EMI sensors are used extensively to detect landmines. These sensors operate by detecting the metal that is present in mines. However, mines vary in their construction form metal- cased varieties with a large mass of metal to plastic-cased varieties with minute amounts of metal. Unfortunately, there is often a significant amount of metallic clutter present in the environment. Consequently, EMI sensors that utilize traditional detection algorithms based solely on metal content suffer from large false alarm rates. This false alarm problem can be mitigated for high-metal content mines by developing statistical algorithms that exploit models of the underlying physics. In such models it is commonly assumed that the soil has a negligible effect on the sensor response. To date, no testing has been done to validate the assumption that when modeling the response of EMI sensors to low-metal mines, the solid effects are negligible. In addition, advanced algorithms have not been applied specifically to the detection of low-metal mines. The Joint UXO Coordination Office (JUXOCO) is sponsoring a series of experiments designed to establish a performance baseline for EMI sensor. This baseline will be used to measure the potential improvements in performance offered by advanced signal processing algorithms. This paper describes the results of several experiments performed in conjunction with the JUXOCO effort. The results indicate that (1) the properties of the soil do affect the response of a broadband EMI sensor to low-metal mines, and (2) advanced algorithms can improve detection performance over traditional algorithms based solely on metal content.

The phenomenology of frequency- and time-domain electromagnetic induction (EMI) is examined in detail, through use of a rigorous electromagnetic scattering model, and through appropriate signal analysis. We demonstrate that both the time- and frequency-domain EMI signatures can be characterized in terms of a few magnetic singularities, thereby significantly reducing the number of features that need be stored for target identification. Further, we examine the aspect-dependent properties of the relative extinction strengths of the magnetic-singularity modes. Finally, we perform a statistical analysis of the relative efficacy of frequency- and time-domain EMI operation, for a class of conducting targets.

The problem of mine localization and characterization form electromagnetic induction data is addressed. We consider processing techniques based on an inductive sensor model originally proposed by Das et. al. Given this model we examine estimation-theoretic methods for determining an object's center, its orientation, and scattering characteristics from low frequency spectroscopic data obtained over a grid of spatial locations. Under this model, the data are linear functions of the multiple moment spectra and non-linearity related to object's location and rotation angles. An estimation procedure based on a low-dimensional non-linear optimization routine for the determination of the object center and rotation angles is employed with a linear lest squares inversion method used to estimate the multiple moment spectra. Examples are provided for ellipsoidal objects.

The magnetic detection of objects with ferromagnetic casings, such as unexploded ordnance (UXO), is typically plagued by high false alarm rates. In the past, only estimates of the magnitudes of the dipoles found in a magnetic search have been used to determine if the dipoles are UXO-like. Following Altshuler, this paper reports on a method aimed at significantly reducing the false-alarm rate, thereby making magnetic detection of UXO more effective. The method assumes that UXO magnetic dipoles are predominantly induced, rather than permanent, and that their ferromagnetic casings can be closely approximated as spheroids. the method relies on the determination of the full magnetic moment of each magnetic dipole. This information can be obtained, even in a multi-target scenario, by magnetic field gradiometers, which measure the five independent components of the gradient of the magnetic field. Two distinct two-parameter classification schemes are described. One assumes that the magnetic object is lying flat and determines the effective magnetic susceptibilities parallel and normal to the symmetry axis. The other scheme makes no assumption about the orientation of the magnetic object, but provides simple and robust limits on moment magnitude and direction that effectively reject clutter. Real magnetic clutter data gathered at sea is employed to demonstrate and estimate the effectiveness of these schemes.

This paper describes data collection efforts carried out at Fort A.P Hill VA with the US Army's hand held metal detector (MD) the AN/PSS-12. Our efforts were directed toward establishing the receiver operating characteristics (ROC) curves for the AN/PSS-12 thereby creating an objective measure of its baseline performance capabilities. Voltage- versus-time waveforms were recorded at two different locations in the AN/PSS-12's receiver circuitry: 1) After the first stage of amplification following the receiver coil, and 2) At the output of the step-2 difference amplifier just before its voltage-to-frequency converter. The latter signal was used to determine the frequency of the detector's audio output frequency-versus-position data was used to determine the baseline ROC for the AN/PSS-12 under the binary mine-no-mine hypothesis. Under this hypothesis, the baseline ROC for the detector is shown to lie close to the change diagonal, an expected result, since the detector's audio output offers the detector operator no way to discriminate between mines and metallic clutter objects. An improved 'ad hoc' detector is presented that has the ability to distinguish between mines and clutter objects based on spatial symmetry. Our symmetry detector operates on energy-versus-position data derived from the first type of data described above. The ROC for out symmetry detector is shown to lie well above the change diagonal.

Two sets of metallic objects are created to provide a standard set of metallic test targets to facilitate an objective comparison and evaluation of metal detectors. The first set of metallic objects is chosen form combinations of small metal parts common to many low-metallic content landmines. The collections of small metal parts are chosen based on an average detection distance measured with five sensitive metal detectors. A second set of metal objects is created using short-circuited coils of wire, INSCOILS. A development of the theory describing the interactions of INSCOILS with a metal detector's transmit and receive coil shows that the coupling and response function of an INSCOIL can be independently controlled. By varying the wire gauge, wire material, and loop size, an INSCOIL can be made to approximate the response of an arbitrary metallic object. A pulse-induction measurement system is used to measure the response of different metallic objects. The pulse-induction measurement system is used to match the response of an INSCOIL to that of the collection of small metal parts. Surrogate landmines are also constructed by matching the response of a coil of wire to that of a specific landmine.

Metal land mines still account for a large percentage of land mines, even with the advent of the so-called plastic mines. The metal detector thus remains a viable tool in the mine detector's bag. The limitation of the metal detector is not in detection of the mines, but in the additional detection of metal clutter. A metal detector has been developed which can largely discriminate the mines for the clutter, thereby greatly reducing false alarm rates. This 'mine detector' is designed to characterize the magnetic polarizability dyadic of the metal objects, and to use pattern recognition to determine the goodness-of-fit to the responses of known mines. Data are presented from test runs conducted for the US Army for buried metal miens. Data are also presented for some non-mine metal targets. The characterization of the mines as threats is performed in a totally autonomous system, with high probability-of- detection and low false alarm rate. We can also generally tell one mine type from another.

Unlike radar-based imaging technologies that use electromagnetic waves, quasistatic imaging technologies operate at lower frequencies where electric and magnetic fields are decoupled. Magnetoquasistatic (MQS) devices, such as metal detectors, that impose magnetic fields satisfy the diffusion equation in conducting media and Laplace's equation in air or poorly conducting soils. Electroquasistatic (EQS) devices satisfy Laplace's equation. In Laplacian or diffusion decay, the amplitude of the magnetic and electric fields decay exponentially with distance from the drive windings or electrode. For quasistatic sensors, objects are detected and imaged through perturbations to the applied magnetic or electric fields that change the mutual transimpedances or transadmittances at the sensor terminals, rather than through time delays of reflected electromagnetic waves as in GPR.

A handhold mine detection system is under development using a 2D array of spin dependent tunneling (SDT) magnetorestrictive sensors, which will measure the x, y and z scalar components of the electromagnetic (EM) field. SDT sensors directly measure the EM field component along an axis of the sensor over a wide frequency and intensity range, which make them ideal EM sensors. The sensors are small and are relatively inexpensive due to the massive investment in this technology by the computer industry for their use in disc storage devices. A system was designed with primary emphasis on the unique capabilities of the sensor elements and sensor array design for landmine detection and discrimination. Much of the early work concentrated on theoretical models verified with measured laboratory time domain EM response of metallic components of typical low metal landmines. The modeling results have provided the information needed to define performance requirements for the SDT sensor and a design of an array of SDT sensor to measure the x, y and z spatial components expected from the landmines. A parallel effort to develop the supporting theory for optimal interpretation of the multi-axis sensor array, has resulted in significant progress in developing an improved methodology for distinguishing the signature of landmine targets from metallic clutter. We have adopted an integrated approach to the sensor design in which the data requirements for effective discrimination have driven the design while meeting the practical and engineering requirements as well.

Recent advances in the development of miniature high power radar sources are presented. The miniature radar sources convert mechanical energy into electrical energy by compressing either piezoelectric materials or magnetic flux. The results of experiments performed to reach 100 kW of converted power are presented and discussed. Estimates are also given indicating the maximum power that can be generated with such devices.

We have recently conducted a series of laboratory and field test to demonstrate the utility of combining active illumination with hyperspectral imaging for the detection of concealed targets in natural terrain. The active illuminator, developed at MIT Lincoln Laboratory, is a novel microlaser-pumped fiber Raman source that provides high- brightness, subnanosecond-pulse-length output spanning the visible through near-IR spectral range. The hyperspectral- imaging system is comprised of a compact, grating-based spectrometer that uses a gateable, intensified CCD array as the detector element. The illuminator and hyperspectral imaging system are mounted on a small platform that is itself mounted on a tripod and scanned in azimuth to build an image scene of up to several hundred spectral bands. The system has been deployed under a variety of environmental conditions, including night-time illumination, and on a variety of target scenes, including exposed and concealed plastic and metallic mine-like targets. Targets have been detected and identified on the basis of spectral reflectance, fluorescence signatures, degree of polarization, and range-to-target information. The combination of laser-like broadband illumination and hyperspectral imaging offers great promise in concealed or obscured target detection. On-going developments include the incorporation of broadband illuminators in the 1 to 2 micrometers and 3 to 5 micrometers spectral bands, with corresponding increases in spectral coverage of the imaging and detection systems.

Buried landmine detection using IR sensor has been investigated by a number of researchers around the world. The technique is promising, particularly for unpaved roads in hot dry countries with little roadside shading. The thermal source for this detection mechanism is the solar heat flux and this imposes constraints on the time of day and weather conditions when this technique is effective. Here we investigate the use of a microwave energy source to drive the thermal flux with the aim of extending the usefulness of IR detection to full day operation in all forms of overcast weather. The microwave radiation penetrates the soil providing a depth to the heating process. We present a 1D model of microwave absorption and heat dissipation by moisture-laden soils which contain landmine-like buried objects. The microwave source for our numerical experiments is a 2.45 GHz, 5 kW. If required, a possible power increase to as high as 50 kW is envisaged. The physical mechanisms employed are suitable for a vehicle mounted detection system. We consider thermal properties of plastic bodied landmines shallowly buried in soil at the peak power point in the ground footprint of the heat source and review some of the factors which govern the thermal difference which will result. In particular we review (a) the factors which determine the radiant power penetrating the ground, and (b) the absorption and conduction processes which subsequently occur within the vicinity of the landmine.

