A new generation of target recognition systems must be based on the principles of image understanding and active vision. The implementation of both principles is possible in the form of Network-Symbolic systems. Instead of precise computations of 3-dimensional models a Network-Symbolic system converts image information into an “understandable” Network-Symbolic format, which is similar to relational knowledge models. The traditional linear bottom-up “segmentation-grouping-learning-recognition” approach cannot provide a reliable separation of a target from its background/clutter, while human vision unambiguously solves this problem. The nature of informational processes in the visual system does not allow separating from the informational processes in the top-level knowledge system. An Image/Video Analysis that is based on Network-Symbolic approach is a combination of recursive hierarchical bottom-up and top-down processes. Logic of visual scenes can be captured in the Network-Symbolic models and used for the reliable disambiguation of visual information, including target detection and identification. View-based object recognition is a hard problem for traditional algorithms that directly match a primary view of an object to a model. In Network-Symbolic Models, the derived structure and not the primary view is a subject for recognition. Such recognition is not affected by local changes and appearances of the object from a set of similar views.

When we digitize data from a hyperspectral imager, we do so in three dimensions; the radiometric dimension, the spectral dimension, and the spatial dimension(s). The output can be regarded as a random variable taking values from a discrete alphabet, thus allowing simple estimation of the variable's entropy, i.e., its information content. By modeling the target/background state as a binary random variable and the corresponding measured spectra as a function thereof, we can compute the information capacity of a certain sensor or sensor configuration. This can be used as a measure of the separability of the two classes, and also gives a bound on the sensor's performance. Changing the parameters of the digitizing process, bascially how many bits and bands to spend, will affect the information capacity, and we can thus try to find parameters where as few bits/bands as possible gives us as good class separability as possible. The parameters to be optimized in this way (and with respect to the chosen target and background) are spatial, radiometric and spectral resolution, i.e., which spectral bands to use and how to quantize them. In this paper, we focus on the band selection problem, describe an initial approach, and show early results of target/background separation.

Thermal infrared target detection and tracking has challenging and useful applications outside of military scenarios. A digital image processing technique is described for the detection and tracking of free flying bats. Uncalibrated video-rate thermal imagery from a stationary FPA micro-bolometric IR imager is captured on 8-bit digital media. Sequential frames are differenced to remove stationary clutter, and thresholded to select pixels outside of the central distribution of differenced pixel values (both positive and negative). Moving objects then appear as pairs of pixel clusters of differing contrast polarity. For the typical case of a warm bat against a cool background, a pixel cluster exceeding the positive threshold indicates a target location in the current frame and corresponding pixel cluster below the negative threshold indicates the target’s location in the previous frame. These location pairs define a motion vector that is updated every frame. Using the updated motion vector, the next position of the bat is predicted. If a similar-sized pixel cluster of the correct polarity is found at this predicted location, within a selectable error tolerance, then a track is established. This process is iterated frame-by-frame generating an output file of individual bat tracks. This process is described in detail and data are presented from an imaging survey of a bat emergence containing several thousand bats.

Polarization adds another dimension to the spatial intensity and spectral information typically acquired in remote sensing. Polarization imparted by surface reflections contains unique and discriminatory signatures which may augment spectral target-detection techniques. While efforts have been made toward quantifying the polarimetric bidirectional reflectance distribution function (pBRDF) responsible for target material polarimetric signatures, little has been done toward developing a description of the polarized background or scene clutter. An approach is presented for measuring the pBRDF of background materials such as vegetation. The governing equation for polarized radiance reaching a sensor aperture is first developed and serves as a basis for understanding outdoor pBRDF measurements, as well as polarimetric remote sensing. The pBRDF measurements are acquired through an imaging technique which enables derivation of the BRDF variability as a function of the ground separation distance (GSD). An image subtraction technique is used to minimize measurement errors resulting from the partially polarized downwelled sky radiance. Quantifying the GSD-dependent BRDF variability is critical for background materials which are typically spatially inhomogeneous. Preliminary results from employing the measurement technique are presented.

