The TACOM Thermal Image Model (TTIM) will be described in some detail in this paper. This model maps a target/background scene radiance map pixel by pixel into an image as seen through a thermal imaging sensor. The development of TTIM was motivated by a need to analyze the thermal signatures of current and concept combat and tactical vehicles. A number of other applications will be discussed for TTIM including sensor performance, battlefield obscurants and natural atmospheric effects.

The synergistic combination of laser radar and infrared sensors segments the man made objects out of cluttered back-ground and enhances the the distinguishing features of those objects. The gradients extracted from laser radar and infrared sensors provide the basic tools. The nearest neighbors' patterns aide in determining the actual location, size and boundary of the targets. The combination of processed laser radar and infrared data through multiple masking presents information that is not visible in either of the sensors by itself. This paper will discuss the gradient based methods of target segmentation and enhancement and present some characteristic results.

Proper registration between infrared (IR) and millimeter wave (MMW) scenes in an air-to-ground dual-mode MMW/IR sensor is discussed. A method to autonomously compensate for mis-registration errors between scenes is given. It is shown, in an example, that this will allow for a high probability of target detection without significantly increasing the probability of false-target detection.

This paper presents a new computer vision technique based on multisensor fusion for interpreting images of outdoor scenes. The technique is based on the synergetic integration of information derived from thermal and visual sensors. Information integration is implemented at different levels of abstraction in the interpretation hierarchy i.e., at the pixel and at the symbolic levels. The approach establishes features based on surface heat flux estimates that describe intrinsic thermal behavior of imaged objects. Heuristic rules are employed in a decision tree classifier to categorize imaged objects as being either vegetation, building, pavement or a vehicle.

Texas Instruments has completed the development of a new, cost-effective family of FLIR systems named "Falcon Eye" for current and future fighter aircraft. Falcon Eye is based on advanced detector technology in a system concept that merges the FLIR function of a navigation pod and many of the functions of a targeting pod within a conformal sensor package. This new FLIR packaging concept will virtually eliminate the drag penalty associated with externally mounted FLIR systems.

Precision munitions employ various electro-optical sensors that enable them to accurately locate and strike targets. These electro-optical sensors perform a variety of functions, including hot spot detection, false target discrimination, stand-off fuzing, range to the target, aimpoint selection and potentially target profiling. Of interest are three primary electro-optic sensor categories: passive IR, active IR and dual mode (IR/MMW) sensors. Examples of fielded and prototype sensors built for these three categories are presented and discussed.

A multi-aperture sensor concept has been developed as a solution to the U.S. Army and Air Force's need for a class of low-cost, adequate performance optical sensors capable of generating detection and tracking information for precision-guided munitions. The Honeywell multi-aperture sensor consists of a number of small apertures. Behind each aperture is a low-cost, single-element IR detector. The concept offers a wide field-of-view while maintaining a high resolution center region. The concept is an attractive solution for classes of munitions where low cost is extremely important. A proof-of-principle device has been built and the concept demonstrated in the lab.

Multifunctional silicon integrated sensors have been fabricated for simultaneous measurement of several physical and chemical variables. These sensors are based on the compatible introduction of thin films of piezoelectric and pyroelectric zinc oxide combined with conventional MOS processing technology in the fabrication of 1) a 64-element infrared sensing array, 2) a mass air-flow sensor, 3) a tactile sensor array for precision robotics applications, 4) carbon monoxide sensor, 5) surface-acoustic wave chemical vapor sensor, 6) microbeam accelerometer, and 7) infrared charge-coupled device imager. This paper specifically addresses technology issues involved in incorporating sensors with high-performance silicon integrated circuits. Focus is directed toward 1) zinc oxide based sensors, 2) silicon micromachining pf sensor membranes, and 3) compatible processing issues necessary for a generic integrated sensor technology. Performance examples for a 64-element room-temperature pyroelectric imager are presented.

The integration of information is a central issue for Artificial Intelligence research and development. The inference process in AI is the fundamental mechanism for combining information, and a significant aspect of most AT systems is the means by which they manage their overall workload by focusing processing attention and controlling which inferences are drawn and when it is appropriate to draw them. Several perspectives on the control of inferential processes and their access to information have evolved. One view of the problem treats the task as a goal-driven perceptual process, where specific information is explicitly sought from the world through selected sensor modalities, translated into a common "vocabulary," fused with other relevant information, and finally translated back into an understanding of critical aspects of the environment. Another view, centers on a flexible structure known as the blackboard architecture for enforcing control and communication activities. In this paper, we first review briefly a variety of AI inference techniques, focusing primarily on logical inference and uncertain reasoning methods. We conclude with a survey of approaches used to control inference processes, to mediate their access to real world information, and to schedule their activities.

This paper addresses methods for high and low level multi-sensor integration based on maintaining consistent labelings of features detected in different sensor domains. Implementation in a concurrent computing environment is discussed. Keywords: Multi-Sensor Integration, Sensor Fusion, Consistent Labeling, Markov Random Field, Concurrent Computing, Hypercube, Simulated Annealing.