This paper presents findings based on IR thermography in the spectral range of 3-5 micrometers of mine-like objects buried in gravel, where the surface is actively thermally stimulated. Enhanced detectability of buried objects, specifically land mines, is demonstrated to be possible. Priority is given to buried plastic AP mines. Results show that it seems possible to actively provoke an IR signature of near surface mine- like objects and that a time-domain method will be able to distinguish mines from other types of buried objects. Actively provoked IR signatures combined with a time-domain treatment method appear to have potential for further improvement of the localization of buried mines.

Microwave heating of the ground can enhance IR signatures of buried objects. The extent of the enhancement depends upon many parameters including wavelength, polarization, angle of incidence, and properties of the solid and the object. We show that angle of incidence and dielectric properties of the object are important, with some analytical and experimental results.

The shape and thermal properties of buried objects can result in a variation in the temperature profile on the surface of the ground. IR imaging has been used to exploit this variation to detect the presence of buried objects. The thermal signature in such cases is normally induced by natural means such as diurnal cycles. This method requires observation at specific times of day and has not in general allowed reliable detection and discrimination, especially for small antipersonnel mines. We have developed a process that uses an array of heated waterjets to rapidly induce a thermal signature of buried objects in the region of interest. The high-pressure, small diameter waterjets penetrate the soil but are deflected by a formed buried objects. A temperature profile on the ground surface is formed due to the radiation and conduction of heat from the water blocked and reflected by the surface of the buried object and the heating of the object itself due to heat transferred from the object to a blurred 2D IR image of the surface. Deblurring and other physics-based image processing techniques are used to correct for the heat diffusion and an estimate can be made of the 3D shape of the part of the buried object which is covered by the waterjet. A time history of the thermal profile is also available when several IR images are acquired after the waterjets are applied. This allows further analysis of the nature of the properties of the buried objects. Known properties of land mines can be used to discriminate them from other buried objects. Shape feature properties based on Fourier descriptors have been developed to allow discrimination of objects.

A hybrid technique is presented that simultaneously uses both electromagnetic and acoustic waves in a synergistic manner to detect buried land mines. The system consists of an electromagnetic radar and an acoustic source. The acoustic source causes both the mine and the surface of the earth to be displaced. The electromagnetic radar is used to detect these displacements and, thus, the mine. To demonstrate the viability of this technique, experimental models have been constructed. The models use an electrodynamic transducer to generate an acoustic surface wave, a tank filled with damp sand to simulate the earth, simulated mines, and a radar to measure the vibrations. The technique looks promising; we have been able to measure the interactions of the acoustic waves with both simulated antipersonnel mines and antitank miens buried in damp sand. We have measured strong resonance in some of the mines; these resonances are shown to help differentiate the mine from clutter.

Airborne acoustic waves coupled into the surface of the ground excite Biot Type I and II compressional and shear waves. This coupling of airborne sound into the ground is termed acoustic-to-seismic coupling. If a land mine or other inhomogeneity is presented below the surface, the ground surface vibrational velocity or S/A ratio will increase due to reflection and scattering of the Type II compressional wave. The dispersion characteristics of this wave in solids determines the mine detection limits. The S/A ratio is read with a laser doppler vibrometer (LDV). The loud speaker and LDV were mounted onto a large forklift at Fort AP Hill. This system was used to scan patches of ground at the Fort AP Hill calibration mine lanes. An investigation on the variability of surface velocity over different background types and mine types is described. The results of these initial field exercises are described.

A truncated ellipsoidal dish is used to translate a point source of acoustic energy, applied at the upper focal point, to a virtual source position at the lower focal pint, located within the truncated zone of the ellipsoid. The position of the lower focal point is, by definition, search depth. When this depth coincides with the depth of burial of a land mine, the height of the virtual source above the surface of the mine is, effectively, zero. Thus, questions relative to reflection efficiency that involve considerations of source wavelength and target size fall mute, since the Fresnel radius of the reflector essentially vanishes. Consequently, low frequencies can be used to detect small targets. Test results in sand, using plastic surrogate mines as targets, confirm realization of concept. The prototype enabled estimation of dish size and source characteristics required for a full-scale field system. The problem of anomalous nose associated with in-homogeneity of near-surface materials, undulations of the air-earth interface, and presence of surface objects was uncovered. 'False anomalies', as a result of these surface conditions, totally mask 'target anomalies'. For a field system to be viable, target discrimination, and/or noise suppression processing algorithms must be developed.

We report here on the use of acoustic pulses generated by a pulsed-laser incident on the ground surface for the depth- and shape-resolution of buried objects. The laser-induced acoustic wave has considerable advantages over other acoustic wave generation techniques for landmine detection applications. (1) It is efficient because the sound is generated directly in the ground. (2) The acoustic source can be precisely positioned or scanned by optical redirection of the laser spot. (3) Remote, non-ground- contact detection can be accomplished with a receiving microphone in the air or by using optical vibrometry of the soil surface for detection. Research has been focused on the data acquisition and signal processing applicable to de- mining scenarios. A de-convolution method using a Wiener filter is introduced to the processing of data. By scanning the laser position and filtering the time-trace of the reflected acoustic pulse, we have obtained 3D images of the underground objects. The images give us the clear discrimination of the shapes of underground objects. The quality of the images depends on the mismatch of acoustic impedance of buried objects, the bandwidth of acoustic sensor, and the selection of filter function.

The goal of the waterjet-based mine location and identification project is to find a way to use waterjets to locate and differentiate buried objects. When a buried object is struck with a high-pressure waterjets, the impact will cause characteristic vibrations in the object depending on the object's shape and composition. These vibrations will be transferred to the ground and then to the water stream that is hitting the object. Some of these vibrations will also be transferred to the air via the narrow channel the waterjet cuts in the ground. Currently the ground vibrations are detected with Doppler radar and video camera sensing, while the air vibrations are detected with a directional microphone. Data is collected via a Labview based data acquisition system. This data is then manipulated in Labview to produce the associated power spectrums. These power spectra are fed through various signal processing and recognition routines to determine the probability of there being an object present under the current test location and what that object is likely to be. Our current test area consists of a large X-Y positioning system placed over approximately a five-foot circular test area. The positioning system moves both the waterjet and the sensor package to the test location specified by the Labview control software. Currently we are able to locate buried land mine models at a distance of approximately three inches with a high degree of accuracy.

Many operations undertaken by the Defense Department must cope with the active or residual effects of a variety of methods of warfare that a defender can use to inhibit maritime use of the oceans. Prominent cases encountered by the naval forces are mien warfare, salvage and recovery operations, and debris clearance. The office of Naval Research has sponsored development of a family of underwater object location sensors which have a strongly enhanced capability to detect, classify, and identify underwater objects of interest. Use of these sensors in a military exercise, in an operation to locate debris from the Swissair crash, and in a test to demonstrate technology for underwater debris location is described.

The detection and removal of buried landmines and unexploded ordnance (UXO) is one of the most important problems facing the world today. Numerous detection strategies are being developed, including IR, electrical conductivity, ground- penetrating radar, and chemical sensor. Chemical sensor rely on the detection of explosive chemical molecules, which are transported from buried UXO/landmines by advection and diffusion in the soil. As part of this effort, numerical models are being developed to predict explosive chemical signature transport in soils. Modifications have been made to TOUGH2, a general-purpose porous media flow simulator, for application to the chemical sensing problem resulting in the T2TNT code. Understanding the fate and transport of explosive signature compounds in the solid will affect the design, performance, timing and operation of chemical sensing campaigns by indicating preferred sensing strategies.

The detection of buried mines is important to both the military and to civilians, and the possibility of chemical detection may provide a more definitive identification. Understanding the transport of explosive vapors through soil or snow is a major step in the detection problem. In cold regions, the presence of freezing ground or a snow cover complicates the situation, yet also may provide temporary storage of the explosive, potentially leading to opportunities for more optimal sensing conditions later. This paper discusses preliminary work towards adapting an existing 2D heat, mass, and chemical vapor transport model to the problem of explosives transport. The model, originally developed for simulating heat and mass transport through snow under a variety of meteorological conditions, shows promise for simulating explosives vapor transport for buried mines.

The qualitative and quantitative determination of nitroaromatic compounds such as trinitrotoluene (TNT) and dinitrotoluene (DNT) in water and soil has applications to environmental remediation and the detection of buried military ordnance. Recent results of laboratory and field test have shown that trace level concentrations of these compounds can be detected in water, soil, and solid gas samples taken from the vicinity of submerged or buried ordnance using specialized sampling and signal enhancement techniques. Solid phase micro-extraction methods have been combined with Ion Mobility Spectroscopy to provide rapid, sub-parts-per-billion analysis of these compounds. In this paper, we will describe the gas. These sampling systems, when combined with field-portable IMS, are being developed as a means of classifying buried or submerged objects as explosive ordnance.

Neutral electron attachment reaction products of selected analytes were characterized with a dual column gas chromatography mass spectrometer system. The characterization confirmed that electron attachment reaction of polychlorinated aromatics lead to the formation of hydrodechlorination products. Nitroaromatics, by contrast, do not yield any discernable neutral products. Nitroaromatic molecules therefore may be amenable to solute modulation through electron attachment reactions.

High-density optical arrays of fluorescent micrometer sized sensors show promise for detecting low level vapor phase explosives and explosives-like compounds. Imaging software and a high-speed CCD detection system are used to capture changes in a temporal response pattern upon pulsatile vapor delivery to the sensors. Nitroaromatic compounds, such as 2,4-dinitrotoluene (DNT) and 1,3-dinitrobenzene, which are often present on the solid surface above buried landmines, are used to train pattern recognition networks for vapor phase detection. We have demonstrated previously that approximately 9 ppb 2,4-DNT can be detected when signal processing schemes are employed.

Thin films of carbon black-organic polymer composites have been deposited across two metallic leads, with swelling- induced resistance changes of the films signaling the presence of vapors. To identify and classify vapors, arrays of such vapor-sensing elements have been constructed, with each element containing a different organic polymer as the insulating phase. The differing gas-solid partition coefficients for the various polymers of the sensor array produce a pattern of resistance changes that can be used to classify vapors and vapor mixtures. This type of sensor array has been shown to resolve all organic vapors that have been analyzed, and can even resolve H₂O from D₂O. Blends of poly(vinyl acetate) and poly(methyl methacrylate) have been used to produce a series of sensor that response to vapors with a change in resistance of a magnitude that is not simply a linear combination of the responses of the pure polymers. These compatible blend composite detectors provided additional analyte discrimination information relative to a reference detector array that only contained composites formed using the pure polymer phases. Vapor signatures from chemicals used in land mine explosives, including TNT, DNT, and DNB, have been detected in air in short sampling time and discriminated from each other using these sensor arrays.

