An algorithm based on Bayesian probability theory is developed to discriminate buried metallic landmines from buried metallic clutter. A binary hypothesis problem is formed using the two hypotheses that the buried object is either a mine-like object or a clutter-like object. The received signal under both hypotheses is modeled as a target function, which is a delayed decaying exponential, plus Gaussian noise. The target functions contain the target's decay rate and coupling strength information. The coupling strength manifests itself as the point where the buried target's response reasons comes out of amplifier saturation. A target with a large coupling strength will fall out of saturation much later in time that a target with a low coupling strength. The decay rate for each buried object is extracted using a differential-corrections routine. The decay rate and fallout time are considered random variables with known distributions under each hypothesis. The distribution for the mine decay rates and fallout times are calculated from four separate measurements taken in a calibration area. The distribution of decay rates and fallout times for all objects in a blind grid is also estimated.

EMI sensors are used extensively to detect landmines, and operate by detecting the metal that is present in mines. However, mines vary in their construction form metal-cased varieties with a large mass of metal to plastic-cased varieties with minute amounts of metal. Unfortunately, there is often a significant amount of metallic clutter present in the environment. Consequently, EMI sensors that utilize traditional detection algorithms based solely on metal content suffer form large false alarm rates. We have at least partially mitigated this false alarm problem for high- metal content mines by developing statistical algorithms that exploit phenomenological models of the underlying physics. The Joint UXO Coordination Office (JUXOCU) is sponsoring a series of experiments designed to establish a performance baseline for a variety of sensors. The experiments to dat have focused on detection and discrimination of low-metallic content mines. This baseline will be used to measure the potential improvements in performance offered by advanced signal processing algorithms This paper describes the result of several experiments performed in conjunction with the JUXOCU effort. In our preliminary work, statistical algorithms have been applied specifically to the problem of detection of low-metal mines, and dramatic performance improvements have been observed with respect to the baseline performance. However, these algorithms improvements have been observed with respect to the baseline performance. However, these algorithms were statistical in nature, did not incorporate phenomenological models, and exploited spatial information. The tradeoffs among these various factors are explored in this paper, along with the performance of alternative statistical approaches. In addition, approaches to classification of the mine-type are discussed and the performance of such classifiers is presented.

Traditionally, field EMI sensors are operated in the time- domain. The time-domain (TD) EMI sensor usually is a pulsed system. It contains both a transmitting coil and a receiving coil. After transmitting an excitation pulse, which generates the primary field, the receiving coil records the secondary field in the late time. Since a TD EMI sensor measures only the late-time responses, the information contained in the early time response is lost thus limiting the types of objects that can be discriminated. Alternatively, EMI sensors can be operated in the frequency- domain (FD). In this case, the excitations are sinusoidal signals and the sensor measures the static response. The advantages and disadvantages of TD and FD EMI sensors are reviewed in this paper. For landmine and UXO detection, discrimination of targets of interest from clutter is required, since the cost of large false alarm rates is substantial amounts of money, labor and time. In order to discriminate targets from clutter, Bayesian optimal classifiers are derived. Traditional detectors for these applications only utilize the energy of the signal at the position under test or the output of a matched world scenario, the depth of the underground objects is uncertain. The optimal classifier that we utilize takes these uncertainties into account also. In this paper, we present classification performance for four metal objects using TD and FD EMI data. Experimental data were taken with the PSS- 12, a standard army issued metal detector, and the GEM-3, a prototype frequency-domain EMI sensor. Although the optimal classifier improves performance for both TD and FD data, FD classification rate are higher than those for TD systems. The theoretical basis for this result is explored.

Target discrimination via decay rate (pole) estimation has been proposed a an effective method for landmine and UXO detection. The physical basis for this strategy is that every object in the target library possesses a unique set of decay rates, or poles. In theory, these poles can be estimated from the measured EMI response and then utilized for target detection and subsequent discrimination and/or identification. Unfortunately, pole estimation is notoriously difficult and this difficulty adversely impacts target discrimination performance. We present simulation results that show that when the sensor/object orientation is not known, the time-domain signature of two objects with distinct sets of poles may be indistinguishable. Furthermore, this can occur even when the two sets of poles are not necessarily 'close' to each other. Since the basis for this approach to target detection and identification is that targets are uniquely characterized by their estimated poles, discrimination performance is dependent upon pole estimation performance. The Cramer-Rao lower bound, which provides a lower bound on the variance of an unbiased estimator, for the pole and amplitude coefficient estimates is utilized to investigate the fundamental limitations on target discrimination via pole estimation. It is shown how both the sampling strategy and the number of poles being estimated affect pole estimation performance. Detection and identification results are presented for simulated data.

The problem of low metal content mine characterization form broadband electromagnetic induction (BEMI) data is addressed. A stochastic model describing the spatial distribution of clutter is developed and methods for estimating and removing this unwanted interference are described and tested. After removing the clutter from the signal, a technique is introduced for extracting from BEMI data information describing the location, orientation and structure of the buried object. Examples are provided for spherical and ellipsoidal mines.