We report experimental and computational results showing spatially dependent intensity patterns for polarized light, diffusely backscattered from highly scattering media. It is demonstrated that obtained two-dimensional polarization patterns can be used to differentiate between various scatterer sizes even if the reduced scattered coefficient for different media is the same. Our experimental technique uses Mueller matrix imaging polarimeter with a sensitive camera providing the signal detection at 12 bit resolution. The computational approach is based on Mie theory, Stokes vector/Mueller matrix formalism and Monte Carlo simulation for propagation of polarized light in scattering media.

A midwave hand-held two-band infrared imager has been developed to measure the emissivity image of aircraft surfaces. Its purpose is to detect changes in IR surface properties of aircraft coatings. This system has been designed for use in field environments such as an aircraft carrier hangaer and other maintenance facilities where the object being measured is static and cooperative but the environment is not well controlled. It thus requires real time monitoring of the environmental reflections off the surface and algorithms to correct for this reflected radiance. This correction for the environment using calibrated paint patchescoupons is a novel feature of this new technology. The camera output includes two band radiance images, temperature images, and the novel emissivity imagery for which eCAM has been named. It has long been known that two-band infrared measurements can be used to optically determine temperature and solid angle-emissivity products of a greybody surface. This measurement becomes considerably more complex when the environmental foreground reflecting off the surface is a significant part of the infrared-based measurement. This paper will describes the theory behind making foreground- reflection- corrected emissivity image measurements. It will includes material on the calibration of the IR sensor and its specially designed optics and the custom emissivity calibrated patches that are an integral part of the design. It will also shows laboratory test results and field test data taken in an aircraft maintenance hangar.

Fundamental to the performance of an active infrared imager, is the Bi-directional Reflectance Distribution Function (BRDF) from potential target and surveillance surfaces. Proper integration of the BRDF also can yield emissivity, a quantity that is important to determine emitted infrared radiance for passive sensors, especially in sunlit conditions in the SWIR and MWIR. A series of BRDF measurements were acquired in up to four bandpasses ranging from 3 to over 11 microns. Results are presented for multiple surfaces in these bandpasses with the incidence angles of the radiation at -45 degrees in elevation, and azimuth angles from 0 to 180. Materials measured included likely target and background materials such as: human skin, clothing materials, wood, paper, water, sand and crude oil.

For UWB (30 Hz - 100 kHz) electromagnetic induction (EMI) sensor discrimination of unexploded ordnance (UXO), we evaluate first the effects of significant magnetic permeability in the surrounding soil. Measured data and theoretical arguments suggest that ground effects can often be accounted for by using a simple halfspace analytical solution. Thus, when target responses are strong enough, free-space target signature shapes can still be used for discrimination if properly compensated. At the same time, even in artificially well-mixed, physically smoothed settings, local variations in soil permeability can be a significant source of signal clutter. Cases with multiple UXO’s beneath dispersed small metallic clutter are also considered as instances in which clutter may dominate. In simulations of two comparably sized UXO’s at comparable depths with a signal to clutter ratio (SCR) of ~ 20, UWB data distinguishes the two objects reliably over a ground surface measurement grid. For similar cases but with the objects at significantly different depths relative to one another, one cannot distinguish the deeper target, even with the same noise level and with UWB data. Measurements illustrate the level of EMI SCR to be expected from dispersed small metallic items collected from a firing range. For cases with a single piece of clutter and a much more massive UXO immediately below, simulations show almost complete obscuration of the UXO, in both frequency and time domains. This is not caused by signal blockage but results from different degrees of proximity to the sensor, i.e. from the consequent signal magnitude disparity.

Infrared emissions from the detonation of three bomb types and four weights in a series of 56 events were recorded by a Fourier transform spectrometer in the midwave IR (1800-6000 wavenumbers) at temporal and spectral resolutions of 0.047 seconds and 16 wavenumbers, respectively. Fifteen time-resolved spectral datasets corresponding to two distinct chemical explosives were selected for this study. The detonation fireball intensities are well described as cooling greybodies, and a single Planckian distribution, modified by atmospheric absorption, has been fit to the spectra. Agreement between the model and data is within a few percent on average. However, the model underestimates the observed intensity by as much as 40% in the 2000-2200 wavenumber window and hot carbon dioxide at the surface of the fireball is a likely source of this spectral emission (spectral assignments have not yet been performed). For the statically detonated munitions, temperature curves are characterized by initial temperatures of 1685-1885 Kelvin and lifetimes of 0.91-1.24 seconds. Temperatures for some air delivered ordnance exhibited secondary maxima. Fireball areas are estimated without imagery. The model provides features which are reproducible within and characteristic of the munition type, providing promise for proposed event classification schemes. The timedependent Planckian fit residual near 2150 wavenumbers versus time provided the best discrimination between the two munition types, indicating that better understanding the non-Planckian behavior is key to the classification problem. A novel method to estimate the atmospheric transmittance function from the time-resolved fireball spectra is also developed.