MEMS technology was used to fabricate arrays of sensor for detecting the explosive micro particulate residue found in mine fields. MEMS devices were fabricated by a surface micromachining process provided by MCNC. The sensor consists of a bimorph structure of polysilicon/gold cantilevers. An optical detection system was designed to detect the deflection of the cantilevers. The sensors were heated by either radiation or conduction using an UV lamp and a small heater under the sensor chip respectively. The deflection of the cantilevers with increasing temperature is presented. Experiments have been performed to detect the response of the cantilevers in the presence of an explosive particle. The cantilevers show a response due to the presence due to the presence of nanograms of TNT and RDX in the vicinity of the cantilevers. Currently it is understood that the response shown by the cantilevers is due to the vaporization of the micro particles, which pulls significant heat out of the temperature sensitive beams causing detectable beam motion. The chemical selectivity of the sensor is provided by the unique melting temperatures of TNT and RDX.

Sub-miniature thin film resonators (TFR) operating near 2 GHz are being developed as mass transducers for high- sensitivity vapor detection. The TFR sensor are coated with species selective vapor absorbing polymers to develop pattern response to the target species for detection and identification. Eight member arrays of TFR sensor have been fabricated and tested for the detection of characteristics explosive vapors including trinitrotoluene, dinitrotoluene, and dinitrobenzene. The TFR sensor use aluminum nitride as the active piezoelectric element and have ben fabricated with resonator quality factors greater than 200. Response patterns and sensitivity measurements are being made using pure vapors, water solutions of the target species, and solid contaminated with the target vapor species.

Recently there has been an increasing demand to perform real-time in-situ chemical detection of hazardous materials, contraband chemicals, and explosive chemicals. Currently, real-time chemical detection requires rather large analytical instrumentation that are expensive and complicated to use. The advent of inexpensive mass produced MEMS devices opened-up new possibilities for chemical detection. For example, microcantilevers were found to respond to chemical stimuli by undergoing changes in their bending and resonance frequency even when a small number of molecules adsorb on their surface. In our present studies we extended this concept by studying changes in both the adsorption-induced stress and photo-induced stress as target chemicals adsorb on the surface of microcantilevers. For example, microcantilevers that have adsorbed molecules will undergo photo-induced bending that depends on the number of adsorbed molecules on the surface. However, microcantilevers that have undergone photo-induced bending will adsorb differently molecules on their surfaces. Depending on the photon wavelength and microcantilever material, the microcantilever can be made to bend by expanding or contracting. This is important in cases where the photo- induced bending of microcantilevers for chemical detection.

We have developed a novel chemical detection technique based on IR micro-calorimetric spectroscopy that can be used to identify the presence of trace amounts of very low vapor pressure target compounds. Unlike numerous recently developed low-cost sensor approaches, the selectivity is derived from the unique differential temperature spectrum and does not require the questionable reliability of highly selective coatings to achieve the required specificity. This is accomplished by obtaining the IR micro-calorimetric absorption spectrum of a small number of molecules absorbed on the surface of a thermal detector after illumination through a scanning monochromator. We have obtained IR micro- calorimetric spectra for explosives such as TNT over the wavelength region 2.5 to 14.5 micrometers . Thus both sophisticated and relatively crude explosives compounds and components are detectable with this technique due to the recent development of ultra sensitive thermal-mechanical micro-structures. In addition to the above mentioned spectroscopy technique and associated data, the development of these advanced thermal detectors is also presented in detail.

An antibody-based competition assay has been developed using a surface plasmon resonance (SPR) sensor platform for the detection of trinitrotoluene (TNT) in soil extract solutions. The objective of this work is to develop a sensor-based assay technology to use in the field for real- time detection of land mines. This immunoassay combines very simple bio-film attachment procedures and a low-cost SPR sensor design to detect TNT in soil extracts. The active bio-surface is a coating of bovine serum albumin that has been decorated with trinitrobenzene groups. A blind study on extracts from a large soil matrix was recently performed and result from this study will be presented. Potential interferant studied included 2,4-dinitrophenol, 2,4- dinitrotoluene, ammonium nitrate, and 2,4- dichlorophenoxyacetic acid. Cross-reactivity with dinitrotoluene will be discussed. Also, plans to reach sensitivity levels of 1ppb TNT in soil will be described.

We report surface-enhanced Raman scattering (SERS) for vapors of 2,4-dinitrotoluene (2,4-DNT), 1,3-dinitrobenzene, 4-amino-2, 6-dinitrotoluene and trinitrotoluene (TNT) adsorbed onto gold metal foils. Detection of 2,4-DNT down to approximately 1 ppb has been demonstrated. A compact field portable Raman unit with fiber optic SERS attachment has been fabricated and field tested for landmine detection. Preliminary results showed little environmental interference to the SERS measurement and detection of a buried landmine. The results demonstrate that SERS can detect buried landmines and, with further improvements, has the potential to be a man-portable field unit for landmine detection.

We develop methods for the automatic detection and localization of landmines using chemical sensor arrays and statistical signal processing techniques. The transport of explosive vapors emanating from buried landmines is modeled as a diffusion process in a two layered system consisting of ground and air. The measurement and statistical models are derived by exploiting the associated concentration distribution. We derive a generalized likelihood ratio detector and evaluate its performance in terms of the probabilities of detection and false alarm. To determine the unknown location of a landmine we derive a maximum likelihood estimation algorithm and evaluate its performance by computing the Cramer-Rao bound. The results are applied to the design of chemical sensor arrays, satisfying criteria specific in terms of detection and estimation performance measure, and to optimally select the number and positions of sensors and the number of time samples. To illustrate the potential of the proposed techniques in a realistic demining scenario, we derive a moving sensor algorithm in which the stationary sensor array is replaced by a single moving sensor. Numerical examples are given to demonstrate the applicability of our results.

A series of chemoselective polymers have been designed and synthesized in order to enhance the nitroaromatic sorption properties of coated acoustic wave devices. Acoustic wave devices coated with a thin layer of chemiselective polymer can provide highly sensitive transducers for the detection of vapors or gases. The sensitivity and selectivity of the sensor depends on several factors including the chemoselective coating used, the physical properties of the vapor(s) of interest, the selected transducer, and the operating conditions. To evalute the effectiveness of the chemoselective coatings a polynitroaromatic vapor test bed was utilized to challenge polymer coated SAW devices. Detection limits with the coated SAW sensors, as tested under laboratory conditions, are determined to be in the lower parts per trillion range. FTIR studies were undertaken to determine the nature of the polymer-polynitroaromatic interactions.

Spin cast films of iptycene-derived aryleneethynylene polymers were synthesized and examined for their ability to detect electron deficient aromatics such as 2,4,6- trinitrotoluene and 2,4-dinitrotoluene. These polymers display a fast, reversible fluorescent response to vapors form these analytes, with the fluorescence attenuation dependent on the time of exposure of the films to the vapors. The iptycene-derived polymers have high fluorescence quantum yields in the solid state and display greater spectroscopic stabilities to heating and solvents than similar polyphenyleneethynylene polymers that lack rigid iptycene functionalities. The syntheses of these polymers will be discussed, as will their photophysical properties and response to different analytes.

The concentration of the chemical vapors emanating form landmines is very low. The equilibrium vapor concentration above pure, crystalline TNT at room temperature is approximately 70 ng/liter. It has been estimated that the TNT concentration in the air over a buried TNT-containing landmine is three to six orders of magnitude less than this value. TNT vapor concentrations three orders of magnitude less than equilibrium are difficult to detect with research quality laboratory instruments and are beyond the capabilities of most commercially available field-portable instruments. Hence, new ultra-sensitive detection technologies for explosives are needed. Collaborators at the MIT have synthesized novel fluorescent polymers that have been implemented as sensory materials in a landmine detection system. When vapors of nitroaromatic compounds of the type found in most landmines bind to thin films of the polymers, the fluorescence of the films decreases. A single molecular binding even quenches the fluorescence of many polymer repeat units, resulting in an amplification of the quenching. Analyte binding to the films is reversible, so the films can be reused. A prototype sensor package has been developed that response almost immediately to sub-picogram quantities of target nitroaromatics. The prototype is portable, is lightweight, has low power consumption, is simple to operate, and is relatively inexpensive. Improvements in the sensitivity of the package are expected. A sample preconcentrator is also being developed for use when the concentration of target analytes is to low to be sensed directly.

This paper presents an innovative microfluidic gas concentrator that uses thermophoresis to concentrate a preselected dilute constituent in air. When used in combination with a gas phase detector, such as an ion mobility spectrometer, the ability of the thermophoretic device to increase the concentration of the target constituent as well as to eliminate other interfering impurities, will substantially increase the sensitivity of the resultant system over the detector alone. The compete sensor could be hand held with an estimated sensitivity of 10^-12 to 10^-15 atmospheres with detection times of 0.1 to 10 s respectively.

A rapid-cycling, low-volume, and inert sampling and enrichment device for organic vapors is described. The device consists of two concentric fused silica capillaries. A small portion of the inter-capillary volume, which is cooled with a burst of compressed carbon dioxide, serves as a trap for the semivolatile organics. The trap's low mass permits rapid sampling and desorption cycles suitable for applications requiring fast monitoring of volatile and semivolatile chemicals. The device is devoid of switching valves in the sampling train and consequently does not suffer from analyte loss due to irreversible adsorption or cross contamination. The device was successfully used for sampling highly adsorptive nitroaromatic explosives in air at low concentrations.

We have developed an innovative method using radiation produced by in flight annihilation of energetic positrons to detect hidden explosives and other illegal substances. The system uses either radioisotope or compact accelerator based methods to generate a high energy positron beam. The high energy positrons annihilate in flight producing a tunable, narrow spectrum beam of high energy photons. The photon energy, which is determined by the positron energy, can be chosen to be resonant with elements of the explosive or other target. The concentration of the target material determines the intensity of the return signal. Standard gamma radiation detection techniques are used to detect the emitted gamma rays. Because of the innovative method we use to generate and monochromatize the positron beam, the entire system is inexpensive, compact and portable.

Nuclear Quadrupole Resonance (NQR) combines the compound specific detection capability offered by chemical offered by chemical detection techniques with the spatial coating capability and convenience of an induction coil metal detector. In this paper we present the first results of the detection of TNT by NQR with sufficient for detection of many antipersonnel mines and essentially all antitank mines. In addition, we review the result of a blind in-field demonstration of the system in detecting RDX in which 28 out of 31 RDX-only targets were found with 1 false alarm in a 110 m test lane, and a second test in which 21 out of 21 RDX mines were found with zero false alarms at a clearance rate of 1.1 m² per minute.

