The Ground Standoff Mine Detection System (GSTAMIDS) is now in the Engineering, Manufacturing and Development (EMD) Block 0 phase for USA CECOM. The Mine Detection Subsystem (MDS) presently utilizes three different sensor technologies to detect buried anti-tank (AT) land mines; Ground Penetrating Radar (GPR), Pulsed Magnetic Induction (PMI), and passive infrared (IR). The GSTAMIDS hardware and software architectures are designed so that other technologies can readily be incorporated when and if they prove viable. Each sensor suite is designed to detect the buried mines and to discriminate against various clutter and background objects. Sensor data fusion of the outputs of the individual sensor suites then enhances the detection probability while reducing the false alarm rate from clutter objects. The metal detector is an essential tool for buried mine detection, as metal land mines still account for a large percentage of land mines. Technologies such as nuclear quadrupole resonance (NQR or QR) are presently being developed to detect or confirm the presence of explosive material in buried land mines, particularly the so-called plastic mines; unfortunately, the radio frequency signals required cannot penetrate into a metal land mine. The limitation of the metal detector is not in detection of the metal mines, but in the additional detection of metal clutter. A metal detector has been developed using singular value decomposition (SVD) extraction techniques to discriminate the mines from the clutter, thereby greatly reducing false alarm rates. This mine detector is designed to characterize the impulse response function of the metal objects, based on a parametric three-pole model of the response, and to use pattern recognition to determine the match of the responses to known mines. In addition to discrimination against clutter, the system can also generally tell one mine type from another. This paper describes the PMI sensor suite hardware and its physical incorporation into the GSTAMIDS sensor modules. This is a time-domain, transient signal metal detector that gives target signal response information of a different nature than that from more conventional continuous-wave (CW) metal detectors. The magnetic design of the GSTAMIDS PMI has very broad-band radiation properties that allow for the required transient eddy current responses in the metallic targets. The design of this detector is unique in that it allows processing of the received signals from targets to begin at the very start of the eddy current decays (t = 0). This then gives the ability to measure and quantify up to three decay terms in the target response, which features unambiguously identify the particular threat target. The results of the data processing algorithms that are used to extract the features used for mine detection are included herein to more clearly show the mine signals.

The International Pilot Project for Technology Co-operation in landmine detection is a multinational effort with participation from government agencies and research institutes from Canada, the USA, the UK, the Netherlands and the European Commission. One goal of the pilot project is to provide technical information on a number of commercial metal/landmine detectors to sponsors and end users of such technology, to help them make informed decisions about equipment selection in humanitarian demining. To this end, a series of laboratory and field tests have been conducted by the project team at various locations. A significant component of these tests was the tests conducted in a controlled laboratory environment at the Defence Research Establishment Suffield, Alberta, Canada. These tests focused on a detector's ability to detect objects in air (also referred to as its in-air sensitivity) and assessed how much this sensitivity would be affected by various parameters that model some real-world conditions, such as the presence of moisture, variation of sweep speed, electronic drift and so on. While a detector's ability to detect objects in air does not always indicate its ability to detect objects buried in the ground, such controlled tests are very useful in comparing certain basic performance factors of the electronics of a given detector and in understanding a detector's performance in the field. Six different in-air tests were conducted on 29 models of commercial-off-the-shelf metal detectors from a number of manufacturers. The paper discusses the methodology and results of these tests.

Nowadays, metal detectors are hardly used to extract target parameters from measurements and often only the indication of a metallic object's presence seems to be sufficient. In this article we present an analysis of the sensitivity of metal detectors with the goal to estimate characteristic parameters of a hidden object. Processed data from practical measurements will be compared to theoretical observations.

Wideband (approximately 0-2000 MHz) electromagnetic scattering from mines and unexploded ordnance (UXO) is considered using a method of moments (MoM) analysis for general targets in a layered medium, with the lossy, dispersive layers representing the typical layered character of many soils. We examine wave phenomenology to identify target features that may be exploited in target detection algorithms. Accordingly, we examine the target signature in both the frequency and time domain, as well as in the SAR image domain. For this study, we restrict ourselves to a single UXO target, i.e., a 155 mm shell, an arbitrary cylindrical metal mine, and a cylindrical plastic mine (TM- 62P3 anti-tank mine). Results show that for the UXO, a strong correlation exists between the target signature in the frequency domain and the target orientation. For the rotationally symmetric mines, despite their small size, results show that these targets contain isolated scattering centers due to the wide bandwidth of the incident pulse. Further, these scattering centers can be used to deconstruct the backscattered time waveform into closely spaced simple wave objects that leads to an observable interference pattern in the frequency domain. This relatively simple scattering deconstruction lends itself well to target detection and discrimination.

This report presents a summary of signal strength testing conducted with the metal detector (MD) subsystem of the Mine H/K (hunter/killer) vehicular mine detection system. An overview of the operational characteristics of the MD subsystem, the VMV16, is provided. Tests are described that assess the variation in sensitivity across the MD coil array. Absolute sensitivity measurements of the MD array are also presented. Results presented show that the array has sufficient sensitivity to detect low metal (LM) mines provided the mines are not located further than 3.5 inches from the plane of array. Laboratory experiments indicate that saturation and a limited temporal sampling window severely restrict any opportunity for discrimination based on eddy current decay predictions/comparisions.

This paper presents wide bandwidth, time decay responses from low metal content (LMC) mines, LMC mine simulates, and ground voids. Measurements were collected both in the laboratory and in the field. The target time decay responses were measured with the Johns Hopkins University Applied Physics Laboratory developed Electromagnetic Target Discriminator (ETD) sensor developed for the US Army CECOM/NVSED. The ETD sensor has demonstrated the ability to measure metal target decay times starting approximately 3 to 5)mus after the transmitter current is turned off and metal target decay time constants as short as 1.4)mus.

This paper describes a prototype electromagnetic induction (EMI) sensor system designed specifically to measure the horizontal component of a metal target's eddy current time decay signature. Instead of creating a vertical magnetic field from a horizontal loop transmitter configuration used by most EMI metal detectors, the prototype transmitter geometry has been designed especially for creating a horizontal magneti field (HMF). One of the potential advantages of the HMF sensor is the relatively uniform magnetic field that is created over a large volume. A second potential advantage is that, compared to a conventional loop antenna, the magnetic field intensity falls off slowly with distance from the plane of the sensor. These two advantages potentially make the HMF sensor well suited for detection and classification of metal targets buried deeply in the ground (e.b., unexploded ordnance, UXO) or from a vehicle-mounted mine detector sensor. Preliminary modeling of the antenna and laboratory data from a time-domain version of the HMF sensor are presented.

In keeping with the Navy's policy to remove humans from harms way, the Autonomous Underwater Vehicle (AUV) is replacing human divers for many missions. The Advanced Marine Systems Lab at Florida Atlantic University (FAU) has developed a small, magnetically friendly, modular plastic AUV called Morpheus designed for coastal applications and especially suited for very shallow water (VSW) mine reconnaissance. Currently employed sensor technologies on AUVs have certain deficiencies and limitations when used across the wide gamut of naval targets and environments, and a strong requirement exists for a sensor or sensors to fill these niches. The Real-time Tracking Gradiometer (RTG) selected for this integration is truly such a niche sensor because its capabilities are not degraded by media interfaces or environmental conditions. It is an experimental prototype fluxgate magnetometer array developed by Quantum Magnetics for the Coastal Systems Station (CSS) and was designed to be man portable and self contained. While limited by physics in detection range, it is capable of detecting ferrous targets under the worst environmental conditions, even when the target is buried. While not having the range of sonar, the RTG does not respond to the false alarms that are indicated by sonar, and since it is capable of also providing range and bearing information, it provides an invaluable niche filling classification tool. The placing of any magnetic sensing system on a conventional AUV is a non-trivial problem. The standard AUV is designed around materials and components that were selected to maximize performance without regard to the magnetic properties of the materials used in its fabrication. To minimize the degradation of sensor performance caused by the platform, several steps must be taken. These include; the substitution of nonferrous components for ferrous, maximizing the separation between the sensor and magnetic field sources, minimizing current loops and using auxiliary current and field sensors capable of generating noise canceling signals. To maximize utility, the magnetic sensor systems should also provide range, bearing and magnetic target strength. While all data and results contained in this paper have been obtained with land-based testing, they are easily adapted to the underwater environment of the AUV. The RTG was recently attached to the Morpheus, and data collected with the unmodified Morpheus powered and undergoing simulated sea motion table. These tests indicate that integration, while not trivial, is indeed feasible, and work is continuing toward mounting the sensor internal to the AUV and implementing the required noise mitigation solutions.

This paper briefly reviews the fundamental operating characteristics of the AN/PSS-12 and describes modifications to the system aimed toward developing the capability to distinguish buried low-metal landmines from buried metallic clutter. Improvements were implemented to three key areas including the AN/PSS-12 hardware, method of data collection at Fort A.P.Hill, VA, and algorithm design. The improvements to the AN/PSS-12 hardware yield higher system bandwidths resulting in the ability to extract the fast decay rates associated with small metallic objects. The improvements in data collection involve exciting and measuring the response of a buried object along its three principle cardinal axes resulting in an increase of characteristic target information that can be used to further separate mine responses from clutter responses. The increase in characteristic target information yields five target parameters that characterize each of the eleven different mine types in the JUXOCO grid. A generalized likelihood ratio test (GLRT) is developed that incorporates the five target parameters. The algorithm, using the additional target information, results in an increase in landmine discrimination performance presented in a receiver operator characteristic (ROC) curve.

Methods for classifying objects based on spatially sampled electromagnetic induction data taken in the time or frequency domain are developed and analyzed. To deal with nuisance parameters associated with the position of the object relative to the sensor as well as the object orientation a computationally tractable physical model explicit in these unknowns is developed. The model is also parameterized by a collection of decay constants (or equivalently Laplace-plane poles) whose values in theory are independent of object position and orientation. These poles can be used as features for classification. The overall algorithm consists of two stages. First we estimate the values of the unknown parameters and then we do classification. Three classification schemes are examined. The first is based on data residuals. The second uses estimated pole values. The third is a blending of the first two. Preliminary results on synthetic data indicate the robustness of the pole estimates as features for classification and point toward the need for further analytical as well as experimental evaluation of the proposed methods.

A new algorithm for automatic detection of mines using a vehicle-mounted metal detector is detailed and demonstrated in this paper. The sensor is the Vallon VMV 16 vehicle- mounted array on the Mine Hunter/Killer (MH/K) system; data were collected at a prepared test site. The resulting data have two notable characteristics. First, the sensor outputs a linear combination of the integrated time decay response over separate time windows, and not the time decay response itself. Second, the data are sampled in two dimensions, along the 16 coil array and along the direction of travel of the vehicle. Thus 2-dimensional spatial processing techniques are used, treating the sensor data as a 2-D array of pixels. Our algorithm also uses adaptive methods to enable detection of both high and low metal content mines. Mine detection results using this algorithm are presented and analyzed for the first time. In particular, the ability to detect mines with low metal content in the presence of noise will be quantified and discussed.

This paper reports the results of the performance assessment of an Electrical Impedance Tomography detector (EIT) for 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 electrical conductivity perturbations underneath the array. In the course of this work, an EIT instrument was built and field evaluated. The array is made of 64 stainless steel stimulating and recording electrodes arranged in an 8x8 grid. Specialized electronics has been built to control the electrode current stimulations and potential measurements. The detection algorithm is tuned to objects of a given size and shape to reduce the false alarm rate. The main mechanical, electronic and algorithm components of the detector will be presented. The EIT detector was originally designed in view of evaluating its potential as a confirmatory detector of AT mines. To that end, mine-like objects representative of some AT mines were used. Results of preliminary field evaluations are presented. The detection capabilities and limitations of mine-like objects are discussed.

Electromagnetic emissions from electronics associated with explosives is a potential detection modality, both passive listening, and stimulated RF emissions. However, the parasitic paths by which energy is coupled off a printed circuit board from an active device, and by which external energy can be coupled onto the board and to a device, must be identified and characterized. One such noise-coupling path is identified in this work, and a modeling approach demonstrated. In particular, coupling of noise from the DC power bus of a multilayer printed circuit board that uses entire metal layers for power and ground, and an I/O line that transitions through the DC power planes is investigated.