Highly-energetic targets such as rocket plumes and detonation fireballs are difficult to locate and track effectively with active sensors that rely on reflection, but traditional passive sensors cannot determine range. Development of a passive ranging sensor will enable accurate target location and tracking while simultaneously improving covertness. Recent work is presented on development of a passive sensor to estimate range using spectroscopic measurements of atmospheric absorption. In particular, advantages of measuring absorption on the O 2(bX) transition near 762nm are discussed. Theoretical predictions are compared with experimental results to verify model performance at short ranges (up to 200m). Range accuracy better than 1% has been demonstrated using a Fourier transform spectrometer at short range. Extension of the theory to long range (in excess of 100km) is also discussed.

This research investigates the classification of battlespace detonations, specifically the determination of munitions type and size using temporal and spectral features from near-infrared (NIR) and visible wavelength imagers. Key features from the time dependence of fireball size are identified for discriminating various types and sizes of detonation flashes. The five classes include three weights of trinitrotoluene (TNT) and two weights of an enhanced mixture, all of which are uncased and detonated with 10% C-4. Using Fisher linear discriminant techniques, these features are projected onto a line such that the projected points are maximally clustered for the different classes of detonations. Bayesian decision boundaries for classification are then established on class-conditional probability densities and are tested using independent test data. Feature saliency and stability are determined by selecting those features that best discriminate while requiring low variations in class-conditional probability densities and high performance in independent testing. Given similar conditions, the most important and stable feature is the time to the maximum fireball area in the near-infrared wavelength band (0.6 to 1.7 microns). This feature correctly discriminates between TNT and ENE about 90% of the time, whether weight is known or not. The associated class-conditional probability densities separate the two classes with a Fisher ratio of 2.9±0.3 and an area under the receiver operating characteristic, A_ROC, of 0.992. Also, t_mp achieves approximately 54% success rate at discerning both weight and type.

Technological progress in the fields of computing hardware and efficient algorithms make it possible to set up real-time exploitation systems for a huge number of applications (e.g. assessment of camouflage effectiveness, or various surveil-lance applications, UAVs, as well as image sequence data reduction, indexing, archiving, and retrieval). The system in question has been developed to cope with highly dynamic situations. Such dynamic situations may be characterized by moving targets acquired by a static, trembling, or moving sensor system. The image sequences may stem from a visual-optical (VIS) or some forward looking infrared (FLIR) sensor. Except for wide-angle lenses (due to their optical distortions) neither sensor nor calibration parameters have to be known to the automatic exploitation system. Furthermore no human interaction is required. The algorithmic approach tries to digitally stabilize the movement of the sensor system. To accomplish this task the algorithm extracts 40-60 tie points from the static nonmoving background, then robustly matches the tie point constellations frame to frame for calculating the 8 parameters of a projective mapping. This is the basis for some sort of background stabilization. The difference image of two consecutive and matched image frames re-veals the moving targets. After the segmentation of the (moving) target signatures, additionally attached tracking and classification components have been tested.

Accurate information on the distribution of sensible and latent heat fluxes is critical for evaluation of background characteristics. Since these fluxes are subject to rapid changes in time and space, it is nearly impossible to determine their spatial and temporal distributions over large areas from ground measurements alone. Therefore, prediction from remote sensing data is very attractive as it enables large area coverage and a high repetition rate. In this study, the Surface Energy Balance Algorithm for Land (SEBAL) is selected to estimate sensible and latent heat fluxes at 30 m resolution in the riparian areas of the Middle Rio Grande Basin (New Mexico), San Pedro River (Arizona) and Owens Valley (California). The objective is to compare SEBAL fluxes derived from LandSat TM images with those measured on the ground with eddy covariance towers. The comparison in arid heterogeneous riparian areas in the southwestern United States clearly demonstrates that SEBAL can be applied for mapping sensible and latent heat fluxes at high spatial resolutions.