Nuclear quadrupole resonance (NQR) is well suited for detecting land mines with non-metallic cases. It provides both spatial localization and chemical identification of explosives. A search coil produces a train of radio frequency (RF) magnetic pulses that perturb the orientation of nitrogen nuclei contained within the explosive material. Following each RF pulse, the nuclei rotate back to orientations of lower energy. As the nitrogen nuclei possess a magnetic moment, their motion following an RF pulse induces a detectable voltage in the search coil. The NQR signal strength depends on the amplitude, frequency, duration and repetition rate of the applied RF pulses. The optimal selection of RF parameters requires knowledge that is not available in practice, such as the location of the explosive with respect to the search coil. Existing NQR detection systems sacrifice signal intensity by using field pulse parameters. We demonstrate that feedback control provides a means for automatically adjusting multiple pulse parameters so that the maximum NQR signal strength is obtained. The advantages afforded to landmine detection using feedback NQR are summarized.

Nuclear quadrupole resonance (NQR) is a technique that discriminates mines from clutter by exploiting unique properties of explosives, rather than the attributes of the mine that exist in many forms of anthropic clutter. After exciting the explosive with a properly designed electromagnetic-induction (EMI) system, one attempts to sense late-time spin echoes, which are characterized by radiation at particular frequencies. It is this narrow-band radiation that indicates the presence of explosives, since this effect is not seen in most clutter, both natural and anthropic. However, this problem is complicated by several issues. First, the late-time radiation if often very weak, particularly for TNT, and therefore the signal-to-noise ratio must be high for extracting the NQR response. Further, the frequency at which the explosive radiates is often a strong function of the background environment, and therefore in practice the NQR radiation frequency is not known a priori. Finally, at the frequencies of interest, there is a significant amount of background radiation, which induces radio frequency interference (RFI). In this paper we discuss several signal processing tools we have developed to enhance the utility of NQR explosives detection. In particular, with regard to the RFI, we exposure least-mean-squares algorithms which have proven well suited to extracting background interference. Algorithm performance is assessed through consideration of actual measured data. With regard to the detection of the NQR electromagnetic echo, we consider a Bayesian discrimination algorithm. The performance of the Bayesian algorithm is presented, again using measured NQR data.

A method of automatically locating mine-like objects in side scan sonar images has been used for building data bases which contain clutter density estimates as a function of geographic location. Such data bases are useful for both operations planning and for subsequent analysis of later side scan surveys of the same area. Since traditional side scan sonar object detection is focused on individual objects rather than a more general description of collections of objects, it is not immediately useful for the problem addressed here. For that reason, we have developed an approach that uses intensity contouring, followed by a simple geometric analysis of the contours, to find clutter. Discrimination is based upon object shape, area, and the presence of nearby shadows. We describe the incorporation of such an algorithm into a processing package known as the Unified Sonar Image Processing System, and we give examples of dummy mine detection and of clutter estimation in a number of side scan sonar images.

A successful Naval power projection campaign may require operations through littoral regions defined by mines. Mines are among the most prolific weapons to other nations intent on inhibiting US Naval forces' ability to project power from the sea. Therefore, mine countermeasures (MCM) is integral to the overall power projection campaign. Countering the mine threat is critical to naval forces ability to shape and dominate the battlespace. Future operations will require rapid transit though the littoral regions.New capabilities need to be developed to use the environment to a tactical advantage. This paper describes how the environment is being used and should be used in support of MCM in the littoral region.

This paper presents the result of an ONR-sponsored ocean- surface reconstruction project. The goal of the LIDAR project is to investigate a method suitable for obtaining the shape, and in particular, the slopes of the large gravitational waves, to be used in a Navy application for under-water mine detection. Towards this goal, Summus has designed, built and tested a laser-based device for water surface measurement. A field test was conducted at the Army Field Research Facility at Kitty Hawk, North Carolina in October 1998. This paper describes the basic design conducted of our method, and the experimental results.

The report consists ofthree parts: 1.Methods and means for observing and 3D imaging of underwater static and dynamic objects (e.g., mines and living beings). The methods are based on optical refraction, diffraction, and the Talbot effects on the water surface disturbed by ultrasound waves reflected or passed by the object16 An algorithm to calculate the shape of disturbed water surface has been developed. It was also used for realization of an ultrasound holography system. Computer — aided cross-section recovery of 3D image of a submerged object is described. The Talbot effect and the algorithm developed are shown to be useful: for measure of the shape, thickness and velocity of oil spreading on the water surface; for investigation drops creation and measuring their forms. This method, for measure the form of any disturbed liquid can be used. For instance, the form of ocean surface disturbed by the wind, which can me used for determining the wind direction. For this goal, for example, the Floating Instrument Platform (FLIP)7, with some additional appliance, may be used. 2. Observation and 3D imaging of static and dynamic objects inside or behind optically opaque solid media using a microwave holographic movie camera recording the radiation passed through the object. 3. The design of the agreement for the feasibility study and technical development of mobile facility for detection of landmines or other objects inside or behind an opaque obstruction using the radiation reflected from the object.

In this paper, five methods for enhancing sonar images prior to automatic detection of sea mines are investigated. Two of the methods have previously been published in connection with detection system and serve as reference. The three new enhancement approaches are a variance stabilizing log transform, nonlinear filtering, and pixel averaging for speckle reduction. The effect of the enhancement step is tested by using the full processing chain i.e. enhancement, detection and thresholding to determine the number of detections and false alarms. Substituting different enhancement algorithms in the processing chain gives a precise measure of the performance of the enhancement stage. The test is performed using a sonar image database with images ranging from very simple to very complex. The result of the comparison indicates that the new enhancement approaches improve the detection performance.

Even under ideal conditions side-scan sonar (SSS) images of targets can vary greatly depending on the target range and orientation, even if their geometries are identical. This complicates target classification algorithms since typically only a small samples of targets are available for training purposes. This under-representation of targets can cause missed classifications and a higher false alarm ratio in the presence of clutter. This problem is addressed by using a priori information about the targets as well as the imaging system embedded in a model for simulating target images. These simulated target images can be added to the training set for a more complete target representation. Another important aspect of this research includes the use of multiple channels extracted from the images using a multi- resolution wavelet decomposition. This multi-resolution analysis is used to first provide for an efficient detection strategy, by filtering the images over the lower resolution channels. Furthermore, providing target features at different scales improves the performance of the neural network classifier. The dependence of the classifier on local image enhancement provided by total variation minimization and Mumford-Shah segmentation is also studied.

An adaptive image regularization algorithm, based on the NoN neural computing theory, is applied to enhance mine signatures. The algorithm, developed by Guan and Sutton (GS), uses vector connections among model neurons to delineate dynamic boundaries corresponding to critical features of images. The boundaries subdivide large networks into many smaller networks, where each smaller network has, in many instances, attractor properties. In this report, the GS algorithm is applied to deblur and segment three sets of underwater mine data. The results suggest that the GS algorithm requires minimal training, performs well under inhomogeneous conditions and generates contours, which can be fed into other NoN architectures for further processing, including classification.

A new approach to detection of sea-mines in sonar imagery that improves the detection density ACF method is presented. The steps are: 1) background normalization, 2) spatially adaptive Wiener filtering, 3) convolution with a 2D FIR filter matched to the target signature, 4) adaptive thresholding to reduce noise, 5) extraction of higher-order spectral features to capture the spatial correlations, 6) extraction of size, strength, and density features, 7) optimal feature selection, and 8) classification. An adaptive Wiener filter is applied to remove noise without destroying the structural information in the mine shapes. The FIR filter is designed to suppress noise and clutter, while enhancing the target signature. A double peak pattern is revealed as the filter passes over highlight and shadow regions. The location, size, and orientation of this pattern can vary. Higher-order spectral features capture the spatial correlations in this pattern and provide invariance to translation and scaling. The approach has been tested on the CSS Sonar 3 database of 60 images with about 84 percent classification accuracy and 11 percent probability of false alarm.

A parameter-free and a priori-information-free preprocessing of sonar images is proposed, which permits a ranking of local extrema in the image according to their likelihood to be amine-like objects. It is shown that an acceptable fully automatic detection algorithm can be built on a variational method which estimates shape information of the possible mines. This algorithm does not use any a priori information on the type of mine or range distance or background type and works without any change on both sonar databases we had available. It therefore can be used as a detection algorithm without any information request the use or designer. Its results could be fed into a classification algorithm like the one proposed. We also think that the features computed by this variational method could serve for both the detection step and the classification step, thus reducing the number of designer's parameters and opening the way to a parameter-free detection-classification algorithm.

An image fusion method is developed for sonar systems that collect multiple images in a single pass. This research builds on our past work with automated detection and classification of sea mines in side-looking sonar imagery. The new method has the following processing step. Each image is processed separately by our automated detection and classification algorithm. Next detections in the images are collated into two general categories: (1) detections appearing in only one image and (2) detections appearing in multiple images. A fuzzy-logic procedure is used to fuse the detections that are common to multiple images. The final 'fused' classifications will fall under two general descriptions: (1) classifications where sufficient mine-like evidence had to be accumulated over multiple images. This fusion process dramatically reduces false alarms.

A complete mine detection/classification (D/C) system has been specified and implemented, which runs in real-time, and has been exercised on the latest available dual-frequency side-scan sonar acoustic image sets. The compete DC system is comprised of a collection of algorithms that has been developed and evolved at Draper Laboratory over the past decade. The detection process consists of image normalization, enhancement, segmentation, and feature extraction algorithms. The enhancement algorithm is a variant of a Markov Random Field based anomaly screener developed in FY-94. The feature that were extracted were those derived in FY-93. A distance constrained matching algorithm, which was developed in FY-95, is used to generate a list of high and low frequency fused tokens. The classification process involves the evaluation of a hierarchy of three multi-layer perceptron neural networks: HF, LF, and HF/LF fused. Research performed in FY-95 also concentrated on the development of several variants of information fusion with hierarchical neural networks. The 'discriminant-combining' variant of fusion was selected as part of this DC system. In addition, a classification post- processing and decision node statistic modification step, which was developed in FY-96, was included. This paper will describe the algorithm that were implemented. However, the emphasis will be on the performance results of processing the latest available side-scan imagery, comparison of single sensor vs dual-frequency sensor results, and the issues that were encountered while exercising the DC system on the new data set.