The UK Remote Minefield Detection System (REMIDS) Technology Demonstration Programme (TDP) aims to demonstrate the applicability of emerging technologies in an integrated system for detecting mined areas from an airborne platform. The multi-sensor demonstrator is being managed by DERA on behalf of the UK MoD with an unmanned air vehicle (UAV) as the intended platform. The TDP is nearing the end of a five year programme that has developed a novel ultra wide band synthetic aperture radar (UWB SAR) and a MWIR time-parallel polarimetric IR camera (IRPC). A ground station processor has also been developed to fuse sensor outputs through a variety of detection and identification algorithms. An extensive series of data collection trials and analysis activities has been performed. This paper reports on the development by DERA of the algorithms required both to pre- process raw data from the IRPC into a useable form, and to discriminate mines from a variety of backgrounds. These algorithms are being integrated into ground station processing that will combine data from both sensors to identify mine areas and accurately delineate their boundaries. The top-level process architecture for this ground station, and how the IPRC fits within it, are also described in this paper.

It has long been recognized that surface-laid land mines and other man-made objects tend to have different polarization characteristics than natural materials. This fact has been used to advantage in a number of mine detecting sensors developed over the last two decades. In this work we present the theoretical basis for this polarization dependence. The theory of scattering from randomly rough surfaces is employed to develop a model for scattering and emission from mines and natural surfaces. The emissivity seen by both polarized and unpolarized sensors is studied for smooth and rough surfaces. The polarized and unpolarized emissivities of rough surfaces are modeled using the solution of the reciprocal active scattering problem via the second order small perturbation method/small slope approximation(SPM/SSA). The theory is used to determine the most suitable angle for passive polarimetric IR detection of surface mines.

Linear polarization of Thermal InfraRed (TIR) radiation occurs whenever radiation is reflected or emitted from a smooth surface (such as the top of a landmine) and observed from a grazing angle. The background (soil and vegetation) is generally much rougher and therefore has less pronounced linear polarized radiation. This difference in polarization can be used to enhanced detection of land mines using TIR cameras. A measurement setup is constructed for measurement of polarized TIR images. This setup contains a rotating polarization filter which rotates synchronously with the frame sync of the camera. Either a Long wave InfraRed (LWIR) or a Mid Wave InfaRed (MWIR) camera can be mounted behind the rotating polarization filter. The synchronisation allows a sequence of images to be taken with a predefined constant angle of rotation between the images. Out of this image sequence three independent Stokes images are calculated, consisting of the unpolarized part, the vertical/horizontal polarizations and the two diagonal polarizations. An initial model is developed that describes the polarization due to reflection of and emission from a smooth surface. This model predicts the linear polarization for a landmine `illuminated' by a source that is either hotter or cooler than the surface of the landmine. The measurement setup is used indoors to validate the model. The measurements agree well with the model predictions.

The technical issues of a standoff electro-optic tripwire detector are discussed. Significant advances in short-wave infrared (SWIR) laser diodes and InGaAs detector technologies have made it possible for the demonstration of a passive and active eyesafe (1.5 micron) laser illuminated tripwire (ELIT) detector. The demonstrated system utilizes COTS laser diodes and cameras. The Hough Transform was used for the detection of tripwires in images. System trade-offs are discussed and images are shown.

Predicting the thermal signature of a buried land mine requires modeling the complicated inhomogeneous environment and the structurally complex mine. It is useful, both in checking such models and in making rough calculations of expected signatures, to have an accurate, easily computed solution for a relatively simple geometry. In this paper, a reference solution is presented for the integral equation that governs the temperature distribution. Our solution procedure uses the method of weighted residuals. The problem comprises a homogeneous cylindrical body (the mine model) buried in an infinite homogeneous half space (the soil model) with a planar interface. Using periodic boundary conditions in time at the planar interface, the temperature distribution in the lower-half space is expanded in a Fourier series. A volume integral equation for the Fourier series coefficients is obtained via Green's second identity. The Green's function for the Fourier coefficients is derived and reduced to a computationally efficient form. The integral equation is reduced to a matrix equation, which is then solved for the unknown temperature distribution. The integral equation solution is compared with a finite element model.

A parametric model for the infrared signature caused by a buried land mine is presented. Further, two ways of modeling the colored background noise, is proposed. In the first, it is assumed the noise can be approximated by an autoregressive process, while in the second, the statistics of the noise is described using recent development in texture modeling, the so called FRAME method. Given an a priori distribution of the mine parameters in combination with a trained noise distribution, a Bayesian detector is derived. Experiments indicate that significant gains in performance can be achieved as compared to the standard detector used, which correlates the infrared image with the known mine shape and thresholds the square of the output.

Many aspects of a buried mine's thermal IR signature can be predicted through physical models, and insight provided by such models can lead to better detection. Several techniques for exploiting this information are described. The first approach involves ML estimation of model parameters and followed by classification of those parameters. We show that this approach is related to an approximate evaluation of an integral over the parameters that arises in a Bayesian formulation. This technique is compared with a generalized likelihood ratio test (GLRT) and with computationally efficient, model-free approaches, in which soil temperature data are classified directly. The benefit of using the temporal information is also investigated. Algorithm performance is illustrated using broadband IR imagery of buried mines acquired over a 24 hour period. It is found that the detection performance at a suitably selected time is comparable to the performance achieved by processing all times. The performance of the GLRT, for which detection is based only on the residual error, is inferior to a classifier using the parameters.

Infrared imagers are capable of the detection of surface laid mines. Several sensor fused land mine detection systems make use of metal detectors, ground penetrating radar and infrared imagers. Infrared detection systems are sensitive to apparent temperature contrasts and their detection capabilities are inversely proportional to the amount of background clutter generated by local surface non- uniformities. This may result in spurious detections, or even cancellation of true detections in a post classification process. Sub-surface mines can be detected when buried not too deeply. Furthermore, soil type and soil water content will influence the detection result. For this reason experiments in various soil types, including vegetation, and soil circumstances are essential for understanding and improving the infrared detection capabilities. We have performed outdoor experiments with different types of soil and weather conditions. Several examples are described and analyzed. Data analysis shows the strong correlation of apparent temperature with thermocouple gradients and solar energy, as well as a correlation of local standard deviation with these parameters. Model based temperature contrasts are predicted for several mines in sandy soils, and these are compared with infrared imaging apparent temperature measurements and thermocouple data. The comparison results are quite good but also show the complexity of the thermal infrared data, in particular due to infrared clutter, diurnal variations, and sky reflectance contributions. Model predictions are made for the application of active heating methods also. Limitations of the model and potential future expansions based on evaluation of experiments are discussed. We discuss the potential use of modeling for thermal infrared detection and sensor fusion applications.

Vehicle-mounted detection (VMMD) systems currently under development depend heavily on ground penetrating radar (GPR) for detection of buried mines. GPR is supported by electro- magnetic induction (EMI) sensors for detection of metal- cased mines, and by forward-looking infrared (FLIR) imagers for detection of surface mines. However, GPR is often subject to unacceptably high false alarm rates due to the presence of naturally occurring anomalies such as surface irregularities and spatial discontinuities below the surface (e.g., variation in soil compaction or moisture content). Thus, there is a need to rely more heavily on EMI and infrared sensors for detection of buried metal and plastic- cased mines. The conventional approach to IR detection is a roof-mounted forward-looking infrared imager (FLIR), viewing a trapezoidal region of front of the vehicle. This approach is subject to several important limitations. First, boresight tolerance, vehicle attitude variation, and terrain undulation limit registration of IR detections with those of the down-looking GPR and EMI sensors. Second, the images of the mine surface thermal effects are foreshortened by the oblique viewing angle and subject to obscuration and distortion by surface heigh variation. Third, the redundant nature of the FLIR framing process makes the data rate unnecessarily high. This paper will discuss an alternate approach, based on an array of discrete, down-looking infrared detectors that are co-located with the GPR antennas and EMI coils. The advantages of this approach, design of a breadboard sub- array, and preliminary test results will be described.

The thermal signatures of surface and buried land mines vary widely with time of day, weather, soil type, soil moisture content, and mine burial depth. There have been recent advances in modeling these effects, but until these models are fully developed and validated we will continue to rely on measured data. This paper witll present signatures in the medium-and long-wavelength infrared (MWIR and LWIR) spectral bands for surface and buried M19, M15 and VS1.6 anti-tank mines at two locations, a temperate site and an arid site. We will show that the apparent contrast of these landmines is substantial throughout the diurnal cycle except during thermal crossover periods after sunrise and sunset. Our results show that the mine signatures are well above sensor noise and that further improvements in sensitivity or resolution are not required. The paper will also present LWIR images of landmines buried in dirt and gravel road environments taken on a cold winter day and discuss the intriguing and unexpected differences observed between the images of landmines buried in dirt and gravel.

This paper identifies the optimal bands in the 3 to 5 micron region for surface mines for the lightweight airborne detection (LAMBD) sysem. Specifically, this paper focuses on the analysis to identify the optimum bands in the 3 to 5 micron region which can be used in a filter wheel implementation in an attempt to add the multi-band capabilities to enhance the detection performance for various background, times of day, while lessening the weight/size/power problem in a lightweight airborne mine detection (LAMBD) system. The analysis includes the hyperspectral signatures of various mines and backgrounds, the contrast between mines and backgrounds for various spectral regions, different times of day, and differential heights simulating sensor airborne scenarios. This paper uses data collected with a Design and Prototypes Fourier Transform Infrared (FTIR) Spectrometer (D&P) mounted 3m above the ground on the Mobile Sensor Platform (MSP) at a temperate test site. The total number of signatures used was 1000 for mines and 1000 for backgrounds.

This paper includes analysis/assessment and development of detection algorithms: (1) the assessment of detectability of surface mines using the RX algorithm implementation, which in turn, provides a first look at the limitation of the algorithm for suitable real-time implementation; (2) the development of the adaptive real-time mine detection algorithm (ARMD) based on statistical analysis of the data. The statistical analysis includes the class distribution between mines and background, the underlying distribution for mines and background based on the quantile-quantile plot. The paper also compares the quantitative performance of probability of detection (Pd) and false alarm rates (FAR) for different detection techniques for different background and mine types. This paper also presents the minefield probability of detection versus minefield false alarm rate to gauge the minefield detection performance trade-off using: (1) only mine density; and (2) mine density with pattern. This paper also demonstrates the importance of the observables that offer the class separability between mine/target and background for automatic target detection/recognition applications. Detection algorithms with high computational capability are not the 'silver bullet' for automatic target detection/recognition as commonly believed. The art of ATR is the ability to be able to pinpoint the observables that distinguish mines and background. Once the observables offer the class separability between classes are established, any simple correlation method can deliver an acceptable performance (demonstrating that highly computational methods, indeed, are wasteful and unnecessary). This paper uses the multiband and broadband data collected with the AMBER (3.5-5)mum) camera in May 2000. This data set contains about 513 (approximately 1.1 in resolution) images covering three spectral regions: 3-5)mum, 3- 4.2)mua and 4.2-5)mua. The total number of mines and the area coverage for these three spectral regions are approximately 579, and 25200m², respectively. Note that each spectral region contains 171 images (of which 53 images contain mines with 131 large mines, and 62 small mines) covering about 8400m². Also note that to stimulate minefields, an image containing 3 mines with a straight line pattern is defined as a minefield opportunity.

Airborne or vehicle-borne sensor-based techniques are potentially attractive approaches for fast detection of landmine field towards efficient and safety humanitarian demining. The measured data in such cases has a rather low spatial resolution due to the altitude of measurements. Landmines in an IR image are either indicated directly due to their temperature difference to background, or indirectly by signs of digging or disturbance patterns. This paper proposes a novel method for automatically detecting landmine candidates by exploiting features associated with landmine point patterns. We describe a special type of multiresolution isotropic bandpass filter for detecting these landmine candidates and other man-made landmarks on the ground surface (which may be used for locating mine fields). The introduction of multiresolution to the detection fiber enables both good detectability and localization of landmine candidates. However, the method cannot distinguish landmine candidates from clutter sharing similar spatial patterns. Therefore, it is only suitable for detecting landmine fields, or candidates of landmines. For reliable individual mine detection, landmine discrimination methods should be subsequently applied. Experiments were performed on several images measured from vehicle-borne and airborne sensors over the test bed scenarios, and some results are included.