Despite ongoing research and development efforts, landmine and unexploded ordnance infestation continues to be a problem. Remediation efforts typically utilize inexpensive handheld metal detectors that rely on the principle of electromagnetic induction but have a limited depth of sensitivity and are unable to discriminate the shape, size, depth, and material type of the detected object. Conventional metal detectors can often detect all relevant metal objects. However, when instrument thresholds are set to a high level of sensitivity, unacceptably high false alarm rates result. To address these issues, a new design for inductive array detectors is under development. This design is founded on an advanced Nondestructive Evaluation (NDE) sensor called the MWM that was originally developed at MIT and has been successfully commercialized for manufacturing and NDE applications. The MWM-Array offers several advantages: (1) high resolution imaging with deep penetration, (2) varied applied field direction across sensor footprint, (3) potential for wide bandwidth continuous wave of pulsed mode operation, (4) simultaneous sampling and parallel processing of data form sensing element arrays for rapid image building, and (5) when measurements of the sensor element responses are combined with model-based measurement grids, quantitative estimates of size and depth for known object shapes can assist in the classification and discrimination of detected objects, as well as the elimination of false detections.

Electromagnetic emissions from electronics associated with explosives - mines, timers, and fuses - have been experimentally observed. The emissions fall into two categories, those resulting from the natural functioning of the electronics themselves - an electromagnetic interference (EMI) signature, and stimulated emissions resulting from RF irradiation of the electronics and re-radiation by the circuit. In order to develop suitable detection modalities for these processes, an understanding of the basic physics of the radiation process is essential. The basic physics of radiation and stimulation of electronic circuits are being studied, and electromagnetic interference characterization and measurement procedures are being developed. Finite- difference time-domain modeling is being applied to gain insight into radiation processes and levels so that the feasibility of detection of these signatures can be evaluated, and detection systems developed.

An object, made partly or wholly of metals, has a distinct combination of electrical conductivity, magnetic permeability, and geometrical shape and size. When the object is exposed to a low-frequency electromagnetic field, it produces a secondary magnetic field. By measuring the broadband spectrum of the secondary field, we obtain a district spectral signature that may uniquely identify the object. Based on the response spectrum, we attempt to 'fingerprint' the object. This is the basic concept of EMIS. From numerous surveys that we have conducted using our multifrequency electromagnetic sensors (GEM-2 and GEM-3), we have accumulated significant evidence that a metallic object undergoes continuous changes in response as the transmitter frequency changes. These observations made over may UXO targets suggest strongly that the EMI anomaly measured in a broad band offers an ability to both detect and identify a target. The frequency-dependent structure of the difference was also reproducible and consisting over a range of depths. Therefore, we have established that the FEM-3 is capable of delivering broadband EMI data with ample target-specific information content for the purpose of target classification and identification.

Studies show that soldiers' mine detection capabilities with the PSS-12 hand-held detector are substandard and that their probabilities of detecting (PD) low-metal mines are dangerously low. Highly experienced PSS-12 operators, however, achieve PDS over 0.90 on high- and low-metal anti- tank (AT) and anti-personnel (AP) mines. Significantly, experts' detection techniques differ from conventional military PSS-12 operating procedures. We report three studies investigating whether instruction based on expert skill could bridge the observed performance gap. Basic research on human expertise has shown that instruction based on detailed scientific analyses of experts' behaviors and thought processes boosts skill acquisition dramatically. These studies tested the effects of an experimental detection training program based on knowledge and techniques learned from analysis of PSS-12 expertise. In Study I soldiers who had completed standard mine detection training participated as operators/trainees. This experiment used a pretest-posttest design. Mine simulants served as targets in testing gand training. Targets simulate d5 different mines and represented high- and low-metal AT and AP mine types. Pretest performance failed to distinguish the treatment and control groups. Both achieved very low PDs on low metal mines. Treatment-group soldiers then received approximately 15 hours of experimental, hands-on training. Posttest results showed that the treatment groups PD on minimal metal targets was more than 6 times that of the control group. Study 2 tested a subset of the treated soldiers in the same setting, now wearing body armor. Results replicated those of Study 1. Study 3 tested treatment group soldiers on real mine targets. Several mines from each mine type were used. The surface of the test lanes was expected to increase detection difficulty. Soldiers nonetheless achieved an aggregate PD of 0.97 and showed significant improvement in detecting low-metal mines.