Remotely sensed images of the Earth’s surface provide information about the spatial distribution of evapotranspiration. Since the spatial resolution of evapotranspiration predictions depends on the sensor type; scaling transfer between images of different scales needs to be investigated. The objectives of this study are first to validate the consistency of SEBAL algorithms for satellite images of different scales and second to investigate the effect of up- and down-scaling procedures between evapotranspiration maps derived from LandSat 7 and MODIS images. The results of this study demonstrate: (1) good agreement of SEBAL evapotranspiration estimates between LandSat 7 and MODIS images; (2) up- and down-scaled evapotranspiration maps over the Middle Rio Grande Basin are very similar to evapotranspiration maps directly derived from LandSat 7 and MODIS images.

Root zone soil moisture is a dynamic variable subject to rapid changes in time as well as space. Accurate, detailed information on the distribution of soil moisture is difficult to obtain. Ground based methods for the measurement of temporal and spatial changes in root zone soil moisture require much time and effort and, therefore, have limited value for soil moisture monitoring at the regional scale. Existing remote sensing methods use microwaves to measure soil moisture near the soil surface (0-10 cm). In this study, the Surface Energy Balance Algorithm for Land (SEBAL) is applied to a series of LandSat TM optical images to determine the regional distribution of the evaporative fraction. From this, soil moisture conditions are derived using an empirical relationship between evaporative fraction and root zone soil moisture. Ground measurements available in the Middle Rio Grande Basin do not cover a sufficiently wide range of soil moisture values for validation of the soil moisture maps derived from optical satellite imagery. Therefore, we conducted a qualitative validation by comparing predicted soil moisture conditions along three transects perpendicular to the Rio Grande with ground observations. This analysis shows that the remote sensing approach is successful in distinguishing between moist and dry pixels in the Middle Rio Grande Basin. Further field investigation will be required to validate the product quantitatively.

CHORALE (simulated Optronic Acoustic Radar battlefield) is used by the French DGA/DET (Directorate for Evaluation of the French Ministry of Defense) to perform multi-sensors simulations. CHORALE enables the user to create virtual and realistic multi spectral 3D scenes, and generate the physical signal received by a sensor, typically an IR sensor. Some assessments concern the study of the duality between a threat (a missile for example) and a target (a battle tank for example) in the battlefield. In these cases, obscurants are special counter measures (clouds), classically used to hide armored vehicles and/or to deceive threatens. To evaluate their efficiency in visible and infrared wavelength, simulations tools, that give a good representation of physical phenomena, are used. The first part of this article describes the elements used to prepare data for the simulation. The second part explains the physical model used in CHORALE for the resolution of the Radiative Transfer Equation when obscurants are set in the scene. Obscurants are modeled by a set of voxels (elementary volume elements). Each voxel contains the spectral absorption and scattering coefficients, phase function coefficient and temperature information. The shape is changing with time to take into account the dynamic evolution of the obscurant. A “photon map” method is used in the ray tracing process to take into account global illumination within the cloud and solve the Radiative Transfer Equation.

The increased interest during the last decade in the infrared signature of (new) ships results in a clear need of validated infrared signature prediction codes. This paper presents the results of comparing an in-house developed signature prediction code with measurements made in the 3 5 μm band in both clear-sky and overcast conditions. During the measurements, sensors measured the short-wave and long-wave irradiation from sun and sky, which forms a significant part of the heat flux exchange between ship and environment, but is linked weakly to the standard meteorological data measured routinely (e.g., air temperature, relative humidity, wind speed, pressure, cloud cover). The aim of the signature model validation is checking the heat flux balance algorithm in the model and the representation of the target. Any uncertainties in the prediction of the radiative properties of the environment (which are usually computed with a code like MODTRAN) must be minimised. It is shown that for the validation of signature prediction models the standard meteorological data are insufficient for the computation of sky radiance and solar irradiation with atmospheric radiation models (MODTRAN). Comparisons between model predictions and data are shown for predictions computed with and without global irradiation data. The results underline the necessity of measuring the irradiation (from sun, sky, sea or land environment) on the target during a signature measurement trial. Only then does the trial produce the data needed as a reference for the computation of the infrared signature of the ship in conditions other than those during the trial.