An advanced, automatic, adaptive clutter suppression, sea mine detection-classification and fusion processing string has been developed and tested with sonar imagery data. The overall string includes pre-processing, adaptive clutter filtering (ACF), normalization, detection, feature extraction, classification and fusion processing blocks. The ACF is a multi-dimensional adaptive linear FIR filter, optimal in the Least Squares sense, and is applied to low- resolution data. It performs simultaneous background clutter suppression and preservation of an average peak target signature. Following 2D normalization, the detection consists of thresholding, clustering of exceedances and limiting the number of detections. Subsequently, features are extracted from high-resolution data and an orthogonalization transformation is applied to the features, enabling an efficient application of the optimal log- likelihood-ratio-test (LLRT) classification rule. Finally, the classified objects of the LF and HF processing strings are fused. The utility of the overall processing string was demonstrated with two new shallow water high-resolution sonar imagery datasets. The processing string classification performance was optimized by appropriately selecting a subset of the original feature set. The overall ACF, detection, feature extraction and orthogonalization, LLRT- based classification and fusion processing string resulted in improved mine classification capability, providing a three-fold false alarm rate reduction, compared to previous results. A wide-sense stationary covariance model was utilized in the ACF algorithm design, significantly reducing the algorithm implementation complexity, and the implementation of the overall processing string in real-time was demonstrated.

Most of the fusion algorithms reported operate in a multi- sensor environment. In this paper, we investigate a novel fusion strategy that operates on images obtained from a single sensor. There exists a variety of methods for mien detection in sonar or laser line scan images, but single method can be considered as the best under all circumstances. Targets missed by certain algorithms might be detected by others. It can be shown that, by applying various algorithms to the same mine detection task and fusing the decisions together, one can obtain a superior detector. However, in certain difficult applications, the weaknesses of various classifiers reinforce each other. This may be due to the fact that when most classifiers are developed, the primary goal is always to achieve the highest detection over false alarm ratio. Thus, certain type of targets might be neglected by most of the detection algorithms in favor of many other targets that can be easily detected. This scenario posts challenges to decision level fusion algorithm developers, who have to develop several complementary detection algorithms. The novel fusion strategy we described in this paper applies different image enhancement algorithms to the same image before the classification stage. Since image enhancement algorithms are more diversified than detection algorithms, they tend to highlight complementary features of the mien targets in the same data set, thus producing a pseudo multi-sensor environment. These detection decisions on multiple looks of the same image are then fused together to create a superior detection scheme.

This paper presents a novel formulation of the adaptive filtering concept ideally suited for operating in highly non-stationary environments. A new set of filtering weights ins designed for each data point to be processed, using solely information extracted from the immediate vicinity of the data pointing under consideration. A single 'snap-shot' of local data, equal in extent to the filter window is used to estimate the filter weights in this procedure, allowing the filtering to adaptive to all kinds of background variations. Numerical stability is achieved by incorporating to the traditional constraints the requirement that the adaptive weight vector be of unit length, thus providing the added benefit of preserving the white noise floor power.

Nearest neighbor classifiers with direct sum successive approximation (DSSA) templates are shown to be effective for detecting and discriminating mines and mine-like objects in forward looking sonar data. DSSA results are demonstrated on data obtained form field measurements with actual mines and calibration targets. The DSSA templates are used in a nearest neighbor classifier that can be characterized as a new type of radial basis function neural network. This neural network is not designed with a preset complexity level as quantified by an a priori determined number of degrees-of-freedom. Rather, the system is constructed incrementally and adds additional degrees-of-freedom as required by the nature of the training data. The neural net system possesses stage structure that result in inherent computational and memory efficiency in searching and storing the DSSA-based radial basis functions.

A local discrimination basis is selected among the possible wavelet packet bases in order to create a two-class classifier for high dimensional signals. We use the classifier for target detection in images and signals. The position of the target is also determined. A method invariant to the localization of the target in the analyzed image is developed in order to reduce the time complexity of the algorithm. We apply the method on mine detection in sonar images.

Good discrimination result have been obtained with an active backscatter data set of mine-like objects, where the task as to distinguish between man-made and non-man-made objects. In this work we introduce a novel method for constructing best basis for discrimination from wavelet packets, and demonstrate the superiority of multiple ensembles of predictors. We achieve far better discrimination results using a wide band FM sweep of 80kHz compared with the earlier work that used a 40kHz FM sweep. We further show improved results by combining several discriminative methods with several wavelet packet representations.

Principal component analysis (PCA) in the wavelet domain provides powerful features for underwater object recognition applications. The multiresolution analysis of the Morlet wavelet transform (MWT) is used to pre-process echo returns from targets ensonified by biologically motivated broadband signal. PCA is then used to compress and denoise the resulting time-scale signal representation for presentation to a hierarchical neural network for object classification. Wavelet/PCA features combined with multi-aspect data fusion and neural networks have resulted in impressive underwater object recognition performance using backscatter data generated by simulate dolphin echolocation clicks and bat- like linear frequency modulated upsweeps. For example, wavelet/PCA features extracted from LFM echo returns have resulted in correct classification rates of 98.6 percent over a six target suite, which includes two mine simulators and four clutter objects. For the same data, ROC analysis of the two-class mine-like versus non-mine-like problem resulted in a probability of detection of 0.981 and a probability of false alarm of 0.032 at the 'optimal' operating point. The wavelet/PCA feature extraction algorithm is currently being implemented in VLSI for use in small, unmanned underwater vehicles designed for mine- hunting operations in shallow water environments.

This paper presents a new adaptive underwater target classification system for in situ robust classification ins presence of environmental and target signature changes. A wavelet packet-based feature extraction scheme is used as a front-end processor to extract pertinent features in each subband. The extracted features correspond to the linear predictive coding coefficients of the signals in each subband. The heart of the system is an adaptive feature mapping subsystem that maps the original feature space in such a way that the mapped feature vector remains invariant to the environmental and sensory conditions. The goal is to minimize the classification error of the neural network classifier on the changing data. The feedback to the adaptation mechanisms is provided by a K-nearest neighbor classifier that can also be updated in situ in face of changing environment. Preliminary results on broadband acoustic backscattered signals collected for six different objects are obtained which reveal the effectiveness of the system when compared with the non-adaptive system.

Many sensors for landmine detection are affected by soil water content, temperature, electrical conductivity and dielectric constant. The most important of these is water content since it directly influences the three other properties. We model water distribution around antitank mines buried in a loam and loamy sand soil under the climatic conditions of Bosnia and Kuwait. In Kuwait the loam and loamy sand have mean soil water contents of about 16 and 7 volume percent, respectively; in Bosnia, the mane water contents are higher with means of 30 and 14 volume percent in the loam and loamy sand. As a result the soil dielectric constant in Kuwait varied from about 4 to 8 in the loamy sand and from 8 to 14 in the loam. In Bosnia the higher water contents result in a soil dielectric constant from 4 to 12 in the loamy sand and from 9 to 50 in the loam. Water contents below the landmine were sometimes higher than above it. The modeling result demonstrate that a solid water content regimes and the resulting distributions of soil dielectric constants around landmines are strongly affected by the interaction between climate, soil type, and landmine geometry.

The detection of non-metallic anti-personnel landmines with ground penetrating radar (GPR) is complicated by low dielectric contrasts with the surrounding background medium. Previous studies have shown that the addition of water can improve dielectric contrasts but also increases loss so that target detectability is not necessarily improved. Previous studies have also shown that the addition of liquid nitrogen to wet soils can reduce background medium loss and restore target visibility. In this paper, further waveguide studies of target detection through a controlled depth of nitrogen penetration are reported, and it is shown that scattering from known depth targets can be significantly enhanced if an optimal amount of nitrogen is added. The procedure can also be generalized to unknown depth targets if measurements are taken as gradually increasing amounts of liquid nitrogen are added. Both analytical models and waveguide experiments are presented to illustrate these ideas. Finally, initial test of the soil modification techniques developed through waveguide experiments are reported with a dielectric rod GPR system; results indicate that these methods should be applicable to general GPR sensors.

We are investigating the environmental effects on radar detection of simulant mines (SIMs). SIMs are standard test targets developed by the US Army Project Manager-Mines, Countermine and Demolitions, and VSE Corporation for testing and evaluation of mine detection equipment. These test targets are filed with RTV silicone rubber, which has similar dielectric properties as TNT and Composition B. Therefore, they interact with radar sensors in a way representative of live mines. We are using broadband frequency modulated continuous wave (FMCW) and impulse radars to obtain signatures of SIMs buried under controlled laboratory conditions and at a test site instrumented with sensors to monitor the state of the ground. We find that anti-tank SIMs buried in frozen soil, in our case a common, silty sand are easy to detect. The dominant resonances included within SIMs by a broadbeam, 1.5 GHz impulse radar are of-nadir responses that appear unique and not predictable by simple ray theories of diffraction. A narrow beam, 2-6 GHz bandwidth FMCW radar induced reflections from the top and bottom of SIMs that were clearly resolved due to the broad bandwidth of the FMCW radar.

Existing and emerging handheld mine detectors do not yet perform at the standards established for full capability. To correct this deficiency we must expand our efforts by developing mine detection systems that find mines at sufficiently high rates and raise acceptably few false alarms so as to be operationally effective. The proposed Handheld Standoff Mine Detector System (HSTAMIDS) provides the potential for a significant capability increase over its predecessor, the ANIPSS- 1 2 . However, the widespread proliferation of low metal and no metal mines has changed the nature of the threat and increased the capability shortfall. In its present form and in the hands of the average soldier, the HSTAMIDS misses too many mines and will likely have far too many false alarms to be practically handled in a genuine mine-clearing mission. In the hands of highly motivated and experienced contractors, the HSTAMIDS still does not meet the detection requirements for the program, and the false alarm rate is even higher than for soldiers. Fortunately, there are ways to improve system performance of technology-limited detectors. We can use existing resources, which, if properly managed, can significantly improve performance of future iterations of this system. The purpose of this article is to set forth that solution and provide an expanded approach to developing a handheld mine detection system.

This paper describes findings from a study conducted to analyze the behavior, knowledge, and thinking that support the highest levels of human land mine detection skill. A recent assessment of land mine detection capability concluded that 'human operators perform better with any detector system than the corresponding fully automated system.' This assessment, plus evidence linking individual differences in detection ability to experience, suggests that methods, data, and theory developed in studying human expertise can be applied to the problems of land mine detection and discrimination. Studies of experts across a variety of skill domains have demonstrated that analyses of experts' skills can yield findings useful for designing efficient and effective training programs and supporting technology development. This initial field study was performed to (a) identify the upper levels of human mine detection capability using currently-fielded hand-held equipment and (b) model the knowledge, thickening, and techniques employed by proficient human operators. Two experienced operators showed sufficiently impressive detection performance to qualify as experts. Data laos show that skilled PSS-12 operator can detect low-metal mines with considerable accuracy. A first-approximation information- processing model of expert operator skill is presented that is based on observation of the operators' activities as they searched for mine targets.