In this paper we revisit and enhance various algorithms for landmine detection, discrimination and recognition. Single- band and multi-band medium wave infrared (MWIR) image data from the May data collection (part of Lightweight Airborne multispectral Minefield Detection-Interim (LAMBD-I) program) is used for the analysis. In particular discrimination based on gray-scale moments is explored and its effectiveness is evaluated for surface mines under IR imaging using receiver operating characteristics (ROC) curves. The discriminatory power of gray-scale moments is compared with the RX and matched fiber based detectors for different terrain (e.g., grass,sand) and different mine types. The performance of single-band (broadband) MWIR imagery is compared with multi- band (short-pass and long-pass) MWIR images. Also direct multi-band detection is compared against fusion of multiple single-band responses. Gray-scale moment based target discrimination at potential target locations, identified by RX or matched fiber detectors, is shown to be computationally efficient and provides better performance in terms of reduced false alarms for comparable probability of detection. An evolutionary framework for minefield identification, in the presence of inevitable false targets, is also presented. Starting from the locations of individual mine targets and false alarms, the evolutionary algorithm is used to identify the underlying structure of the minefield. Issues in the detection of different minefield layouts are discussed. Preliminary implementation shows the promise of this approach in identification of a wide variety of minefields.

This is a follow-up work to analyze completely the detectability of the buried mines for the spectral regions extending from Visible/Near IR (VNIR) to Longwave IR (LWIR). Similar to previous work focusing on the VNIR region (1) this paper presents the quantitative detectability of the buried mines in the 3-5)mum and 8-12)mum regions. Specifically, this paper presents a statistical analysis for the buried mines in specified spectral regions for various soils and burial durations. As shown in the previous work (1) the performance based on the single hypothesis test using the distance measure was better than the intensity thresholding method. This paper focuses on only the distance measure method for statistical analysis of the data, and subsequently, classification to quantify the detectability of the buried mines in the 3 to 5 and 8 to 12 micron regions.

Algorithms are presented for detecting surface mines using multi-spectral data. The algorithms are demonstrated using visible and MWIR imagery collected at Fort A.P. Hill, VA under a variety of conditions. For imagery with a resolution of a few centimeters there is significant correlation in the clutter. Using a first-order Gauss Markov random field model for the clutter, an efficient pre-whitening filter is proposed. A significant improvement in detection is demonstrated as a result of this whitening. Further improvement in the detection of specific mine types is demonstrated by using a random signal model with a known covariance matrix. That approach leads to an estimator-correlator formulation, in which the random signature estimate is the output of a Wiener filter. It is suggested that by fusing the output of a bank of such filters one could improve detection of all mine types.

This paper discusses some preliminary results of the application of simple neural networks to the problem of landmine detection in IR imagery. A large data set of IR imagery (3-5)mum) collected as part of the U.S. Army's Lightweight Airborne Multispectral Minefield Detection (LAMD) system is used as the basis for the analysis. The data set is divided into training and testing subsets then used to train and evaluate the performance of some neural networks. A single neuron perceptron is trained and evaluated using two different types of input feature. The first type of input feature is based on the raw pixel values with typical maximum vale normalization. The second type is based on the unity vector of the inputs to take advantage of the angular displacement feature of the vector [1]. A more complex, multiple neuron network is also trained and evaluated. The results are compared to determine whether the increased computational complexity of the multiple neuron network is justified in terms of improved performance.

An unsupervised algorithm is proposed for land mine detection in heavily cluttered multispectral images, based on iterating hybrid multi-spectral morphological filters. The hybrid filter used in each iteration consists of a decorrelating linear transform coupled with a nonlinear morphological detection component. Targets, extracted from the first pass, are used to improve detection results of the subsequent iteration, by helping to update covariance estimates of relevant filter variables. The procedure is stopped after a predetermined number of iterations is reached. Current implementation addresses several weaknesses associated with previous versions of the hybrid morphological approach to land mine detection. Improvement in detection accuracy and speed, robustness with respect to clutter inhomogeneity, and a completely unsupervised operation are the main highlights of the proposed approach. Our experimental investigation reveals substantially superior detection performance and lower false alarm rates over previous schemes. Properties of a graphical user interface (GUI), based on the proposed iterative morphological detection scheme, are also discussed.

Supported by the Army Humanitarian Demining MURI, we most recently have focused on determining the unique strengths of passive IR sensing as a function of attribute diversity. Our initial findings identify polarimetric hyperspectral imaging.as a robust means to rapidly survey and detect partially exposed, non-metallic anti-personnel (AP) mines. We are investigating the discrimination gains expected from the combined polarimetric hyperspectral attributes under laboratory and field conditions. A principal components analysis of our earliest data indicates that this combination of attributes is about three times more effective in discriminating AP mines or mine-like materials than conventional hyperspectral sensing. In addition, we have uncovered a distinguishing spectral behavior of the Fresnel reflectance across resonance features that can be measured only by spectrally-resolved polarimetry.

The detection and identification of small surface targets with Electro-Optical sensors is seriously hampered by ground clutter, leading to false alarms and reduced detection probabilities. Active ground illumination can improve the detection performance of EO sensors compared to passive skylight illumination because of the knowledge of the illumination level and of its temporal stability. Sun and sky cannot provide this due to the weather variability. In addition multispectral sensors with carefully chosen spectral bands ranging from the visual into the near IR from 400-2500 nm wavelength can take benefit of a variety of cheap active light sources, ranging from lasers to Xenon or halogen lamps. Results are presented, obtained with a two- color laser scanner with one wavelength in the chlorophyll absorption dip. Another active scanner is described operating at 4 wavebands between 1400 and 2300 nm, using tungsten halogen lamps. Finally a simple TV camera was used with either a ste of narrow band spectral filters or polarization filters in front of the lamps. The targets consisted of an array of mixed objects, most of them real mines. The results how great promise in enhancing the detection and identification probabilities of EO sensors against small surface targets.

A ground vehicle-based, real-time, surface mine detection system, utilizing a Compact Airborne Spectrographic Image (casi), efficient mine detection algorithms, and real-time processing systems, was designed and tested. The combined real-time system was capable of 'learning' the in-situ spectra of various mines, thus providing a spectral library for the detection algorithms. The real-time processing of the casi data involved three steps. The first step was the radiometric correction of the raw data. The second step involved the application of the mine detection algorithms to the corrected data, referencing the spectral library. In the final step, the results of the real-time processes were stored and displayed, usually within a few frame times of the data acquisition. To the authors knowledge, this system represents the first hyperspectral imager to detect mines in real-time. This paper describes the generation of the in-situ mine spectral library, the collection of the scene data, the real-time processing of the scene data and the subsequent display and recording of the detection data. The limitation and expansion capabilities of the real-time system are discussed as well as various techniques that were implemented to achieve the goals. Planned future improvements that have been identified to create a more robust and higher performance, yet simpler processing systems are also discussed.

Humanitarian landmine detection and clearance is one of the most challenging, difficult and time-consuming tasks to be completed with existing technologies. Infrared (IR) Imagery has been used to find differences in heat transfer on the surface of the soil due to a buried object. In this paper, we will describe a method, Dual Frequency Microwave Enhanced Infrared Thermography (MEIT). Heating with microwaves instead of natural sunlight leads to a number of advantages, such as more efficient heating to enhance the thermal signature, and the ability to sense electromagnetic as well as thermal properties of the buried object. However, like other IR techniques, it is limited by surface roughness. Thus, the two frequency technique is used to minimize the clutter introduced by the rough, irregular surface of the ground itself, and vegetation covering the ground. The dependence of scattered waves on frequency is weak enough to makes this possible. A 2-D computational model of this method has been developed to simulate real-world landmine detection. Moreover, ROC (Receiver Operating Characteristic) curves are used to evaluate the performance of the system applying this method.

The objective of this study is to expand our exploration of the effects of the soil environment on landmine detection by investigating the influence of soil texture and water content on surface soil temperatures above antitank mines buried at 15 cm depth and away from it. Temperature distributions in July were calculated in six soil textures for the climatic conditions of Kuwait and Sarajevo. We evaluated the temperature distributions in typical dry and wet soil profiles. The simulated temperature differences varied from .22-.63 degree Celsius in Kuwait to .16-.37 in Sarajevo. Temperature differences were - with one exception - larger in the wet than in the dry soils which suggests that soil watering may help improve thermal signatures. A major finding of this study is that the thermal signature of an anti tank mine strongly depends on the complex interaction between soil texture, water content, and geographical location. It is very difficult to predict the exact time or even the approximate hour of the appearance or nonappearance of a thermal signature. Therefore, this modeling study indicates that the use of a thermal sensor in a real mine field for instantaneous mine detection carries a high risk. On the other hand if a given area can b monitored constantly with a thermal sensor for twelve hours or longer the thermal signature will be detected if the signal to noise ratio of the mine environment allows so. Field experiments are needed to validate the results of this modeling study.

Most mine detection sensors are affected by soil properties such as water content, temperature, electrical conductivity, and dielectric constant. The most important of these is water content since it directly influences the three other properties. The variability of these properties may be such that either potential landmine signatures are overshadowed or false alarms result. In this paper we present the results of field measurements in the Netherlands, Panama, and New Mexico on spatial variability of soil water content. We also discuss how the variability of soil water content affects the soils electrical conductivity and dielectric constant and the resulting response of a ground penetrating radar system.

Land mines are a major problem in many areas of the world. In spite of the fact that many different types of land mines sensors have been developed, the detection of non-metallic land mines remains very difficult. Most landmine detection sensors are affected by soil properties such as water content, temperature, electrical conductivity and dielectric constant. The most important of these is water content since it directly influences the three other properties. In this study, the ground penetrating radar and thermal IR sensors were used to identify non-metallic landmines in different soil and water content conditions.

Soil water content, dielectric constant, electrical conductivity, thermal conductivity and heat capacity affect the performance of many sensors and therefore the detection of landmines. The most important of these is water content since it directly influences the other properties. We measure soil water distribution around an antitank and an antipersonnel mine buried in a sand soil under varying moisture levels. After a period of two days with 38 mm precipitation the water content below the AP-mine increased from 0.07 to 0.12. The water content above and below the AT- mine increased from 0.09 to 0.17 and 0.09 to 0.13, respectively. Below the AT-mine it was 0.02 to 0.04 dryer than above the mine. The dielectric constant of the soil was estimated from the soil water content. After a dry period of two weeks the dielectric contrast between the AT-mine was approximately 2 (F/m). After a period of 38 mm precipitation the contrast between AT-mine and background increased to 6 (F/m). Differences in soil water distribution around the AT- mine caused a maximum dielectric contrast 4.5 (F/m) between background and mine. This effect was less apparent around the AP-mine. Differences in measured and simulated soil water distribution around an AT-mine urge for further investigation.

This paper presents a parametric study on the influence of target and soil properties, including depth of burial, on features extracted from ground penetrating radar (GPR) data. Understanding this influence is crucial for designing a classifier that uses these features for mine detection and identification. Two types of features have been studied. These are the Wigner-Ville distribution and geometric moments. Using a fast forward modeling program, synthetic GPR data were created for six buried objects, including two plastic minelike objects, for a wide range of soil properties and depths of burial. Both non-lossy and lossy soils were considered. From the computed data the above features were extracted and correlated with each other. The results show that the Wigner-Ville distribution performs much better in discriminating between objects than geometric moments. Furthermore, the features were found to be practically invariant to changes in mine-soil permittivity contrast and depth of burial provided that the soil is non-lossy. In the presence of losses, the GPR pulse is reshaped at the air-ground interface and as it propagates through the soil. As a result of the reshaping, the target response and hence the features can differ substantially from the non-lossy case.

