to future spaceflight instruments, such as the Visual Infrared Mapping Spectrometer (VIMS) and the High Resolution Imaging Spectrometer (HIRIS) are described.

On the Lunar Geoscience Observer (LGO) project a Visible and Infrared Mapping Spectrom-eter (VIMS) instrument has been proposed. This instrument will have science data input rates in the hundreds of kilobits per second (kbps) and an average telemetry output data rate of 4 kbps. Techniques that can be used to reduce the throughput of the instrument are editing, summing and averaging, data compression, data preprocessing, pattern recognition and snapshot data taking. Due to instrument limitations in the buffer memory size and processing speeds, a careful selection of the available techniques must be made.

Median filters are of interest because they eliminate impulse noise and preserve image edges. The output of these nonlinear filters depends upon the local statistics and therefore, no analytic expression exists to describe filter performance. Performance can only be inferred by using representative inputs. For imaging systems operating at or near the Nyquist frequency, median filters will eliminate unresolved and small targets. The effects of median filters are noticeable on periodic targets such as the four-bar resolution target used for minimum resolvable temperature measurements.

A noninvasive technique for the measurement of image jitter has been developed. This technique has been applied to a common module thermal imager to show that scanner position oscillations about linearity are exacerbated by power supply fluctuations.

We have measured the noise voltage histories of adjacent and nonadjacent detectors at the system's level of a common module FLIR in synchronism in order to form estimates of the channel interactions. We used the Bartlett method to form estimates of the cross correlation function and the squared coherency spectrum. Unambiguous evidence for correlations between channels was not found.

This paper keynotes the "Infrared Automatic Target Recognition" session. Its thesis is that: 1) To win in the battlefield of the future against a numerically superior enemy equipped with weapons of unprecedented capabilities, we must have the "force multiplier" promised by technology 2) Automatic target recognition (ATR) is one such technology with attributes that provide this leverage 3) ATR can provide this enhanced performance for several current, specific, combat missions 4) ATR can be applied to broader applications if several technical challenges can be met 5) DoD is recognizing the potential of ATR to provide "the winning edge."

This paper describes a suite of target cueing algorithms which has been developed for the recognition of ship targets in the open ocean through FLIR Imagery. Imaging prepro cessing is first used to remove pattern and temporal noise. A relaxation technique is implemented to extract the target's silhouette. The superstructure profile is then obtained and classification is performed based on low-order coefficients of the discrete Fourier transform of the profile. This classification approach was found to have a 93% accuracy for short ranges (7-11 miles) and 70% accuracy for long ranges (11-20 miles) for eight target classes tested against 11398 images. Finally, a terminal homing algo rithm is described which incorporates scene tracking for maintaining track on a selected aimpoint which demonstrates superior performance over more conventional approaches.

Ambiguity in the use of the phrase "model-based" in the discussion of automatic target recognition (ATR) algorithms is identified. Some terminology designed to allow more precise discussion of model-based concepts is proposed, with the meaning of the terms delineated by discussion and example.