Most handled landmine detection systems use tones or other simple acoustic signals to provide detector information to the operator. Such signals are not necessarily the best carriers of information about the characteristics of hidden objects. To be effective, the auditory signals must present the information in a manner that the operator can comfortably and efficiently, the auditory signals must present the information in a manner that the operator can comfortably and efficiently interpret under stress and high mental load. The signals must also preserve their audibility and specific properties in various adverse acoustic environments. This paper will present several issues on optimizing the audio display interface between the operator and machine.

Although the ability of EMI sensor to detect landmines has improved significantly, false alarm rate reduction remains a challenging problem. Improvements have been achieved through development of optimal algorithms that exploit models of the underlying physics along with knowledge of clutter statistics. Moreover, experienced operators can often discriminate mines from clutter with the aid of an audio transducer, the method most commonly used to alert the sensor operator that a target is presented. Assuming the basic information needed for discriminating landmines from clutter is largely available from existing sensors, the goal of this work is to optimize the presentation of information to the operator. Traditionally, an energy calculation is provided to the sensor operator via a signal whose loudness or frequency is proportional to the energy of the calculation is provided to the sensor operator via a signal whose loudness or frequency is proportional to the energy of the received signal. Our preliminary work has shown that when the statistic used to make a decision is not simply the signal energy the performance of mine detection systems can be improved dramatically. This finding suggests that the operator could make better use of a signal that is a function of this more accurate test statistic, and that there may be information in the unprocessed sensor signal that the operator could use to effect discrimination. In this paper, we investigate and quantify, through listening experiments, the perceptual dimensions that most effectively convey the information in a sensor response more appropriately to the listener, discrimination, as opposed to simple detection, can be achieved.

The effectiveness and robustness of any landmine detector ultimately depend on its operator. This is especially true for hand-held landmine detectors, since the operator handles both the scanning motion and the interpretation of the data. Therefore, it is important that the human-in-the-loop issues are addressed as an integral part of the detector design, not as an afterthought. Two critical issues that we have identified are the lack of position feedback for the operator and the lack of 2D map of the detector output. The position feedback will allow the operator to obtain feedback with respect to the sweep rate, detector height and orientation. The position feedback can also be integrated with the detector output to generate a 2D map for the operator. In addition, the 2D map enables 2D image processing techniques, which are more robust and effective than 1D signal processing techniques.

The feasibility of detecting surface-laid and, in some circumstances, buried mines by analysis of visible to near- IR (VNIR) hyperspectral imagery has been demonstrated by the authors in previous studies. An important factor in the practical success of such technology is being able to achieve the necessary spatial and spectral resolution to allow discrimination of mines from background. With some restrictions, both can be improved by increasing the instrument data output rate or decreasing the platform speed and both can be traded off against one another. The optimum trade-off must be determined for a given problem, including the choice of algorithm. Airborne VNIR hyperspectral data were collected over several controlled surface-laid mine fields using a casi hyperspectral imager and a helicopter. The combination of the imager's high speed data recording coupled with the low airspeed of the helicopter enabled the collection of hyperspectral data ranging from four 136 nm wide spectral bands at 10 cm resolution to nine 60 nm wide spectral bands at 20 cm resolution. Each mine field contained a variety of mines ranging form small anti- personnel mines to large anti-vehicle mines. An assessment of the feasibility and practicality of using airborne hyperspectral data to detect various surface-laid mines and mine fields was conducted. In addition, the trade-offs between spectral and spatial resolution for the detectability of surface-laid miens and mine fields are discussed.

The objective of the US Army Hyperspectral Mine Detection Phenomenology program was to determine if spectral disciminants exist that are useful for the detection of land mines. A primary goal wa to determine the presence and persistence of spectral features produced by buried anti- tank mines as associated with soil properties and vegetation changes over time. Details of the collections are documented in the ERIM International Technical Report 10012200-15-T, 'Mine Spectral Signature Collections and Data Archive', March 1999. This paper describes the HMDP project and focuses on the initial phase of controlled experimental measurements of spectral mine signatures in ground-based US collections. The foreign data collections are not addressed in this paper. Some of the HMDP project's mine spectral signature result are highlighted here. Detailed analyses of these data were performed and is described in a companion paper in this conference titled 'Detection of Land Mines with Hyperspectral Data'.

Automatic mine detection is an area of intense research due to the implications in humanistic and battlefield management related issues. In this paper, we describe a fully automatic and iterative implementation of the nonlinear MM-MNF algorithm and review its performance for detecting landmines in multi-spectral images provided by the Coastal Battlefield Reconnaissance and Analysis program. The MM-MNF algorithm utilizes a powerful linear multi-spectral enhancement tool, called the Maximum Noise Fraction (MNF) transform, in conjunction with a nonlinear detection device based on mathematical morphology. The iterative implementation of this algorithm improves the accuracy of the clutter covariance estimation, which is turn decreases the number of false alarms, as compared to a previously reported implementation. The result are significantly better than the ones obtained from a constant false alarm rate algorithm, known as the RX-algorithm, whose performance was also inferior to the previous implementation of the MM-MNF algorithm.

An IR mine detection system has been developed which reliably detects buried land miens in certain environmental conditions. The system uses two commercial IR cameras, one using the 3-5 micron band and the other using the 8-12 micron band. The cameras are mounted above a HMMWV and are tilted down to look 2.5 to 7 meters ahead of the vehicle, covering almost a four meter wide swath. Software algorithms are used to de-warp the raw images to remove keystoning and lens curvature effects, producing rectilinear images of the terrain. The mines are observed in the images as cool spots or hot spots on the surface of the soil, with temperatures depending upon the mine type, the recent temperature history of the soil, and the moisture content and soil type. Raw video images are presented which show some of these effects, including when the contrast is so low to be visibly hidden in the background. Filtering algorithms are utilized to perform background identification and removal, producing an equalized image, which enhances the contrast of the buried mines. Statistical order filters are employed that further enhance the mines, with examples again shown. Threshold and object detection algorithms have been developed that autonomously detect mine-like objects in the images without operator intervention. Feature extraction algorithms then search for features that distinguish the mines sought, including such features as size and shape. The objects are classified as a mine or a non-mine and this decision passed on to the registration and hi-level inference detection subsystems of the mine detection platform.

The US Army has a major interest in the development of an airborne system capable of detecting surface and buried anti-tank mines under all-weather, day/night conditions. While the hardware components are essentially an airborne reconnaissance system generating high resolution image data, it is the processing algorithm which provides the mine/minefield detection capability. During the spring and summer of 1997, flight test of several minefield detection systems were concluded at Ft. Huachuca, AZ. The systems tested represented three major technology approaches, namely passive thermal IR, active near IR laser imager for polarization based discrimination and hyperspectral. Image data generated by these sensors were recorded to tape or other media and subsequently processed using various minefield detection algorithm approaches.

In this study we investigate the detection of plastic anti- personnel mine simulants of several sizes using radar and IR imaging. The problem is first studied numerically with the finite-difference, time-domain method and an inverse synthetic aperture radar imaging algorithm to estimate the optimum conditions in dielectric contrast between the mine (epsilon) _m and the soil (epsilon) _s. It is found that to obtain a recognizable shape, the size of the simulant should be about two wavelengths when the dielectric contrast is low. Range of soil permittivity considered is from 2.8 to 6.0 corresponding to dry and wet soil conditions up to a volumetric soil moisture of 10 percent. The diameter of the mines ranged from 6 cm to 12 cm and their height ranges from 3.5 cm to 5 cm. It is found that when a mine diameter is larger than a wavelength, its image is discernible and becomes clearer when its diameter exceeds two wavelengths. It is found that a plastic mine with a dielectric constant of 2.5 embedded in a dry soil medium with a dielectric constant of 2.8 and locally flat surface can generate an image with the correct geometric shape. An experimental study of microwave-IR imaging is also conducted in this study. The key impact on IR imaging is the change in the amount of moisture in the soil medium caused by the presence of the simulant. Thus, different amounts of moisture are being heated by the microwave in regions with and without the simulant. This in turn causes a temperature difference leading to the IR image. Since the simulant contains no appreciable moisture, the temperature over the simulant is generally lower than the region around it. The soil inhomogeneity and heating pattern effects are discernible in the IR images.

The ultimate goal of the authors is apply inverse problem methods to image land mines using a electromagnetic GPR signal. Specifically, the intention is to use the recently developed Elliptic Systems Method, which has been successfully applied by these authors to the problem of laser imaging of biological tissues. As the first step, however, one should develop a fast and accurate numerical method for the solution of the forward problem to simulate the data for the inverse problem. The main difficulty of the latter consists of the requirement of solving a Helmholtz- like equation for high frequencies which is very time consuming using standard direct solution techniques. A novel accurate and rapid numerical procedure for the solution of this equation is described in this paper.

A vehicle-width array of metal detectors is one of the sensor systems used in most present day vehicle-mounted mine detectors. Data furnished by such a metal detector array consist of an output from each sensor channel as a function of time which is usually converted to a function of position. In multisensor systems where target-level data fusion is used, there is a need for techniques to process such data in order to detect and locate targets in realtime as the array scans the ground surface. One conventional way of processing such data is to apply a thresholding algorithm to data from each sensor channel separately and infer the presence of a target under a given coil or a number of coils. Such as approach could be very limited and cumbersome particularly when one has to consider large arrays with complex interaction between sensor and targets that produce a response in a number of sensor channels simultaneously. In this paper we model the data from the detector array as a scrolling image and develop a target detection and location scheme based on image processing concepts. Modifications of multiresolution and template-matching algorithms of 'peak' detection are developed using domain -specific knowledge of metal detector arrays. The resulting technique, which also uses dynamic thresholding to allow realtime operation, is illustrated using measured data from a 24-element, 3-meter wide metal detector array.

This paper reports the results of a preliminary assessment of Electrical Impedance Tomography (EIT) technology to detect the presence of mine-like objects in soils. EIT uses an array of electrodes to inject low frequency currents in the soil and measure the resulting electrical potentials. The measurements are then used to reconstruct the conductivity distribution underneath the array. In the coarse of this work, an EIT system was built and evaluated. The array is made of 30 stainless-steel stimulating and recording electrodes arranged in a 5 by 6 grid. A data acquisition card, under computer control, is used for the current stimulation and potential measurements. An algorithm was also developed to produce the image reconstruction of the conductivity distributions. The algorithm uses a linear regularized approach to invert a sensitivity matrix. The objective of this work was to assess the major difficulties involved in using EIT as a candidate technology for landmine detection, namely the repeatability of the measurements and the practical feasibility of the image reconstruction. The evaluation was performed on a scaled down lab prototype, using soil containers and a mine-like object representative of an anti-tank mine. Measurements were taken in black earth, clay and sand environments with a mine-like object having a width of 2 inter-electrode spacings. The result indicate that the EIT prototype could detect the presence of the mine-like object down to a depth of about 1.5 inter- electrode spacing, for the range of medium conductivities considered.