A simple layered medium model for microwave thermal emission from a buried object shows that multiple frequency emission measurements can potentially provide an effective means for target detection. Object detection is obtained form a search for oscillatory features in multiple frequency brightness temperatures, which occur due to interference effects between the surface and buried object interfaces. Previous studies have considered simple homogeneous temperature and water content models of the soil medium, and show that oscillatory features versus frequency are not obtained in the absence of a target even with medium temperature or soil moisture variations. However, the more realistic case of non-constant temperature and water content versus depth was not considered in previous studies; these effects can potentially modify interference phenomena. In addition, subsurface objects have typically been modeled as layers whose horizontal dimensions are infinite; models including the effects of finite targets size are thus of interest.

Thermal IR signatures of buried land mines are affected by various environmental conditions as well as the mine's composition, size and burial geometry. In this work we present quantitative relations for the effect of those factors on the signature's peak contrast and apparent diameter. We begin with a review of the relevant phenomena and the underlying physics. A three-dimensional simulation tool developed by the authors is used to simulate signatures for the case of a static water distribution. We discuss efforts to validate the model using experimental data collected at Fort A.P. Hill, VA. Using this simulation tool a variety of factors are considered, including soil water content, soil sand content, wind speed, mine diameter and mine burial depth.

A particularly difficult problem in using GPR for land mine detection is the shallow depth at which many mines are buried. The scattered energy from a mine may be very small, whereas the specular scattering form the surface, as well as the direct coupling between transmitter and receiver, are often quite large. If the mine is shallow, the difference in propagation delay between these responses is very small, and resolving the mine form the specular surface reflection is difficult. Even if the radar bandwidth is sufficient to resolve closely spaced targets, detecting the smaller target is problematic.

Ground penetrating radar (GPR) generates a cross-sectional profile of the soil by transmitting electromagnetic waves that reflect back in a manner associated with the electrical properties and geometry of the objects buried underground. The responses of the reflected waves are processed using a variety of digital signal processing and image processing techniques. In this paper we compare an energy detector, matched filter, and a proposed Hough Transform approach. The results from each of the algorithms are compared using receiver operating characteristic (ROC) curves. Comparatively, the matched filter method has the lowest false alarm rate, however it is essentially providing a performance bound since for this analysis we derived the matching template from the data to be tested. Thus, in this case the Hough transform method may be more robust when the testing and training sets are separate, as it is inherently integrating over the uncertainty associated with the subsurface object detection problem.

Previous experiments with Hidden Markov Models showed that they could be used as an algorithm for landmine detection. In this paper we propose a basic adaptive algorithm for discrete Hidden Markov Model for landmine detection. The performance of adaptive HMM is investigated using GPR data from gathered during the Vehicle Mounted Mine Detection Advanced Technology Demonstrations. In both cases the adaptive HMM outperforms the baseline HMM model, trained offline using Baum-Welch algorithm.

We propose in this paper an improved correlation based GPR algorithm for a hand-held landmine detector. The previously proposed correlation based detector (CorrDet) generates prediction errors at different frequency bins and sums the prediction errors to generate a statistical quality for detection. The improved CorrDet applies weighting on the prediction errors before they are added to generate the detection value. The purpose of weighting is to emphasize the prediction error values in the frequency bins that are less affected by clutter. The weighting technique is found to be very effective, especially in detecting deep anti-tank mines and small anti-personnel mines. The performance improvement of the proposed algorithm is demonstrated using seven data sets obtained at different sites and soil conditions that contain over 2300 mine targets.

The Mine Hunter/Killer Close-In Detector (MH/K CID) uses Ground Penetrating Radar (GPR) as it's primary sensor. The GPR processor requires a sensitive detection algorithm to detect anomalies that may indicate the presence of a buried land mine. A general formula for a statistical detector is presented, consisting of a median filter to eliminate outliers, a local mean estimator using a Blackman window and a local covariance estimator. Advanced methods for robust estimation of the covariance matrix are presented and evaluated using data collected by the CID over buried land mines. This GPR detector is used as a preprocessor for image processing and mine classification algorithms that are used by a sensor fusion processor to determine when to activate the 'Killer' mechanism to neutralize the buried mine.

The problem of scattered and transmitted electromagnetic wave distortion by random rough ground surfaces can be reduced by using a lightweight dielectric matching layer. For mine detection applications, it is essential for this layer to be lightweight, low loss, readily conformable, and adaptable to different soil types. Arrays of metal-coated plastic spheres act as lossless artificial dielectrics with impedance determined by the volume packing fraction. By controlling the thickness of insulator surrounding each sphere, a close-packed array with the dielectric properties of soil can be created inside a compliant rolling bag that will conform to the rough surface of the ground. Since this artificial dielectric is matched to the soil, the ground surface interface is 'softened', without an abrupt transition from soil to air. Signals transmitted and received by GPR antennas immersed in the artificial dielectric within the bag will not be corrupted by ground surface clutter. Alternatively, an artificial dielectric layer on the ground with a planar air interface could be used to ensure that the surface reflection is a constant, well-calibrated signal. Computational models indicate complete removal of the ground clutter, even with occasional gaps between the artificial dielectric and the ground. Experimental studies with swept-frequency measurements and impulse GPR indicate that using this dielectric layer matching to a rough loamy soil ground surface is results in signals that are practically indistinguishable from those of an equivalent layer of the same type of soil.

The Steepest Descent Fats Multilevel Multiple Method (SDFMM) is used to analyze the distorting effects of random rough ground surfaces on scattered electromagnetic waves from buried TNT mines. The SDFMM method is an integral equation- based fast algorithm that is well suited for 2D penetrable rough surfaces in the frequency domain, and it is used to calculate the unknown surface currents on both the rough ground and the buried target as well. In this study all interactions between the rough interface and the buried target are taken into account. The scattered near field E- patterns of an incident Gaussian beam are calculated at different locations above the mean plane of the dielectric rough interface. The receiver locations are chosen to simulate GPR measurement protocols. The dimensions and burial depth of the TNT mine are smaller than the free space wavelength with material slightly different from the surrounding soil. The average and the standard deviation of the scattered fields for just the target are calculated and results showed that the presence of the rough interface tremendously distorts the target signal even for the small roughness parameters. Moreover, results showed the degradation of signal as the TNT mine is located away from incident beam. This knowledge can significantly contribute to inventing better sensing systems for less false alarm detection strategies.

The project motivating this paper is the deployment of a frequency independent antenna on the transceiver of a monostatic ground-penetrating radar used to detect mines. The design goal is that the radiation pattern and input impedance to nearly uniform over a band from 1 GHz to 5GHz if the antenna is partially immersed in a typical soil medium. The contemplated method of deployment is to have the antenna straddle the air-soil interface i.e. partly in free space and partly underground, radiating into the ground. The particular subclass of frequency-independent antenna under investigation for this application is the conical equiangular-spiral antenna, in which thin wires are wound around a conical frame and the radiation is from the apex and reaches its peak in the axial direction. The conical structure, about 50cm long and with a maximum diameter of 12cm, is thrust into the ground apex-first at an angle of about 70 degrees to the vertical.

Acoustic-to-seismic (A/S) coupling has been used successfully to locate anti-personnel (AP) mines with a high probability of detection (P_d). This work builds on previous efforts that have demonstrated the high P_d and low false alarm rate capabilities of A/S coupling in finding ant-tank (AT) mines. This paper discusses the initial results obtained from applying A/S coupling mine detection on anti-personnel mines. Due to the smaller size of AP mines, AP mine detection is more challenging than AT mine detection. The analysis results in this paper are based on A/S coupling mine detection data fro AP mines collected using a laser Doppler vibrometer-based mine detection system. The primary challenge in AP mien detection is to maintain a low false alarm rate while retaining this high probability of detection.

Over the past three years a system has been under development at Georgia Tech that utilizes a seismic interrogation signal in combination with a non-surface- contacting, radar-based displacement sensor for the detection of buried landmines. Initial work on this system investigated the workability of the system concept. Pragmatic issues regarding the refinement of the current experimental laboratory system into a system which is suitable for field testing and, in turn, one which would be suited to field operations have been largely ignored until recently. Both field operations and realistic field testing require a system that is different from the original laboratory system in two crucial ways. One of these is that a field system needs a sensor standoff from the ground surface larger than the original 1 to 2 cm. This is necessary in order to account for small-scale topography, to avoid ground cover such as grass, and to minimize the risk to the operator. A second difference is that the scanning speed of a field system must be substantially greater than that of the original laboratory system, which takes several hours to image 1 m² of ground surface. From an operational standpoint, the reason for this is obvious. From an experimental standpoint, it is also important because ambient conditions are difficult to control on long time scales outdoors. Both of these new requirements must be met within the design parameters that were established empirically during the development of the laboratory system.

A 3D finite-difference time-domain model for elastic waves in the ground has been developed and implemented on a massively parallel computer. The numerical model has been developed as part of a project in which elastic and electromagnetic waves are used synergistically to detect buried land mines. The numerical model is used to study the interaction of elastic waves with buried land mines. As a first approach, a simple model for a TS-50 antipersonnel mine has been developed, and the interaction of elastic waves with the buried mine has been investigated. In both experimental results and numerical simulations, resonant oscillation occur at the location of buried mine. To further explore the resonant behavior of a buried land mine, a refined mine model has been developed that includes more details of the actual mine. Using the refined mine model, the nature of the resonance is explained, and the parts of the mien that influence the resonant oscillations are identified. Results are presented which describe the resonance as a function of burial depth and soil parameters.

A system has been developed that uses high frequency seismic waves and non- contacting displacement sensors for the detection of land mines. The system consists of a moving displacement sensor and a stationary elastic-wave source. The source generates elastic waves in the earth. These waves propagate across the minefield where they interact with buried mines. The sensor measures the displacements at the earth's surface due to the passage of the waves and the interactions of the waves with mines. Because the mechanical properties of the mine are different from those of the earth, the surface displacements caused by the interaction are distinct form those associated with the free-field propagation of the waves. This provides the necessary cue for mine detection. The system has been demonstrated in a controlled laboratory environment, and efforts are currently underway to transition this work into field tests. Moving the experimental effort into the outdoor environment is a critical milestone toward the ultimate goal of this research effort, which is the design of a field-operable mine detection and classification system. There are many issues associated with this transition. Foremost among these is the propagation characteristics of seismic waves in the field environment and, particularly, the mechanisms that limit the energy which can be coupled into the seismic signal that is used to search for mines. To investigate this, a measurements was undertaken to determine the effects of environmental factors at both sites on the generation and propagation of seismic waves. At both sites, strong non-linearity was observed which limited the energy content of the incident signal.

Seismo-acoustic detection has demonstrated a high potential for the detection of land mines with a low probability of false alarms. A key element in the implementation and optimization of this new detection approach is the physical model of the mine-soil system. The validated model of the mine-soil system employs a mass-spring approach, which characterizes the dynamic response of the system using very few parameters derived from the dynamic mechanical impedances of the soil and the mines. This presentation describes the model and the results of the impedance measurements of live antitank and antipersonnel mines. The paper also deals with the optimization of the detection algorithm and its performance based on mine types, burial depth, and soil condition.

Acoustic waves can be a viable tool for the detection and identification of land mines, unexplored ordnance and other buried objects. Design of acoustic instruments and interpretation and processing of acoustic measurements call for accurate numerical models to simulate acoustic wave propagation in a heterogeneous soil with buried objects. Compared with the traditional seismic exploration, high attenuation is unfortunately ubiquitous for shallow surface acoustic measurements because of the loose soil and the fluid in its pore space. To adequately mode such acoustic attenuation. , we propose a comprehensive multidimensional finite-difference time-domain model to simulate the acoustic wave interactions with land miens and soils based on the Biot theory for photoelastic media. For the truncation of the computational domain, w use the perfectly matched layer (PML). The method is validated by comparison with analytical solutions. Unlike the pure elastic wave model, this efficient PML-FDTD model for photoelastic media incorporates the interactions of waves and the fluid-saturated pore space. Several typical and mine detection measurements are simulated to illustrate the application.