The last decade has seen an increase in the awareness of the infrared signature of naval ships. New ship designs show that infrared signature reduction measures are being incorporated, such as exhaust gas cooling systems, relocation of the exhausts and surface cooling systems. Hull and superstructure are cooled with dedicated spray systems, in addition to special paint systems that are being developed for optimum stealth. This paper presents a method to develop requirements for the emissivity of a ship's coating that reduces the contrast of the ship against its background in the wavelength band or bands of threat sensors. As this contrast strongly depends on the atmospheric environment, these requirements must follow from a detailed analysis of the infrared signature of the ship in its expected areas of operation. Weather statistics for a large number of areas have been collected to produce a series of 'standard environments'. These environments have been used to demonstrate the method of specifying coating emissivity requirements. Results are presented to show that the optimised coatings reduce the temperature contrast. The use of the standard environments yields a complete, yet concise, description of the signature of the ship over its areas of operation. The signature results illustrate the strong dependence of the infrared signature on the atmospheric environment and can be used to identify those conditions where signature reduction is most effective in reducing the ship's susceptibility to detection by IR sensors.

The Irma synthetic signature prediction code is being developed to facilitate the research and development of multisensor systems. Irma was one of the first high resolution Infrared (IR) target and background signature models to be developed for tactical weapon application. Originally developed in 1980 by the Munitions Directorate of the Air Force Research Laboratory (AFRL/MN), the Irma model was used exclusively to generate IR scenes. In 1988, a number of significant upgrades to Irma were initiated including the addition of a laser (or active) channel. This two-channel version was released to the user community in 1990. In 1992, an improved scene generator was incorporated into the Irma model, which supported correlated frame-to-frame imagery. A passive IR/millimeter wave (MMW) code was completed in 1994. This served as the cornerstone for the development of the co-registered active/passive IR/MMW model, Irma 4.0. In 2000, Irma version 5.0 was released which encompassed several upgrades to both the physical models and software. Circular polarization was added to the passive channel and the doppler capability was added to the active MMW channel. In 2002, the multibounce technique was added to the Irma passive channel. In the ladar channel, a user-friendly Ladar Sensor Assistant (LSA) was incorporated which provides capability and flexibility for sensor modeling. Irma 5.0 runs on several platforms including Windows, Linux, Solaris, and SGI Irix. Since 2000, additional capabilities and enhancements have been added to the ladar channel including polarization and speckle effect. Work is still ongoing to add time-jittering model to the ladar channel. A new user interface has been introduced to aid users in the mechanism of scene generation and running the Irma code. The user interface provides a canvas where a user can add and remove objects using mouse clicks to construct a scene. The scene can then be visualized to find the desired sensor position. The synthetic ladar signatures have been validated twice and underwent a third validation test near the end of 04. These capabilities will be integrated into the next release, Irma 5.1, scheduled for completion in the summer of FY05. Irma is currently being used to support a number of civilian and military applications. The Irma user base includes over 130 agencies within the Air Force, Army, Navy, DARPA, NASA, Department of Transportation, academia, and industry. The purpose of this paper is to report the progress of the Irma 5.1 development effort.

When using prediction programs for optical signatures, it is necessary to include validations to find estimates of the uncertainties and define the regions of validity. In this paper we present two paths of development of validation methods: The objective of the first path is to analyze and validate the differences between simulated and measured images, through image features such as edge concentration and different energy measures. In particular, aspects that are important for detection, classification and identification of targets are considered. The second path concerns development of methods for quantifying the propagation of input data uncertainties to output parameters in computational predictions. Two commercial codes have been used for the modeling: RadThermIR for thermal predictions of the targets and CAMEO-SIM for the radiometry and rendering. A recently developed interface between the two codes has been utilized. For the validation of spatial statistics, several feature values have been computed for a measured image and for the corresponding simulated image. It was found that the agreement was quite good. The work on propagation of uncertainties in computational predictions has resulted in a number of proposed methods. In this paper we present two different methods: one based on linear error propagation and one based on the Monte Carlo method. The results are according to expectations for both types of methods and show that a large part of the uncertainty in predicted temperature emanates from input parameter uncertainties for the considered test case.