A series of buried land mine detection measurements were performed at the University of Florida using x-ray lateral migration radiography with 12 difference types of actual antitank and antipersonnel mines. The resulting images posses extraordinarily definitive detail. The signatures are so unique that not only can positive mine detection be accomplished with this technique, but also mine identification. The mine's exterior shape combined with the interior air volumes yield easily recognized image signatures. The emphasis of this paper is on mine-type discrimination from image data. The reported results indicate that the lateral migration radiography technique provides a land mine detection method with the potential of near-zero false positive alarm probability. A practical systems, which is under current design and fabrication, is described and allows for one square meter interrogation in 35 seconds, antitank and antipersonnel mine imaging and recognition in respectively 12.6 and 1.4 seconds. This approximately 75-kilogram system can be attached to a small two-wheel carrier and requires only 140 watts of electric power.

The US Army Research Laboratory (ARL), as part of customer and mission-funded applied research programs, has been evaluating the use of low-frequency, ultra-wideband imaging radar to detect fields of buried mines. An instrumentation- grade measurement system has been designed and implemented by ARL. Four data collection campaigns in support of ground- penetrating radar objectives have led to the establishment of a significant and unique database of radar imagery. We are using these data to develop mine-detection algorithms that can aid an operator in separating mines from background clutter. This paper reviews recent findings and result from ARL's modeling, phenomenology, and algorithm development efforts. At SPIE '98, we reported on the performance of a physics-based mine-detection algorithm using data collected at Yuma Proving Ground (YPG) in January 1996. Subsequent measurements were made using the ARL BoomSAR at YPG in October 1997, January 1998, and June 1998. Most of the mines from the January 1996 experiment were still in place during the 1997/1998 experiments. Additional mines and unexploded ordnance were added to the YPG test after the January 1996 experiment. This paper discusses the difference in soil conditions from these data collections and the impact that may have on a mine's radar cross section (RCS) and detection performance. Detection results for M20 mines under different soil conditions will be shown. The detection algorithm invokes phenomenologically sound feature that exploit the expected mine RCS, texture, frequency dependent scattering, and model-based image correlation. Performance assessments, in terms of receiver operating characteristics, detail the detection capabilities at various false alarm rates. Finally, new imagery will be presented that shows the positive contrast of low metal content above dielectric background.

This paper describes an automatic mine detection algorithm (AMDA) based on the template matching technique. Specifically, this paper demonstrates that regardless of sensor artifacts and other perplexities including environment effects such as terrain variation or weather conditions, there will be distinctive information between targets and clutter imbedded in the signatures for the discrimination. This paper also includes the data analysis of the ground penetration radar signatures and quantifies the AMDA performance. This paper use a subset of the DARPA clutter data collected with the Geo-Centers ground penetration radar at Fort A.P. Hill and Fort Carson. This subset contains anti-personnel, and anti-tank mines buried from 1 to 6 inches deep with the size of the mine ranging from 2 to 12 inches in diameter. The total number of mines and the area coverage of this subset are about 30 and 600m², respectively.

In order to separate buried land mines from clutter a multi- channel stepped-frequency ground penetrating radar has been developed. The system operates over the frequency band 800 MHz to 2.0 GHz. The radar incorporates advanced digital signal processing and radio frequency integrated circuit components. It uses an all-digital modulator coupled with a coherent digital quadrature receiver for making precise magnitude and phase measurements. The control interface to the radar consists of an Ethernet TCP/IP link. A parallel bank of transmit-receive antennas is used to achieve cross track sampling. System motion is used to achieve along track data sampling. Synthetic aperture near field beamforming techniques are used to image buried objects. The system is designed to detect shallowly buried metallic and non- metallic mines. A system overview is presented and result from data collection exercises are included. Images and analysis of data from a mine lane is presented.

We describe several automatic mine detection algorithms in this paper. These methods were tested on real Ground Penetrating Radar (GPR) data and showed dramatic improvement in terms of probability of detection and false alarm rates compared to energy based techniques. The main contributions of this paper are as follows. (1) Only background clutter data, instead of mine data, are needed for the development of the algorithms, which makes collection of data for training and adapting the algorithms to new environment much easier than methods requiring both clutter and mine data for training. (2) The mine detection algorithms are developed in a fairly general form, and thus can be ported to other sensor platforms or future generations of mine detection hardware with little modification. (3) The algorithms require little on-line computation. (4) Adaptation of the algorithms to new environment or mine-fields is done automatically, which reduces human resources and the cost of training.

Ground Penetrating Radar signatures of shallowly buried objects tend to be obscured by the return from the air-soil interface. In this paper an estimate of this background signal is computed locally and, abrupt changes from this estimate, assumed to correspond to the returns from a buried object, are detected using a translation invariant wavelet packet decomposition. The detection statistic is defined as the sum of several peak energy values of the resulting translation invariant wavelet signal representation. The computation takes place in a running window which allows the algorithm to adapt to variations in ground conditions and antenna height.

It has been demonstrated that GPR is an effective sensor for detecting the presence of sub-surface objects. Although this ability is important in regard to mine detection and location, in high clutter environments the false alarm rate from GPR can potentially be unacceptably large. Therefore, an ideal sensor would provide sufficient information not only to detect subsurface objects, but additionally to identify the objects as either a mine or a benign clutter object. To accomplish this for GPR sensor, it is noted that mines exhibit a level of vertical symmetry that is not prevalent in other clutter targets. These symmetric targets have specific polarimetric and bistatic response that are independent of size, shape, dielectric, or target depth. A set of bistatic, polarimetric, GPR observations can be used to identify these specific responses and thus potentially identify objects with planes of vertical symmetries. To examine the efficacy of this target identification technique, an FDTD solution was used to simulate the polarimetric and bistatic scattering form symmetric and asymmetric targets. Analytical measures were developed in order to express numerically the apparent symmetry of a subsurface object, and the detection performance of these measures were evaluated in terms of Receiver Operation Curves. Finally, results using measured GPR data were evaluated.

In pattern recognition, it is crucial to be able to represent objects with feature that contain as much of the information as possible in compact form. A typical 8-bit grayscale digitized image can be sorted using M by N values that represent the intensity levels of individual pixels where M and N are image dimensions. Pattern recognition algorithms use various methods for feature extraction, like chain codes, Fourier descriptors, and invariant moments. We will propose features that will characterize objects much more efficiently. Our feature scan be viewed as basis functions that lead to a set of images within an equivalence class. In order to illustrate the method with an application, these features are then used to train a set of learning RNNs which can be used to detect targets within clutter with high accuracy, and to classify the targets or man-made objects from natural clutter. Experimental data from SAR imagery is used to illustrate the performance of the proposed algorithm. Currently, we are investigating the applicability of this approach to a set of GPR mine data.

A study was conducted for the use of a synthetic aperture interferometric radiometer for mine detection at near-field ranges using a two-element, 1.4-GHz laboratory prototype. Images are presented of an aluminum plate buried at different depths, generate using a backpropagation technique. The parameters required for mine detection are also discussed.

The US Army Research Laboratory (ARL), under the sponsorship of the Strategic Environmental Research and Development Program, is conducting experiments to establish and enhance the ability of low-frequency, ultra-wideband synthetic aperture radar (SAR) to detect and discriminate unexploded ordnance (UXO). Preliminary investigations using ARL's BoomSAR - a UWB radar mounted atop a mobile boom lift platform - concluded that the radar image texture and frequency-dependent scattering from mines and mine-like targets could be exploited in the development of automatic target detection algorithms. To support further investigations, ARL established extensive UXO test sites at the US Army Yuma Proving Ground, Arizona, and Eglin AFB, Florida. The soils at both test sties have been characterized in terms of physical, chemical and electromagnetic properties. Precise location, depth, and orientation information was recorded for each of the approximately 500 inert ordnance test targets at each site. This information helps researchers to better understand the phenomenology associated with UXO target scattering and to more accurately evaluate and modify data processing programs. The ultimate goal is to develop innovative automatic target detection algorithms that provide a high probability of detection with an acceptable false-alarm rate under varying environmental conditions and operational scenarios. This paper present details on the design and characterization of the two test sites and some initial results from BoomSAR data collections.

A full-wave model is developed for electromagnetic scattering from buried and surface land mines, taking rigorous account of the lossy, dispersive and potentially layered properties of soil. The theoretical results are confirmed via synthetic aperture radar (SAR) measurements, performed using the US Army Research Laboratory's BoomSAR, with which fully polarimetric ultra-wideband SAR imagery is produced. The theoretical model is used to predict wave phenomenology in various environments. Since the efficacy of radar-based subsurface sensing depends strongly on the soil properties, we perform a parametric study of the dependence of such on the target RCS and on possible land-mine resonances.

A high-resolution ground penetrating radar system was designed to help define the optimal radar parameters needed for the efficient standoff detection of buried and surface- laid antitank mines. The design requirements call for a forward-looking GPR capable of detecting antitank miens in a 5 to 8 meter wide swath, 7 to 60 meters in front of a mobile platform. The system has a resolution goal of 15 cm both in range and azimuth. The range and azimuthal resolutions are achieved by using a 2.7 GHz bandwidth and a 4 meter synthetic aperture, respectively. The system uses a fully coherent homodyne stepped-frequency approach with a modulation scheme that produces range dependent power gain to partially offset range losses. Transmit power of 1 to 10 W is available over the entire band, and a large effective dynamic range was built into the receiver. The antennas are mounted on separate transmit and receive computer-controlled high-precision linear drives for creating the synthetic aperture. A data scan entails stepping through all the frequencies, polarizations, and antenna positions before the van is driven forward for the next scan. Preliminary data, the resulting images, and preliminary work on automatic target detection will be presented.

We consider the problem of detecting and localizing buried landmines from a ground penetrating radar (GPR) array. A simplified, ray-optics-based physical mode for time domain GPR returns is presented. Under this model in the absence of an object from the field of view of the array, there exist well defined symmetries in the structure of the radar returns. In particular, for a bistatic system composed of one length M transmit array and a second length M array of receivers, we identify M subsets of signals from the M² transmit array and a second length M array of receivers, we identify M subsets of signals from the M² total transmitter/receiver paris such that the mean value of the signals within each subset should be the same when no object is present. This relationship then forms the basis for a modified Hotelling's T²-test to detect the presence of objects when there is noise in the signal. Simulation results demonstrate the validity of these methods.

Algorithms for detecting land mines using the GEO-CENTERS Energy Focusing Ground Penetrating Radar (EFGPR) are presented. Key elements of the system include normalization, down- and cross-track feature extraction, fuzzy set membership based confidence assignment, and false alarm testing via transition and number of hyperbolae features. The system has been implemented in real-time in the GEO- CENTERS Vehicle Mounted Mine Detection System and can be used to perform real-time land mine detection or to analyze data stored to disk. Results are presented on calibration lane data from Aberdeen Proving Grounds, Maryland and the Energetic Materials Research and Testing Center, New Mexico in the summer of 1998.