Land mines buried a few inches below the surface of the ground can be found by acoustic excitation of the porous ground surface and measuring the particle velocity at the surface. There are various theoretical models describing the ground: from a rigid porous frame model to a compete layered poroelastic description. The goal of this paper is to use the approach of Berry et al. to calculate the acoustic field at points on the ground surface in the vicinity of an object buried in a rigid, porous soil. The excitation is point sound source placed in the air above the ground, which is modeled a rigid, porous frame. A boundary element method is used for numerical integration to calculate the scattered acoustic field due to the presence of the object. This study represents the first step towards developing a complete model of acoustic scattering from near-surface objects embedded in a layered poroelastic material. The predicted disturbance associated with the buried object is much smaller than observed in field measurements.

We establish that weak impulses can be effectively used to detect and image buried objects in disordered granular beds. The image carries information about the approximate shape of the surface of the buried object and its approximate location. Impulse backscattering can serve as a powerful tool to detect and to image the identity of buried metallic and non-metallic objects.

A desirable characteristic for a landmine detection system is the ability of the detector to 'look' out in front of the vehicle a significant distance. The obvious reason for this is to reduce the risk to the vehicle and its operators and to allow a safe stopping distance for the vehicle. Several experiments were conducted at Fort A. P. Hill to investigated the feasibility of a forward-looking system based on acoustic-to-seismic coupling. The system, developed at the National Center for Physical Acoustics, insonifies the ground with high amplitude (120 dB), broadband (80-300 Hz) sound and measures the resulting ground vibration with a scanning Laser Doppler Vibrometer (LDV). Images produced by these scans show a distinct contrast in several frequency bands between ground vibrations over a buried mine and those not over a buried mine. In a forward-looking system, both the sound source and the LDV are moved farther from the scanned area. This configuration both reduces the sound pressure level at the scanned area and decreases the angle at which the LDV beam strikes the ground. These effects reduce the contrast between the over-mine and off-mine signals. In addition, the image is distorted at the shallower LDV-ground angles. However, the results from the experiments demonstrate that the acoustic-to-seismic forward-looking approach is feasible once these technical hurdles are overcome.

The use for subsurface buried object detection of high-frequency (15-30 kHz) acoustic waves generated by CO2 laser pulses incident on the surface of dry sand has been demonstrated previously. In this work, field tests of the technique have demonstrated imaging of landmine simulants buried 2.5 cm below the surface in an outdoor test track. Acoustic finite-difference time-domain calculations have given insight into the observed acoustic lineshapes and verified that the over-estimate of the target dimensions in the outdoor field trials may be related to the lower frequency detector used in these measurements. The models also suggest that a large increase in detected signal may potentially be gained by the use of a Laser Doppler Vibrometer interfacial velocity detector in the place of the present airborne microphone.

Landmine detection can be cast as a model selection problem in which probability theory is used as logic for inductive inference. Using this method, the landmine detection decision is based on the values of calculated posterior probabilities for two propositions: 'The received signal is from a landmine' and 'The received signal is from the background.' The posterior probability for a proposition is the probability for the proposition given the observed data signal and the information known prior to the observation. Calculation of the posterior probability requires the numerical integration of a multi-dimensional probability density function. Until the beginning of the last decade, there were few robust methods available to perform these numeral integrations and no methods that could be generally applied. As a result, probability theory as logic for inductive inference found only infrequent use in practical detection algorithms. Because of the increasing power of computers and new research in the areas of Markov chain Monte Carlo and multi-dimensional adaptive-quadrature integration methods, practical detection algorithms based on the use of probability theory as logic for inductive inference are now being developed and used. This paper describes our model selection formulation of the landmine detection problem and presents results obtained using multi-dimensional adaptive quadrature.

Calculation of the electromagnetic scattering form a buried object, such as a landmine, under acoustic vibration requires a scattering solution for an object beneath an interface with acoustically-induced surface roughness. An analytical solution is presented for the electromagnetic scattering from a dielectric circular cylinder embedded in a dielectric half-space with a slightly rough interface. The solution utilizes the spectral representation of the fields and accounts for all the multiple interactions between the rough interface and the buried cylinder. First order coefficients from the small perturbation method are used for computation of the scattered fields from the rough surface. The derivation includes both TM and TE polarizations and can be easily extended for other cylindrical buried objects. Scattering scenarios are examined utilizing the new solution for a dielectric cylinder beneath both flat and arbitrary surface profiles.

The Ground Standoff Mine Detection System (GSTAMIDS) is now in the Engineering, Manufacturing and Development (EMD) Block 0 phase for USA CECOM. The Mine Detection Subsystem (MDS) presently utilizes three different sensor technologies to detect buried anti-tank (AT) land mines; Ground Penetrating Radar (GPR), Pulsed Magnetic Induction (PMI), and passive infrared (IR). The GSTAMIDS hardware and software architectures are designed so that other technologies can readily be incorporated when and if they prove viable. Each sensor suite is designed to detect the buried mines and to discriminate against various clutter and background objects. Sensor data fusion of the outputs of the individual sensor suites then enhances the detection probability while reducing the false alarm rate from clutter objects. The metal detector is an essential tool for buried mine detection, as metal land mines still account for a large percentage of land mines. Technologies such as nuclear quadrupole resonance (NQR or QR) are presently being developed to detect or confirm the presence of explosive material in buried land mines, particularly the so-called plastic mines; unfortunately, the radio frequency signals required cannot penetrate into a metal land mine. The limitation of the metal detector is not in detection of the metal mines, but in the additional detection of metal clutter. A metal detector has been developed using singular value decomposition (SVD) extraction techniques to discriminate the mines from the clutter, thereby greatly reducing false alarm rates. This mine detector is designed to characterize the impulse response function of the metal objects, based on a parametric three-pole model of the response, and to use pattern recognition to determine the match of the responses to known mines. In addition to discrimination against clutter, the system can also generally tell one mine type from another. This paper describes the PMI sensor suite hardware and its physical incorporation into the GSTAMIDS sensor modules. This is a time-domain, transient signal metal detector that gives target signal response information of a different nature than that from more conventional continuous-wave (CW) metal detectors. The magnetic design of the GSTAMIDS PMI has very broad-band radiation properties that allow for the required transient eddy current responses in the metallic targets. The design of this detector is unique in that it allows processing of the received signals from targets to begin at the very start of the eddy current decays. This then gives the ability to measure and quantify up to three decay terms in the target response, which features unambiguously identify the particular threat target. The results of the data processing algorithms that are used to extract the features used for mine detection are included herein to more clearly show the mine signals.

The paper describes the design of a cheap ultra wideband GPR front-end suitable for subsurface imaging with resolution sufficient for antipersonnel mine recognition. The front-end comprises a generator section, a multi-static antenna system and a receiving unit based on a multi-channel sampling converter. In comparison with commercially available video impulse GPR systems the key advantages of the front-end are considerably larger bandwidth, high precision of measurements of scattered field, ability to measure polarimetric structure of the scattered field and high pulse repetition rate resulting in fast data acquisition.

A cavity backed coplanar waveguide to coplanar strip - fed logarithmic uniplanar spiral antenna, which covers a 9 to 1 band-width with a return loss better than 10 dB from 0.4 to 3.8 GHz is presented. A wideband balun, with an insertion loss of less than 3 dB in the frequency band of operation, was developed for the balanced antenna feed. To aid the balun and antenna design, a method of moment computer program, was used to predict the performance of the spiral antenna. Measurements in an anechoic are made in order to verify the simulated far-field radiation pattern, the simulated polarisation, and the simulated input impedance. Additional advantage of the fabricated antenna is the low cost FR-4 substrate used for the antenna combined with the advantages of the uniplanar circuit, makes this configuration suitable as a low-cost wideband antenna. The constructed uniplanar spiral antenna is very well suited to be used in a stepped frequency ground penetrating radar for humanitarian demining due to the very wide bandwidth, relative small size. Successful detection of a small 5.4 cm non-metallic AP-mines in a pseudo minefield are presented.

We consider here a selection of Impulse Radiating Antennas (IRAs) that may find use in mine detection. Such devices generally consist of a paraboloidal reflector and a broadband feed. This configuration generally allows the radiation of a clean impulse with approximately two decades of bandwidth, when driven by a step-function source. In certain systems, IRAs may prove to be preferable to other antennas that might be used in mine detection. IRAs radiate a focused plane-wave onto the target, in both the near and far fields. Because the radiation pattern is focused, clutter in the received signal may be reduced. In addition, a focused beam may allow a better look-ahead capability. Two designs in particular will be explored, with either a solid reflector and a collapsible reflector fabricated from a conducting mesh. We provide extensive data on both designs with respect to gain, impulse response, beamwidth, and crosspol performance. We also provide detailed comparisons of IRAs, with two different feed arm locations, +/- 45 degrees, and +/- 30 degrees to the dominant polarization. For IRAs with either collapsible reflectors or solid reflectors, we demonstrate improved performance when the feed arms are posited at +/- 30 degrees to the dominant polarization. In both cases, optimization of the feed arm geometry improved both the gain and the crosspol rejection of the antennas. A mild side effect was a slight increase in TDR reflections at the end of the feed arms.

The paper describes the design of an antenna system for a video impulse radar dedicated to landmine detection. The developed antenna system consists of a dielectric wedge antenna as transmit antenna and loop antennas as receive ones. The receive antennas are situated below the transmit one. The dielectric wedge antenna has been designed on the basis of the transmission line model, which has been later verified by means of FDTD modeling. The loop antenna has been designed on the basis of a semi-analytical model. The transient behavior of the antenna system has been investigated both experimentally and theoretically. It is shown that the developed system illuminates a limited spot on the ground surface with a short electromagnetic pulse. The field scattered by objects is received by the antenna system in a local point and without integration over large antenna aperture. The output of each receive antenna reproduces the waveform of the scattered field in a local point. To avoid any mechanical contact with hazardous objects the antenna system should be elevated at least 10cm above the ground.

IN previous papers, we reported on the high-resolution ground-penetrating radar (GPR) system designed, built, and deployed by SRI under contract to the Night Vision and Electronic Sensors Directorate at Fort Belvoir. This fully- polarimetric 300 to 3000 MHz stepped frequ3nyc radar is configured to act as a forward-looking synthetic aperture system with resolution approaching 5 cm. The system is being used as a test bed in a program to define the optimal radar parameters and supporting image processing needed for the efficient standoff detection of buried and surface-laid antitank mines. In this paper, we report in detail on the latest test result from the recent field demonstration performed at a government test site. The test site had been carefully designed to produce statistically significant results, by employing many samples of a few representative metal and plastic mine types, buried at several depths. Statistics for baseline performance will be presented for both metal and plastic, buried and surface mines. Improvements in performance above the baseline have been realized by using EM modeling results to tailor polarimetric, spectral, and matched-filter processing techniques. These efforts are reviewed and the results are presented.

SRI International is investigating change detection for the US Army Night Vision Laboratory using a forward-looking ground-penetrating radar (GPR) sensor as a technique for monitoring cleared roadways against further emplacement of landmines. In the course of evaluating the utility of the technique against buried mines, we have noted that the sensitivity of GPR change detection appears to be sufficient to discern small changes in a scene (footprints, soil disturbance, etc.). Experiments have shown that this sensitivity can decrease significantly over a period of several hours. We have undertaken to learn whether this temporal degradation is because of system and processing artifacts or the inherent temporal change of the clutter field. We have found that the degradation appears to be related to a nonzero phase of the correlation peak observed when the 'before' and 'after' images are cross correlated. This behavior suggests that the images may be misregistered. A simple subpixel registration algorithm was formulated to correct the misregistation. The differencing of two nominally identical images finds the clutter suppression, as determined from the ratio of mean pixel magnitudes in difference and sum images, to have been improved by 7 to 10 dB following subpixel registration. In this paper we present results form recent experiments and improvements to our processing algorithms demonstrating the utility of GPR change detection.