The most promising technology for staring infrared focal plane arrays today is based on Schottky barrier detectors. The focal planes are fabricated using platinum silicide (PtSi) photosensors on a p-type silicon substrate. The wide availability of high quality, VLSI grade substrates makes arrays extremely uniform in photoresponse and relatively easy to fabricate. The current state of the art in staring focal planes is at more than 1/4 million detectors on a single chip in a (512x512) format'. Other arrays such as 39,0002 detectors (244x160) are two year old technology and 64,0003 detectors (256x256) have been developed in the last year. PtSi diodes for use as photodetectors have been in development for several years. However, there is a very large base of information available since these same diodes can also be used as high reliability interconnects for VLSI integrated circuits. Work on PtSi for interconnect diode structures has been ongoing since the early 1960's.

The process of internal photoemission is analyzed from both a theoretical and an experimental point of view. A recent model of the internal photoemission process is compared with measured data. The comparison lends credence to the model and yields results which can be used to optimize the performance of Schottky barrier infrared detectors.

Hughes Aircraft Company has successfully developed and demonstrated MWIR staring focal plane array (FPA) technology using Schottky barrier detectors with arrays consisting of 30 micron pixel spacings in 256 x 256 array format. The hybrid Schottky barrier FPA is a key technology which will permit implementation of a simple, low-cost, single field-of-view IR sensor design. The use of Schottky barrier detectors in FPAs is an outgrowth of earlier work conducted by RADC. A 128 x 128 hybrid Schottky barrier FPA was demonstrated in September 1984, with good detection and recognition performance and image quality against tactical targets. The Schottky barrier 256 x 256 MWIR hybrid focal plane array is a result of an on-going developmental process which has evolved from a 62 x 58 FPA, through a 128 x 128 FPA. Evolution of these arrays has included both improvements in the detector arrays as well as the readout or signal processing structure. The readout has been redesigned to reduce the number of clocks and biases necessary for operation. Reported is the requirement, design, fabrication, and test results of this high density hybrid FPA based upon platinum silicide infrared detector technology. The hybrid approach has advantages of ease of fabrication, high optical fill factor, compatibility with existing multiplexer technology, and excellent imaging performance. We review past Schottky FPA development at Hughes Aircraft and discuss the technical trade-offs of our approach. Discussed is the design, fabrication, and test results of our most recent Schottky FPA.

Monolithic infrared imagers with Schottky-barrier detectors (SBDs) can be constructed with silicon VLSI technology into large and high density focal plane arrays (FPAs). This paper presents a brief review of the available Pd2Si and PtSi SBDs, of the reported SBD FPAs, and of the general consideration for the construction of high fill-factor and high-density monolithic FPAs. The four FPA multiplexer architecture reviewed are: an interline transfer CCD, a charge sweep device, a column-readout MOS, and a row-readout MOS.

Schottky barrier diode charge coupled device infrared focal plane arrays exist with thousands of pixel elements. Fabrication of these large 2-D imagers is made possible by the simple design of the unit cell which can be repeated using modern CAD equipment as well as the maturity of the -2μm silicon process. Large 2-D imagers of the kind will invariably have different characteristics than their 1-D (-500 pixel element) scanning counterparts. The quantity of pixels in the Schottky array makes their characterization on a one by one basis an intractable problem. Instead, statistics can be used to quantify the array performance on a global scale. The use of statistics simplifies the extraction of uniquely two dimensional imager characteristics and system parameters.

A noise model for detectors operated in the capacitive discharge mode is presented. It is used to analyse the noise performance of the ESO nested timing readout technique applied to a linear 32 element InSb array which is multiplexed by a silicon switched-FET shift register. Analysis shows that KTC noise of the videoline is the major noise contribution. It can be eliminated by weighted double correlated sampling. Best noise performance of this array is achieved at the smallest possible reverse bias voltage (< 20mV) whereas excess noise is observed at higher reverse bias voltages.

Recent advances in focal plane array technology have made an immense impact on infrared astronomy. Results from the commissioning of the first infrared camera on UKIRT (the world's largest IR telescope) are presented. The camera, called IRCAM 1, employs the 62x58 InSb DRO array from SBRC in an otherwise general purpose system which is briefly described. Several imaging modes are possible including staring, chopping and a high-speed snapshot mode. Results to be presented include the first true high resolution images at IR wavelengths of the entire Orion nebula.

This paper describes aspects of the design and test results of the AVIRIS focal plane assemblies. These focal plane components consist of InSb and Si arrays multiplexed by silicon analog shift registers.

To fulfill the goal of high performance advanced staring infrared (IR) sensors,. development of low-noise focal plane arrays is critically important. Progress on focal plane array (FPA) development over the past decade has produced the capability to demonstrate midwave IR arrays of moderate size which may satisfy the requirements of some scanning and staring applications. In the vast majority of cases, electronic noise levels, array uniformity and dynamic range are major limiting factors with regard to the ultimate usefulness of the FPA. The purpose of this paper is to describe selected methods for measuring and characterizing spatial and temporal noise in staring FPAs and then to show how these results can be used in simulations and analytic models to predict the performance of selected staring sensors. The emphasis of the paper is on mide-wave IR FPAs for use in the detection and tracking of point sources. Measurement techniques are described generically, including necessary equipment and associated computations. Actual MWIR data from our own measurement program, or industry programs which we are monitoring are used to illustrate the current state-of-the-art. We discuss how the results of spatial and temporal noise measurements can be incorporated into simulations of sensors having staring FPAs. We show how methods for predicting performance of selected staring sensor systems are derived using representative spatial and temporal noise.