Algorithms for detecting land mines using the GEO-CENTERS, Vehicle-Based Energy Focusing Ground Penetrating Radar (EFGPR) are presented. The algorithms rely on the use of hidden Markov models to model the time-varying signatures produces by the interaction of the EFGPR and the landmines as the vehicle moves. Results are presented on isolated land mine signatures isolated from calibration lane data from Fort A.P. Hill in the summer of 1998.

In this paper a signal processing technique is developed to reduce clutter due to ground bounce in ground penetrating radar (GPR) measurements. This technique is especially useful when a GPR is used to detect subsurface anti- personnel mines. The GPR clutter is modeled using a simple parametric model. Buried mine and clutter contributions are separated through a set of iterative subspace projections. The algorithm outperforms existing clutter reduction approaches and also yields target feature which are useful for detection and identification of these mines. It is shown that the proposed technique increases the probability of detection and at the same time significantly decreases the false alarm rates when combined with a matched filter detector.

A team of US Army Corps of Engineers, Omaha District and Engineering and Support Center, Huntsville, JPL, Stanford Research Institute (SRI), and Montgomery Watson is currently in the process of planning and conducting the largest ever survey at the Former Buckley Field, in Colorado, by using SRI airborne, ground penetrating, SAR. The purpose of this survey is the detection of surface and subsurface Unexploded Ordnance (UXO) and in a broader sense the site characterization for identification of contaminated as well as clear areas. In preparation for such a large-scale survey, JPL has been developing advanced algorithms and a high-performance testbed for processing of massive amount of expected SAR data from this site. Two key requirements of this project are the accuracy and speed of SAR data processing. The first key feature of this testbed is a large degree of automation and maximum degree of the need for human perception in the processing to achieve an acceptable processing rate of several hundred acres per day. For accuracy UXO detection, novel algorithms have been developed and implemented. These algorithms analyze dual polarized SAR data. They are based on the correlation of HH and VV SAR data and involve a rather large set of parameters for accurate detection of UXO. For each specific site, this set of parameters can be optimized by using ground truth data. In this paper, we discuss these algorithms and their successful application for detection of surface and subsurface anti-tank mines by using a data set from Yuma Proving Ground, AZ, acquired by SRI SAR.

This paper presents the current program of activities of the Defence Evaluation and Research Agency in the area of Airborne Minefield Detection. An overview of the program status is given along with result obtained from sensor trials.

In our previous work, we have shown theoretically that model-based Bayesian approach to the detection of landmines affords significant performance gains over standard thresholding techniques. These performance gains hold for both time- and frequency-domain electromagnetic induction (EMI) sensors. Our methodology merges physical models of the evoked target response with a probabilistic description of the clutter. Under a specific set of assumptions, our technique provides both an optimal detection algorithm and performance evaluation measures expressed as probability of detection and probability of false alarm. This approach also provides a formal framework for incorporating target and/or environmental uncertainties into the processing algorithms. The significant performance improvements observed theoretically have been verified on both time-domain and frequency-domain EMI data collected in the field. In this paper, we review our previous theoretical work, and we use actual data collected in the field to illustrate the improvement obtained by appropriately accounting for environmental uncertainties. We present new results in which a suboptimal processor provides nearly identical performance to that of the optimal processor but with much greater computational efficiency. We also present result that indicate that such an approach can be applied successfully to ground penetrating radar data. Specifically, we consider data taken by the BRTRC/Wichmannn system. In addition to processing the data from each type of sensor individually, as well as the combination of sensor, will be discussed.

This paper provides an overview of performance results achieved by the EG and G version of the Vehicle Mounted Mine Detector at the Army - directed Advanced Technology Demonstration (ATD) conducted in the summer of 1998, and to post test analyses conducted since those test. The VMMD is described as it was configured for the ATD. The ATD is then described in terms of sites, lanes, targets, their emplacement and environments. A summary of result is presented, and comparisons made between those factors perceived as most relevant to detection and false alarms. Conclusions and recommendations for future work follow.

This paper reports on selection of suitable sensor fusion techniques for use in hand carried and vehicle mounted mine detection systems for HOM-2000. Various techniques like Bayes, Dempster-Shafer, voting logic and neural networks are considered. Guidelines for selection of an appropriate fusion level are formulated. Experimental data are presented to illustrate the advantages that can be obtained through the use of sensor fusion.

A robust method of performing information fusion in processing ground penetrating radar (GPR) sensor data in landmine detection will be described. The method involves running multiple automatic target recognition algorithms (ATRs) in parallel on the GPR data. The outputs from each of the ATRs are spatially correlated and a feature set for each potential radar target is automatically generated. The feature set is provided as input to Mamdani style fuzzy inference systems. The fuzzy inference systems' output is a mine confidence value. The major advantage of this technique is that it provides consistent mine detection performance independent of road type, GPR hardware settings, and ATR setup parameters. This paper will first describe the individual ATRs and the process of spatially correlating target reports and generating a feature set. This will be followed by a description of the fuzzy inference system used for target classification. THe paper will conclude with test result from various Fort AP Hill calibration mine lanes.

In this paper, we develop a statistical detection system exploiting sensor fusion for the detection of plastic A/P miens. We design and test the system using data from Monte Carlo electromagnetic induction (EMI) and ground penetrating radar (GPR) simulations. We include the effects of both random soil surface variability and sensor noise. In spite of the presence of a rough surface, we can obtain good result fusing EMI and GPR data using a statistical approach in a simple clutter environment. More generally, we develop a framework for simulation and testing of sensor configurations and sensor fusion approaches for landmine and unexploded ordinance detection systems. Exploiting accurate electromagnetic simulation, we develop a controlled environment for testing sensor fusion concepts, from varied sensor arrangements to detection algorithms, In this environment, we can examine the effect of changing mine structure, soil parameters, and sensor geometry on the sensor fusion problem. We can then generalize these results to produce mine detectors robust to real-world variations.

Over the past year with the support of the Army Humanitarian Demining MURI, Aerodyne has substantially moved forward in developing and demonstrating the value of an affordable and fieldworthy IR polarimetric hyperspectral imager for inclusion in multisensor demining. Such technology promises powerful clutter suppression and enhancement of man made objects, particularly applicable to the reliable detection of scatterable mines, especially plastics, and any UXO that are partially exposed. We have achieved the first 3 steps of a 4 step, controlled-risk program defined as follows: (1) LWIR Spectral Polarimeter to demonstrate the effectiveness of combined polarimetric and hyperspectral discrimination capabilities in observations on static scenes; (2) LWIR Uncooled FPA Imaging Polarimeter to verify the sensitivity of an affordable Uncooled FPA in a broadband configuration against static scenes; (3) Multispectral IMaging Polarimeter to quantify clutter rejection performance improvements to be realized from multispectral imaging polarimetry; and (4) IR Polarimetric Hyperspectral Imager designed with optimal spatial and spectral resolution and sufficient throughput to achieve the reliable performance required in surface mine and UXO detection applications. We present results for Steps 1 and 2, and initial result for Step 3 from the ongoing demonstrations in simulated surface mine detection.

We present and compare methods for pre-detection and post- detection fusion of multi-sensor data. This study emphasis methods suitable for data that are non-commensurate and sampled at non-coincident points. Decision-level fusion is most convenient for such data, but this approach is sub- optimal in principle, since targets not detected by all sensor will not achieve the maximum benefits of fusion. A novel feature-level fusion algorithm for these conditions is described. The optimal forms of both decision-level and feature-level fusion are described, and some approximations are reviewed. Preliminary result for these two fusion techniques are presented for experimental data acquired by a metal detector, a ground-penetrating radar, and an IR camera.

The aim of this paper is to develop a framework for multi- sensor data fusion for the detection and identification of anti-personnel mines as a part of humanitarian demining project. A two-stage hybrid architecture is proposed to integrate non-homogeneous and dis-similar sensor data from various sensor be in developed as a part of the project. The first stage is used to extract significant information from individual sensor data. Self-organizing neural networks are used to define natural and significant clusters embedded in the sensor data. In this regard two popular self-organizing NN architectures of ART2 and DigNet are studied. The second fusion stage is used to integrate this local sensor information into a global decision. The global decision could be binary as in mine/no-mine decision set, or it could be more complex where identification of the underground mine may be involved. For the present paper, reliable data from different sensor was not available. Extracting different shape feature like moment invariants and Fourier descriptors simulates dis-similar sensor data for simulated shapes. Some results for the performance of the clustering algorithms and the fusion architecture are presented.

In this paper the multi sensor fusion results obtained within the European research project GEODE are presented. The layout of the test lane and the individual sensors used are described. The implementation of the SCOOP algorithm improves the ROC curves, as the false alarm surface and the number of false alarms both are taken into account. The confidence grids, as produced by the sensor manufacturers, of the sensors are used as input for the different sensor fusion methods implemented. The multisensor fusion methods implemented are Bayes, Dempster-Shafer, fuzzy probabilities and rules. The mapping of the confidence grids to the input parameters for fusion methods is an important step. Due to limited amount of the available data the entire test lane is used for training and evaluation. All four sensor fusion methods provide better detection results than the individual sensors.

An approach to describe gravitational effect of particles based on the statistical theory has been proposed. In this model, the time being fixed, bodies are shown to have fuzzy contours and are represented by spheroidal forms. The tension, force, potential and energy of the gravitating spheroidal body have been determined to be of probability character. It has been pointed out that a spheroidal body has a clearly outlined form if the potential energy of gravitational interaction of its particles is sufficiently great and the body's mass itself is relatively small. The statistical model being consistent with the Einstein general relativity has been shown.

This paper introduces an advanced robotic system which combines both a sophisticated super resolving detection algorithm and a sate of the art magnetic sensor used for a real time unexploded ordinance detection and mapping.

The Defence Evaluation and Research Agency (DERA) has a requirement for an IRPC system to detect surface laid and buried anti-tank landmines in support of Phase 2 of the REmote Minefield Detection System Technology Demonstration Program. Nichols Research Corporation is currently under contact to DERA to design and fabricate the IRPC system for integration in the REMIDS TDP. The IRPC is a Stokes 4-vector IR camera system designed to operate form a static tower, a moving elevated surface platform or a moving airborne platform and will be used to demonstrate the usefulness of passive IR polarimetry for mine and minefield detection. DERA will use the IRPC system to investigate the feasibility of using polarimetric techniques to detect buried and surface laid mines from an airborne platform when operated in conjunction with an ultra wideband SAR.