The Mine Hunter/Killer system employs a ground penetrating radar (GPR). Twenty antennas sample a 3m swath to measure a 3D depth return from the earth as the vehicle moves forward in a lane. Data has been collected on shallow and deep, metal and low metal landmines. Samples signatures from a metal and plastic cased landmines buried at 6 inches are presented. In each example a hyperbolic signature is observed. Two feature sets that exploit the hyperbolic shape for false alarm reduction are presented. The first uses a pixel clustering technique to isolate the hyperbola in 3D. A vector of size/shape features is extracted and combined with a quadratic polynomial discriminant into a single value. The second feature set utilizes the radon transform. The radon transform sums the tails of the hyperbola allowing the algorithm to differentiate between surface clutter, which tends to be oriented horizontally in depth, and the diagonals of the hyperbola. Performance curves for both the 3D size/shape features and the radon feature are presented.

In this paper, we examine the two-arm conical spiral antenna (CSA) for use in ground-penetrating radars (GPR). Some of the performance characteristics of the CSA in free space such as unidirectional radiation, circular polarization, and frequency independence, would be beneficial in GPR applications when the antenna is operated in a stepped frequency mode. In previous work, it was shown that these characteristics are preserved when the CSA is placed over the ground. The finite-difference time-domain method (FDTD) is used to analyze the CSA over the ground. The model contains all of the details of the antenna, dissipative soil, and buried object. The FDTD analysis is validated by comparing numerical results for the input impedance and the realized gain with measured results that were made wit a pari of identical conical spiral antennas. A parametric study is performed to determined the best antenna geometry for use in a GPR application. The main criterion for this study is to maximize the power transfer into the ground while preserving frequency-independent performance of the CSA over the ground. Antennas with small cone angles are found to be most suitable for this application. The results form this study are used to model a monostatic GPR that uses a single two-arm CSA for transmission and reception. Two different objects buried in the ground are used to illustrate the sensitivity of the CSA to the polarization of the scattered electric field of the target. In the first case, the targets are thin metallic rods that predominantly scatter linear polarization, while in the second case, the targets are buried plastic mines.

In this paper, we use an optimized frequency-wavenumber (F- K) migration method to localize subsurface objects form ground penetrating radar (GPR) array dat. F-K migration coherently processes waves collected at different positions by a GPR array by back-propagating the recorded waves to the underground objects, according to the wave equation. Performance of F-K migration on GPR measurement depends on accurate estimation of wave propagation velocity. Due to measurement noise and random ground surface, F-K migration may lose its resolution and accuracy. We propose an optimized method to improve the performance of F-K migration. The optimized method searches for a better velocity estimate in the framework of Tikhonov regularization. The Tikhonov regularization uses minimum entropy as the regularizer. By trying to minimize entropy of F-K migration results, better performance is achieved in terms of resolution and accuracy. Examples from applying the optimized F-K migration on real data are used to demonstrate its performance.

The generally rough interface between air and ground produces two adverse effects on the performance of a forward-looking ground-penetrating radar system. The first is that the rough surface scatters some of the transmitted radar energy back to the receiver, raising the effective noise level. The second is the reduction in the coherent signal level scattered by a buried target and received by the radar, owing to the passage of the electromagnetic field once through the rough surface as it propagates toward the target and once again as it propagates back to the receiver. We address the second of these effects in this paper. The interface is modeled as a random phase screen, and we specifically consider the problem of backscattering form a buried conducting disk. We quantify the coherent and incoherent components of the backscattering cross-section of this target, using the physical-optics approximation to address the scattering problem itself. It is assumed that the variance of the surface roughness is small over the frequency range of interest. Representative numerical results are shown over the frequency band 300-1900 MHz, derived from the assumption of a Gaussian autocorrelation function of the surface roughness.

Techniques using ground-penetrating radar (GPR) for the detection of targets such as abandoned landmines or unexploded ordnance (UXO) buried under the ground surface continue to receive considerable attention especially in the area of signal processing. In this paper we consider the problem of eliminating the so-called ground-bounce effect, which is due to the specular ground surface reflections of the radar signal. The ground-bounce returns are often significantly stronger than the reflection from a target and pose a challenging problem. Existing techniques commonly assume that the ground response is constant as the radar equipment moves along a track. By using measured data, we show that this is, for several reasons, an unrealistic assumption. Instead, we consider a semi-parametric model for the ground-bounce that is in better agreement with observed data. Furthermore, we show how this model can be used to derive an accurate and robust but yet conceptually simple algorithm for the removal of the ground return. We demonstrate our technique using data recorded by an ultra-wideband GPR on a U.S. Army test range.

This paper presents a novel polarimetric near-field two-dimension (2D) synthetic aperture focusing technique (SAFT) suitable for ground penetrating radar (GPR) application. The imaging algorithm is intended for locating metallic and non-metallic anti-personnel (AP's) mines using an ultra-wide-band stepped frequency radar. A radar image can be formed by coherently integrating the backscattered field over the measured frequency spectrum and cross-range scan. The coherent integration is essentially a convolution of the collected data and a focusing (test) function, which only depends on the geometry of the measurement. Wavefront curvature must be taken account of when attempting to image an object within 1-2 wavelengths off an antenna(s) phase center. Applying conventional far-field SAR imaging using a direct Fourier inversion may result in images which are increasingly blurred and shifted at points more distant from the point of rotation of the focusing function. Here, a focusing function is first derived based on a conventional far-field geometrical optic propagator for a two-media problem. Then to correct for geometric distortion in the focusing function when applied in the near-field zone we introduce higher order terms to the range function. In order to verify and augment the technique described two field studies were conducted, over different frequency spectrums, the finding of which demonstrates the utility of the technique and the experimental practices.

The Ground Standoff Mine Detection System is now in the Engineering, Manufacturing and Development (EMD) Block 0 phase for USA CECOM. This paper describes the data processing algorithms for the GPR that are used to extract the features used for anti-tank (AT) mine detection; those used for pre-processing the data re included herein to show the enhancement of the mine signals. A key feature of the processing is the acquisition of a clean radar return signal from undisturbed soil, which is then deconvolved from each data frame waveform. This soil signal is an estimate of the system impulse response function, save for the magnitude of the reflection coefficient of the soil, which is a scalar to first order. Deconvolution thus gives the impulse response function of the buried mines, a strong enhancement over their raw measured signals. A matched filter test statistic is generated to discriminate between mines and background. Discrimination algorithms using hidden Markov model processing are describe in a paper by PD Gader et al. These processes were developed in MATLAB using dat files acquired and stored from prototype GPR systems and then refined with data form production units. The MATLAB code is then converted into C code for use on the real-time processor on GSTAMIDS. The C code modules are run as dynamic library links in MATLAB for verification. The GPR sensor suite hardware and its physical incorporation into the GSTAMIDS sensor modules are described fully in a companion paper.

The structure of a common land mine typically exhibits many types of geometric symmetries. These structural symmetries translate into symmetries of the bistatic scattering response of a mine target. A bistatic radar array can therefore be constructed that identifies the presence of these symmetries, and thus can distinguish mines from other asymmetric subsurface objects, such as rocks. To fully exploit this phenomenon, a mathematical description of the types of geometric symmetries exhibited by mines, as well as the resulting effect on bistatic scattering, is required. Geometric symmetry can be described mathematically in terms of group theory. An object that is invariant under ever operation (e.g. reflection, rotation) ofa finite symmetry group is said to posses the symmetry ofthat group. In this paper, this concept is extended to bistatic measurements, such that invariant bistatic measurements can be identified for each mine type. For example, from a single bistatic observation of a square mine, fifteen other bistatic observations of that mine can be determined. The bistatic scattering response is therefore determined to be invariant over a group of order 16. By contrast, the scattering response ofan asymmetric object is invariant over a group oforder two. As a result, the proper collection of bistatic observations can provide a powerful method for classifying subsurface objects as either symmetric (e.g., mines) or asymmetric (e.g., rocks). This paper demonstrates how group theory can be used construct the necessary bistatic array processing to accurately determine different types of target symmetries at arbitrary subsurface locations. This processing is applied to an array of bistatic GPR measurements, and the results demonstrate the efficacy of this idea in identifying symmetric subsurface objects.

This paper investigates the fusion of the confidence outputs of the Energy Based Processing (EBP) algorithm from the BAE Systems and the HMM GPR algorithm from the Univ. of Missouri to increase the performance of the Mine Hunter/Killer (MH/K) vehicle mounted landmine detection system. The EBP algorithm is based on the energy changes in GPR signal for detection. The HMM algorithm, on the other hand, is a feature based technique that relies on hyperbolic signatures to detect landmines. When fusing the detection confidences of the two algorithms properly, the performance is improved dramatically. The detection performance after fusion is demonstrated using data measured at a prepare test site during February and June 2000. Similar diagonal features used in HMM have been implemented and fused with EBP algorithm. Official offline scoring shows that the MH/K exit criteria of 92 percent Pd at 0.013/m2 FAR is met.

Proper clutter reduction is essential for Ground Penetrating Radar data since low signal-to-clutter ratio prevent correct detection of mine objects. A signal processing approach for resolution enhancement and clutter reduction used on Stepped-Frequency Ground Penetrating Radar (SF-GPR) data is presented, and the effects of combining clutter reduction with resolution enhancement are examined using simulated SF-GPR data examples. The resolution enhancement method is based on methods from optical signal processing and is largely carried out in the frequency domain to reduce the computational burden. The clutter reduction method is based on basis function decomposition of the SF-GPR time-series from which the clutter and the signal are separated.

We describe our continuing work on the development and use of collaborative virtual environments (CVEs) in support of Naval Special Warfare mission rehearsal activities for the very shallow water mine countermeasures mission. These multi-participant CVEs are built form multiple data streams that include archived numerical circulation model results, 3D model files, moving entities and observations from advanced instrumentation and autonomous underwater vehicles. A room-sized multi-walled projection device known as a CAVE is used to provide an immersive quality to the simulations thereby adding to the realism of data presentation activities. Asynchronous collaborative capability allows users at many remote sites to partake in a many-to-many session that takes place in a common virtual world. A participant is able to view, navigate through and interact with the data and with other users in the 3D environment. Typical data types represented in the mine countermeasures CVE include graphical representations of bathymetry/topography, above-surface images, in-water objects, high-resolution data and hydrographic characteristics.

As landmines get harder to detect, the complexity of landmine detectors has also been increasing. To increase the probability of detection and decrease the false alarm rate of low metallic landmines, many detectors employ multiple sensing modalities, which include radar and metal detector. Unfortunately, the operator interface for these new detectors stays pretty much the same as for the older detectors. Although the amount of information that the new detectors acquire has increased significantly, the interface has been limited to a simple audio interface. We are currently developing a hybrid audiovisual interface for enhancing the overall performance of the detector. The hybrid audiovisual interface combines the simplicity of the audio output with the rich spatial content of the video display. It is designed to optimally present the output of the detector and also to give the proper feedback to the operator. Instead of presenting all the data to the operator simultaneously, the interface allows the operator to access the information as needed. This capability is critical to avoid information overload, which can significantly reduce the performance of the operator. The audio is used as the primary notification signal, while the video is used for further feedback, discrimination, localization and sensor fusion. The idea is to let the operator gets the feedback that he needs and enable him to look at the data in the most efficient way. We are also looking at a hybrid man-machine detection system which utilizes precise sweeping by the machine and powerful human cognitive ability. In such a hybrid system, the operator is free to concentrate on discriminant task, such as manually fusing the output of the different sensing modalities, instead of worrying about the proper sweep technique. In developing this concept, we have been using the virtual mien lane to validate some of these concepts. We obtained some very encouraging results form our preliminary test. It clearly shows that with the proper feedback, the performance of the operator can be improved significantly in a very short time.