Vehicles concealed in highly-cluttered, vegetated scene environments pose significant challenges for passive sensor systems and algorithms. System analysts working hypersectral exploitation research require and at-aperature simulation capability that allows them to reliably investigate beyond ther highly-limited scenarios that expensive field data sets provide. To be useful to the analyst, such a simulation should address the following requirements: (1) the ability to easily generate scene representations for abritrary Earth regions of tactical interests; (2) the ability to represent scene components, like terrain, trees and bushes, to an extremely high spatial resolution for calculation of accurate multiple spectral reflections, occlusions and shadowing; (3) the ability to stimulate the 3D scene with realistic natural irradiances for arbitrary model atmospheres; (4) the ability to appropriately integrate improving, rigorous thermal, spectral signature and atmospheric propogation models; (5) the ability to effectively render at-apurature hyperspectral data sets in a reasonable run-time. herein the authors describe their continuing work toward a comprehensive ray-tracer-based simulation archetecture and prototype capability that addresses these requirements, with emphasis on new techniques for high fidelity thermal modeling, and recent improvements in atmospherically scattered irradiance modeling, manmade light source modeling, and GIS-based database generation, including automated material classification of terrain and scene elements.

This paper addresses advances in techniques used to produce credible Infrared (IR) optical target images and point source intensities for effective simulation testing and validation. Integral to a credible simulation process is the ability to accurately generate and inject synthetic imagery into various simulation topologies for model verification, validation and accreditation. This research exploits improvements in computational power and refines the modeling algorithm in response to the demands for significantly increased detectivity requirements in target discrimination. The software architecture is Commercial-Off-The-Shelf (COTS)-based and results from judicious implementation of performance sensitivity analysis indicators of how the threat and background signatures vary as a function of changes in factor values. A novel approach to determine the minimal image set for a trajectory is presented, using the Intensity Variation Threshold (IVT). Several illustrative examples are presented to show how inherent limitations in the COTS software have been effectively mitigated for a cost-effective solution. This approach is well suited to provide enhanced target imagery and improved flexibility and control over threat geometry and thermal effects within an end-to-end simulation. A discussion of recent advances in modeling target and background synthetic imagery modeling is of increased interest in the military community because of the use of simulation for validation of weapon system performance.

The statistic results for a digital terrain model are presented that closely match measurements for 77% of the 189 possible combinations of 7 radar bands, 3 polarizations, and 9 terrain types. The model produces realistic backscatter coefficient values for the scenarios over all incidence angles from normal to grazing. The generator was created using measured data sets reported in the Handbook of Radar Scattering Statistics for Terrain covering L, C, S, X, Ka, Ku, and W frequency bands; HH, HV, and VV polarizations; and soil and rock, shrub, tree, short vegetation, grass, dry snow, wet snow, road surface, and urban area terrain types. The first two statistical moments match published values precisely, and a Chi-Square histogram test failed to reject the generator at a 95% confidence level for the 146 terrain models implemented. A Sea State model provides the grazing angle extension for predictions beyond the available measurements. This work will contain a comprehensive set of plots of mean and standard deviation versus incidence angle.

Anybody can mount a zooming lens on his camera and observe distant objects. FarAway penetrates the blanket of haze that obscures such objects, regardless of the origin - mist, smoke, dust, rain, aerosols, etc. The system works with live Video, 25 frames/second, performing the restoration in real time on a PC hardware. No pre-knowledge about the obscurants, targets or distances is required. The novelty lies in performance; the system does not use image-processing tricks or contrast-stretching, but rather restores the original true image with its true colors. The current version is capable of relative-contrast enhancement of up to 90 times in color and up to 130 in B/W or NIR, the limit dictated by electrical noise. Facing the sun, this means a three-fold and more increase in detection and recognition ranges. Both color and NIR systems were extensively and successfully tested under rain, mists, haze and dust storms at ranges from 0.2 to 65 km. Turbulence effects are treated crudely, reducing apparent turbulence dance and smear by a factor of 2 to 4. Tests have proved its superiority over existing top-rank military systems. A related technology is Very Far Away. While Far Away uses commercial as-is cameras, the Very Far Away uses a specially designed color camera. The special camera allows extension of the visibility range by a further 70%.