Although the ability of EMI sensors to detect landmines has improved significantly, false alarm rate reduction remains a challenging problem. However, experienced operators can often discriminate mines from metallic clutter with the aid of an audio transducer. The goal of this work is to optimize the presentation of information to the operator and to determine whether information as to the presence of metal can be co-presented with information regarding mine/non-mine belief. Traditionally, an energy calculation is provided to the sensor operator via a signal whose loudness and/or frequency is proportional to the energy of the received signal. This information codes information as to the amount of metal present. However, there is information in the unprocessed sensor signal that the operator could use to effect discrimination. We have experimentally investigated the perceptual dimensions that most effectively convey the information in a sensor response to a listener using simulated data. Results indicated that, consistent with the auditory warning literature, pulsed audio signals with a distinct harmonic pattern which rise in fundamental frequency can be used to provide information which provides better performance than simple single-frequency tones. Additionally, the data indicated that the amount of metal could be coded in the rising pitch of the complex, and that the mine/no-mine probabilities could be coded in a separate dimension - the pulse rate. In this paper, we describe these results in detail.

We report results from an experiment designed to study the perceptual and learning processes involved in the detection of land mines. Subjects attempted to identify the location of spatially distributed targets identified by a sweeping a cursor across a computer screen. Invisible screen 'objects' were identified by either tones (A) or clicks (B) or both. Objects defined by a tone or a click only are distracters; the single object defined by both is the target (mine). We looked at the effect on target detectability of the number and spatial distribution of distracters. As expected from theoretical analysis, target detectability was highest when A and B distracters were negatively correlated; lowest when they were positively correlated. Under these conditions, detectability is was also inversely related to the number of A distracters (which were spatially diffuse) but was largely unaffected by the number of B distracters (which were punctate). Adding a second sensor channel greatly enhanced target detectability, especially if A and B distracters were spatially uncorrelated or negatively correlated.

Sensing the chemical signature emitted from the main charge explosives from buried landmines is being considered for field applications with advanced sensors of increased sensitivity and specificity. The chemical signature, however, may undergo many interactions with the soil system, altering the signal strength at the ground surface by many orders of magnitude. A simulation code named T2TNT was developed specifically to evaluate buried landmine chemical transport issues. A vapor-solid partitioning parameter that is strongly dependent on soil moisture content is included in T2TNT. Laboratory soil vapor flux experiments were conducted to provide data to validate the T2TNT model under well-constrained laboratory testing conditions. The landmine source release, soil transport and surface flux was simulated by aqueous phase injection of DNT, evaporation induced upward water flux and solid phase microextraction sampling of headspace vapor in an air flowing plenum. The surface soil moisture content was reduced by suction removal of soil water followed by artificial rain to evaluate the soil-vapor partitioning function in T2TNT. The data showed the dramatic decline in DNT vapor concentrations expected as the surface soil moisture declined; and, then rebounded upon wetting. This phenomenon was modeled with T2TNT and showed excellent correlation.

Environmental fate and transport studies of explosives in soil indicate that 2,4,6-trinitrotoluene (TNT) and similar products such as dinitrotoluene (DNT) are major contributors to the trace chemical signature emanating from buried landmines. Chemical analysis methods are under development that have great potential to detect mines, or to rapidly classify electromagnetically detected anomalies as mines vs. 'mine-like objects'. However, these chemical methods are currently confined to point sensors. In contrast, we have developed a method that can remotely determine the presence of nitroaromatic explosives in surface soil. This method utilizes a novel distributed granular sensor approach in combination with uv-visible fluorescence LIDAR (Light Detection and Ranging) technology. We have produced prototype sensor particles that combine sample preconcentration, explosives sensing, signal amplification, and optical signal output functions. These particles can be sprayed onto soil areas that are suspected of explosives contamination. By design, the fluorescence emission spectrum of the distributed particles is strongly affected by absorption of nitroaromatic explosives from the surrounding environment. Using ~1mg/cm2 coverage of the sensor particles on natural soil, we have observed significant spectral changes due to TNT concentrations in the ppm range (mg TNT/kg soil) on 2-inch diameter targets at a standoff distance of 0.5 km.

Electronic noses and similar sensors show promise for detecting buried landmines through the explosive trace signals they emit. A key step in this detection is the sampler or sniffer, which acquires the airborne trace signal and presents it to the detector. Practicality demands no physical contact with the ground. Further, both airborne particulates and molecular traces must be sampled. Given a complicated minefield terrain and microclimate, this becomes a daunting chore. Our prior research on canine olfactory aerodynamics revealed several ways that evolution has dealt with such problems: 1) proximity of the sniffer to the scent source is important, 2) avoid exhaling back into the scent source, 3) use an aerodynamic collar on the sniffer inlet, 4) use auxiliary airjets to stir up surface particles, and 5) manage the 'impedance mismatch' between sniffer and sensor airflows carefully. Unfortunately, even basic data on aerodynamic sniffer performance as a function of inlet-tube and scent-source diameters, standoff distance, etc., have not been previously obtained. A laboratory-prototype sniffer was thus developed to provide guidance for landmine trace detectors. Initial experiments with this device are the subject of this paper. For example, a spike in the trace signal is observed upon starting the sniffer airflow, apparently due to rapid depletion of the available signal-laden air. Further, shielding the sniffer from disruptive ambient airflows arises as a key issue in sampling efficiency.

An automated, rapid-cycling vapor concentrator and sensor was designed and evaluated for detection of nitroaromatic compounds. The concentrator consists of an inert deactivated fused silica capillary loop. The temperature of the loop was manipulated through contact with a cold plate or a hot plate. Plates were maintained at pre-selected temperatures with a thermoelectric cooler and a cartridge heater. The low thermal mass of the loop permitted rapid trapping and desorption cycles. The chemical inertness of the fused silica tubing not only led to efficient transfer of nitroaromatics from the air stream to the detection system but also minimized cross contamination between samples. The detection system consisted of a tandem arrangement of electron attachment detectors and an electron attachment reactor. These devices were operated with compact custom electronics. The use of the electron attachment reactor and thermoelectric cooler permitted enhanced selectivity. The device was evaluate for rapid determination of nitroaromatic from air streams at trace concentrations. Detection limits down to sub parts per billion were obtained.

Thin films of carbon black-organic polymer composites have been deposited across two metallic leads, with sorption of vapors producing swelling-induced resistance changes of the detector films. To identify and classify vapors, arrays of such vapor sensing elements have been constructed in which each element of the array contains a different polymer as the insulating phase and a common conductor, carbon black, as the conducting phase. The differing gas-solid partition coefficients for the various polymers of the detector array produce a pattern of differential resistance changes that is used to classify vapors and vapor mixtures. The performance of this detector array system towards 2,4-dinitrotoluene, the predominant signature in the vapor phase above land mines, in the presence high concentrations of water or of acetone has been evaluated.

This paper is concerned with the problem of target geo- location when using forward-looking vehicular-mounted sensors for landmine detection. Intermediate and downward- looking sensor may also be used, but the geo-location problem is most complex for the forward-looking sensor. A nonlinear states model for the vehicle position and attitude. Knowledge of these sensors specifications along with information as to the location and orientation of the sensor on the vehicle combined with knowledge of the vehicle position and attitude make it possible for one to compte the sensor field-of-view or footprint. Given this, one can then analyze sensor frames and for any detected mines, convert their locations from sensor-frame coordinates to ground coordinates.

Mine Hunter/Killer (MH/K) is an Advanced Technology Demonstration (ATD) program directed by the Army Night Vision Electronic Sensors Directorate (NVESD). The MH/K system consists of a vehicle-mounted system that detects and neutralizes surface and buried anti-tank (AT) mines. The detection element in this program consists of a Close-In Detection (CID) System that relies on a multi-sensor configuration. The CID System consists of three sensors: a ground penetrating radar (GPR), a metal detector (MD) and a forward-looking IR imaging system. TRW S and ITG has provided support for analysis, testing and algorithm development for Automatic Target Recognition and sensor fusion processing. This paper presents a multi-sensor fusion approach developeby TRW under this effort. In this approach, the incoming alarms from the three sensors are segregate into five classes, based on spatial coincidence of GPR and MD alarms, and on the presence of a surface null in the GPR depth profile. This GPR null, or 'notch', is indicative of shallowly buried objects or clutter, and helps in the discrimination against false alarm density, attempting to maintain a constant false alarm rate. This paper will describe this fusion methodology and the adaptive threshold method in detail, show the target and clutter probability density functions for each class, and show result form recent field test. Fused results will be compared with single sensor performance, and strengths and weaknesses of each sensor will be discussed.

Current minefield detection research indicates that operationally no single sensor technology will likely be capable of detecting mines/minefields in a real-time manner and at a performance level suitable for a forward maneuver unit. Minefield detection involves a particularly wide range of operating scenarios and environmental conditions, which requires deployment of complementary sensor suites. We have focused, therefore, on the development of a computationally efficient and robust detection algorithm that exploits robust image processing techniques centered on meaningful target feature sets applicable to a variety of imaging sensors. This paper presents the detection technique, emphasizing its robust architecture, and provides performance results for image data generated by complementary sensors. The paper also briefly discusses the application of this detector as a component of fusion architectures for processing returns form diverse imaging sensors, including multi-channel image data from disparate sensors.

As in many areas, performance of landmine detection algorithms is judged in terms of detection and false alarm rates. For the landmine detection problem, it is often the case that detectors satisfy one requirement at the cost of poor performance with regard to the other. It is widely accepted that single sensors cannot simultaneously achieve both high detection rates and low false alarm rates, since every sensor has its advantages and disadvantages when dealing with a large variety of landmines, from large metal-cased mines to small plastic-cased mines. Thus, in this paper we consider two types of sensors, EMI and GPR. In its most common instantiation, time-domain EMI is essentially a metal detector and thus detects mines with high metal content as well as metal debris in the environment. More advanced EMI systems have begun to show potential for the discrimination of such debris from mines. GPR is also used for landmine detection since it can detect and identify low-metallic subsurface anomalies. In our previous work, we have shown that Bayesian detection approach can be applied to EMI data and provide promising results. In this paper, we present results that indicate that statistical signal processing technique applied to GPR data can also yield performance improvements. Theoretical results are verified by data collected with a developmental mine-detection system, which consists of co-located metal detectors and GPR sensors. Thus, in addition to discussing individual sensor data processing, we also present result of data fusion of both the EMI and the GPR data using the detection system.

In this paper, probabilistic fusion of multi-sensor data is applied to mine detection. Probabilistic fusion combines information in the form of scores from automatic target recognition (ATR) algorithms for each sensor. This fusion method has previously demonstrated improved mine detection performance when used with multi-sensor data from the Mine Hunter/Killer system. The sensor suite includes a ground- penetrating radar, metal detectors, and an IR camera; data were collected at a prepared test site. Results of applying the probabilistic fusion method to recent MH/K multi-sensor data using various new ATR algorithms are presented and analyzed in detail. Changes in detection performance are quantified for different combinations of the various ATR algorithms and sensors. It is shown that fusion improves mine detection performance even when the individual sensor and ATR algorithms have very different performance levels. This implies that multi-sensor approaches to mien detection should continue to be pursued.

We investigate the potential for improving land mine detection by fusing data from ground penetrating radars (GPRs) and sensors of acoustically induced soil motion. We present a case study involving data from the SRI forward-looking radar and a laser Doppler vibrometer (LDV) system developed by the University of Mississippi. The LDV sensor detects acoustically induced soil vibrations, while the GPR detects scattering from dielectric discontinuities or metal objects in the soil. Since these sensors exploit different target physical properties, it is reasonable to expect a benefit in fusion. Although the sensors are relatively new, the LDV and GPR data exhibit evidence for complementarity, in that the GPR is significantly better at detecting metal mines, while the LDV is somewhat better at detecting plastic mines. Decision-level fusion is shown to improve performance. A simple OR fusion approach is found to perform similarly to an optimum hard decision fusion algorithm.

Data fusion from two separate and orthogonal mine detection sensors developed independently by the University of Mississippi and Planning Systems Inc. has been performed. The University of Mississippi's acoustic/seismic coupling detection is based on the measurement of ground surface vibration velocity by means of acoustic excitation and a laser Doppler vibrometer. Differences in absolute surface vibration velocity, caused by the present of buried mines, are used to infer the presence of buried land mines. Planning Systems Inc. uses ground-penetrating, synthetic- aperture radar to detect subsurface electromagnetic anomalies. Detection with the GPSAR sensor is based on differences in the dielectric constant of the ground medium and that of a buried land mine. The spatial resolutions of the two measurements are similar and the two sensors measure completely different physical properties. Dat form each system are described in detail and independent examples of performance are presented. A common geo-spatial grid is defined for both sensor systems given their respective resolving capability. Methods of simultaneous display are presented and situations in which the two systems are complementary are identified.

Sensor fusion issues in a streamlined assimilation of multi-sensor information for landmine detection are discussed. In particular multi-sensor fusion in hand-held landmine detection system with ground penetrating radar (GPR) and metal detector sensors is investigated. The fusion architecture consists of feature extraction for individual sensors followed by a feed-forward neural network training to learn the feature space representation of the mine/no-mine classification. A correlation feature from GPR, and slope and energy feature from metal detector are used for discrimination. Various fusion strategies are discussed and results compared against each other and against individual sensors using ROC curves for the available multi-sensor data. Both feature level and decision level fusion have been investigated. Simple decision level fusion scheme based on Dempster-Shafer evidence accumulation, soft AND, MIN and MAX are compared. Feature level fusion using neural network training is shown to provide best results. However comparable performance is achieved using decision level sensor fusion based on Dempster-Shafer accumulation. It is noted that, the above simple feed-forward fusion scheme lacks a means to verify detections after a decision has been made. New detection algorithms that are more than anomaly detectors are needed. Preliminary results with features based on independent component analysis (ICA) show promising results towards this end.

A neural network is applied to data collected by the close-in detector for the Mine Hunter Killer (MHK) project with promising results. We use the ground penetrating radar (GPR) and metal detector to create three channels (two from the GPR) and train a basic, two layer (single hidden layer), feed-forward neural network. By experimenting with the number of hidden nodes and training goals, we were able to surpass the performance of the single sensors when we fused the three channels via our neural network and applied the trained net to different data. The fused sensors exceeded the best single sensor performance above 95 percent detection by providing a lower, but still high, false alarm rate. And though our three channel neural net worked best, we saw an increase in performance with fewer than three channels, as well.

In this paper, two methods for fusion of mine detection sensors are presented, based on belief functions and on voting procedures, respectively. Their application is illustrated and compared on a real multisensor data set collected at the TNO test facilities under the HOM-2000 project. This set contains data acquired by metal detector, infrared camera and ground penetrating radar. The data acquisition and preprocessing are briefly described. For some typical cases presented in this data set, the characteristics extracted and used by both methods are discussed, as well as the answers given by each method and possible causes of potential differences in results. Also, it is shown how the different voting schemes compare to belief functions modeling in various situations, based on the knowledge that is put into the belief functions. Since the roots of the two methods are different, i.e. belief functions involve expert knowledge while voting is a simple approach, the explanations involve these differences. Problems that arise when comparing and evaluating different methods are also addressed. Finally, it is shown that both of the methods have their advantages and drawbacks, depending on the measurement and operational conditions. This paper is a result of a joint work at three European institutions towards a common goal: humanitarian demining.

The objective of the Joint Multi-sensor Mine-signatures (MsMs) campaign is to organize and execute an experimental campaign for collecting data of buried land-mines with multiple sensors. These data sets will then be made widely available to researchers and developers working on sensor fusion, signal processing for improved detection and identification of land-mines, assessing the role of the operator in the detection process, etc. The outdoor test facility of the Joint Research Facility of the European Commission, located at Ispra (Italy), houses the test minefield. Six test strips of 6 by 6 m consisting of different soil types (cluttered grassy terrain, loamy soil, sandy soil, clay soil, soil with high content of organic matter, and ferromagnetic soil) are complemented with one reference test strip of 6 by 6 m consisting of pure sand. The list of objects buried in the minefield includes mine simulants of three different dimensions with either a low or a high metal content, reference targets for position referencing and calibration checking, and clutter objects including empty bullet cartridges, metal cans, barbed wire, stones, wood, plastic boxes, etc. This test minefield is going to be left intact for a long period, in order to be able to perform multiple runs on it. For the test campaign of the year 2000, the core sensors were a metal detector, a ground penetrating radar, a microwave radiometer, and thermal infrared imagers. Later, other (more experimental) detectors will also be tested on the same test minefield. The first data sets are in the process of being released right now.

Simple theoretical models can be constructed to study the behavior of sensor-fused systems using idealized sensor suites. Models are available for feature-level and decision-level fusion, both of which are now being used with demining sensors. These models are attractive as design tools and for estimating the expected performance of new sensor suites, since their performance can be evaluated with relatively little effort. In this paper we review some simple idealized models and their predictions for fused system performance. The data produced by demining sensors are often correlated, and the effect of correlation is explored for both feature-level and decision-level fusion.

Data fusion is a process of combining evidence from different information sources in order to make a better judgement. However, multiple sources can provide complementary information that can be used to increase the performance in detection and recognition. There are many frameworks within which to combine these pieces into a more meaningful answer. However, new information added might be redundant or even conflicting with the existing information. These questions arise: can we predict the value added by fusing their outputs together, if we know the general characteristics of a set of sensors. Can we specify the needed characteristics of a new sensor/algorithm to add to an existing suite to gain a desired improvement performance. The characteristic of a new sensor can be in any forms, e.g., the ratio of a target's signal to the clutter's signal, the position resolution etc. In this paper, we consider these questions in the context of fuzzy set theory and in particular, a soft decision level fusion scheme we developed for land mine detection scenarios. Here, we primarily consider the ratio of a target's signal. We develop a tool to estimate a final d-metric when the information form several sensor is fused through the linguistic Choquet fuzzy integral. We utilize this tool in the examination of the performance of d-metrics in a simulation environment. The approach is demonstrated for data obtained from an Advanced Technology Demonstration in vehicle-based mine detection.

In recent years different sensor data fusion approaches have been analyzed and evaluated in the field of mine detection. In various studies comparisons have been made between different techniques. Although claims can be made for advantages for using certain techniques, until now there has been no single method identified with clearly outstanding performance in all scenarios. In this paper we describe a fusion approach based on a combination of modeling data and feature extraction. By using scenario and environmental data, model predictions are made of sensor data, performance data and mine feature data. These data are then compared with the sensor pre-processing as well as made available for use in the sensor fusion processing. These comparisons take into account the expected and measured sensor features for each object. In scenarios with sufficient a prior knowledge it is expected that detection is improved by applying the model based feature fusion algorithm. The new concept is described and first results primarily based on thermal IR test data recorded at the TNO test facility are presented. Particular attention was paid to the detection pre- processing and the feature fusion stage.

In this paper we introduce the concept of depth fusion for anti-personnel landmine detection. Depth fusion is an extension of common sensor-fusion techniques for landmine detection. The difference lies within the fact that fusion of sensor data is performed in different physical depth layers. In order to do so, it requires a sensor that provides depth information for object detections. Our ground-penetrating radar fulfills this requirement. Depth fusion is then taken as the combination of the output of sensor fusion of all layers. The underlying idea is that sensor fusion for the surface layer has a different weighing of the sensors when compared with the sensor fusion in the deep layers because of apparent sensor characteristics. For example, a thermal IR sensor hardly adds information to the sensor fusion in the deep layers. Furthermore, GPR has difficulties suppressing clutter in the surface layer. As such, the surface fusion should emphasize on the TIR sensor, whereas sensor fusion in the deep layers should have a higher weighing of the GPR. This a priori information can be made explicit by choosing for a depth-fusion approach. Experimental results form measurements at the TNO-FEL test facility are presented that validate our depth-fusion concepts.

In this paper, we work towards a robust approach for imaging weak-contrast objects buried under a rough soil/air interface using data from an electromagnetic GPR array. A major source of variability in observed GPR signals is due to reflection from a rough and random ground. Our approach to imaging is based on us of physical and statistical modeling techniques to estimate and compensate for this rough soil/air interface. An approximate physical model based on Gaussian beams is used to model the interaction of the illumination with the ground and estimate the surface profile. This estimated surface profile is then used to correct the raw data for the effects of the rough surface. The corrected data may subsequently be used to reconstruct the subsurface region and localize anomalies. In this stage, statistical models can be used to account for both noise and residual unmodeled effects.

The fusion of multiple detection/classification algorithms is proving a very powerful approach for dramatically reducing false alarm rate, while still maintaining a high probability of detection and classification. This paper presents insights on how algorithm fusion achieves these benefits.

The automated detection of sea mines remains an increasingly important humanitarian and military task. In recent years, research efforts have been concentrated on developing algorithms that detect mines in complicated littoral environments. Acquired high-resolution side-looking sonar images are often heavily infested with artifacts from natural and man-made clutter. As a consequence, automated detection algorithms, designed for high probability of detection, suffer from a large number of false alarms. To remedy this situation, sophisticated feature extraction and pattern classification techniques are commonly used after detection. In this paper, we propose a nonlinear detection algorithm, based on mathematical morphology, for the robust detection of sea mines. The proposed algorithm is fast and performs well under a variety of sonar modalities and operating conditions. Our approach is based on enhancing potential mine signatures by extracting highlight peaks of appropriate shape and size and by boosting the amplitude of the peaks associated with a potential shadow prior to detection. Signal amplitudes over highlight peaks are extracted using a flat morphological top-hat by reconstruction operator. The contribution of a potential shadow to the detection image is incorporated by increasing the associated highlight amplitude by an amount proportional to the relative contrast between highlight and shadow signatures. The detection image is then thresholded at mid-gray level. The largest p targets from the resulting binary image are then labelled as potential targets. The number of false alarms in the detection image is subsequently reduced to an acceptable level by a feature extraction and classification module. The detection algorithm is tested on two side-scan sonar databases provided by the Coastal Systems Station, Panama City, Florida: SONAR-0 and SONAR-3.

The performance of Computer Aided Detection/Computer Aided Classification (CAD/CAC) Fusion algorithms on side-scan sonar images was evaluated using data taken at the Navy's's Fleet Battle Exercise-Hotel held in Panama City, Florida, in August 2000. A 2-of-3 binary fusion algorithm is shown to provide robust performance. The algorithm accepts the classification decisions and associated contact locations form three different CAD/CAC algorithms, clusters the contacts based on Euclidian distance, and then declares a valid target when a clustered contact is declared by at least 2 of the 3 individual algorithms. This simple binary fusion provided a 96 percent probability of correct classification at a false alarm rate of 0.14 false alarms per image per side. The performance represented a 3.8:1 reduction in false alarms over the best performing single CAD/CAC algorithm, with no loss in probability of correct classification.

An advanced, automatic, adaptive clutter suppression, sea mine detection, classification and fusion processing string has been developed and tested with high resolution sonar imagery dat. The overall computer-aided-detection/computer- aided-classification (CAD/CAC) string includes pre- processing, adaptive clutter filtering (ACF), normalization, detection, feature extraction, feature orthogonalization, subset feature selection, classification and fusion processing blocks. The ACF is an adaptive linear FIR filter, optimal in the Least Squares (LS) sense, and is applied to low-resolution data. Data pre-normalization, clipping and mean subtraction, allows application of a range dimension only ACF that is matched both to average highlight and shadow information, while simultaneously suppressing background clutter. Following post-ACF normalization, and 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 three processing strings, developed by three different researchers, are fused, using an LLRT-based fusion rule. Processing string improvements have been developed over previous CAD/CAC and fusion string versions. The utility of the overall processing strings and their fusion was demonstrated with very shallow water high-resolution sonar imagery data sets, form a difficult environment. The processing string classification performance was optimized by appropriately selecting a subset of the original feature set. The fusion of the CAD/CAC processing strings resulted in improved mine classification capability, providing a three-fold false alarm rate reduction, compared to the best individual CAD/CAC processing string results.

We describe here the current form of Alphatech's image processing and neural network based algorithms for detection and classification of mines in side-scan sonar imagery, and results obtained form their application to three distinct databases. In particular, we contrast here results obtained from the use of a currently employed 'baseline' multilayer perceptron classifier training approach, with the use of a state of the art commercial neural network package, NeuralSIM, developed by Neuralware, Inc.