We report preliminary 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. Each point on the screen was associated with a certain tone intensity; targets were louder than 'distractor' objects. We looked at the effects on target detection and false-alarm rates of the intensity difference between target and distractor signals, the number of distractor signals, the number of distractors and training order. The time to detect 50 percent of targets was measured by a rapid adaptive technique which generated reliable thresholds within few trials. The result are consistent with a simple model for the detection of cryptic prey by foraging predators: search was slower with more distractors, and the effect of distractors was greater when S/N ratio was lower. Although subjects got no accuracy feedback, performance improved somewhat with experience and was slightly better in the low S/N condition when it followed the high S/N condition. The procedure seems to be a useful one for studying more complex mine-related detection tasks with a range of signal types and numbers of concurrent detection signals.

Landmine detection is a complex and highly dangerous task. Most demining operations are done using hand-held detectors, which means that the operator is always at risk of serious injury or death. One of the most important factor that determines the probability of detecting is the operator performance. Therefore, it is very important that we train the operator well and are able to assess their performance accurately. To achieve these objectives, we have been developing two training tools, the 3D tracker for real-time feedback during training, and the virtual mien lane for interactive training. We have been using the 3D tracker successfully to assess the performance of an operator as a prat of a successful training program.

Although the ability of EMI sensors to detect landmines has improved significantly, false alarm rate reduction remains a challenging problem. Improvements have been achieved through development of optimal algorithms that exploit models of the underlying physics along with knowledge of clutter statistics. Moreover, experienced operators can often discriminate mines form clutter with the aid of an audio transducer. Assuming the basic information needed for discriminating landmines form clutter is largely available form existing sensors, the goal of this wok is to optimize the presentation of information to the operator and to be able to predict improved performance prior to extensive experimental testing. Traditionally, an energy calculation is provided to the sensor operator via a signal whose loudness or frequency is proportional to the energy of the received signal Our preliminary theoretical work indicated that when the statistic used to make a decision is not simply the signal energy the performance of mine detection systems can be improved dramatically. This finding suggest that the operator could make better sue of a signal that is a function of this more accurate test statistic, and that there may be information in the unprocessed sensor signal that the operator could use to effect discrimination. We then experimentally investigated the perceptual dimensions that most effectively convey the information in a sensor response to a listener using simulated data. Results indicated that by supplying the sensor response more appropriately to the listener, discrimination, as opposed to simple detection, could be achieved. In this paper we discuss an additional theoretical treatment of these experimental data in which we show that we can predict such improvements. These results are verified in a follow-on listening experiment with actual data measured from landmines.

Selection of temporal and spectral properties of the auditory signal directly affects the listener ability to detect and recognize the signal. A properly designed auditory signal for mine detection operations should be resistant to predominantly low frequency environmental masking noises and easily hear by most users. Many of whom suffer form high-frequency noise-induced hearing loss. In fact, high frequency h earing loss caused by noise exposure is the moist common occupational illness in DOD as well as the USA. Regrettably, many current hand-held mine detection systems use a high-frequency pure tone or a highly- unbalanced high-pitch complex tone to alert the user to the present of an object buried in the ground. Inability to hear the auditory signal or even small variations of the auditory signal can result in an operator failing to detect a mine, a potentially fatal operator error. Unfortunately, proper engineering of the auditory signal is not as simple as selecting another lower frequency pure tone and the use of complex stimuli many be needed. The purpose of this study was to investigate changes in detectability of a specific auditory stimulus as a function of the number of components added to the basic signal. Obtained data provide support to the doctrine that mine detector warning signals should be complex signals expanding over several octaves.

The thermal properties and shape of a buried land mine can, by natural means such as diurnal cycles, result in a temperature profile on the ground surface. By exploiting the presence of this thermal signature, IR imaging has demonstrated the ability to detect buried mine-like objects. Of importance to the practical success of this technology is the ability to obtain a spatial resolution which allows discrimination of mine signatures from background clutter. This paper describes findings from a study conducted to establish the clutter statistics of natural occurring backgrounds. A novel approach is presented: the use of 2D autoregressive models to detect the unnatural variations in the background caused by buried miens. With this knowledge we have developed a process to estimate the camera resolution necessary to reliably detect and discriminate a thermal signature originating from a buried mine-like object in various terrain types.

This paper presents the work on the detection of land mines using IR-images. Experiments have been performed where outdoor time series of IR contrast have been measured for wax filled antitank mines in sand and for real mines in a gravel road. For antitank mines in the sand box the contrast dependence of time lap between burial and measurement has been analyzed for a period of four months. The diurnal contrast variation of an anti tank mine buried for two and a half year in a gravel road has been calculated. Statistical correlation between apparent temperatures and weather parameters for different cases have been calculated. The purpose is to understand the origin of the contrast and to be able to predict the contrast at different times and under different conditions.

3D thermal and radiometric models have been developed to study the passive IR signature of a land mine buried under a rough soil surface. A finite element model is used to describe the thermal phenomena, including temporal variations, the spatial structure of the signature, and environmental effects. The Crank-Nicholson algorithm is used for time-stepping the simulation. The mine and the surroundings are approximated by pentahedral elements having linear interpolation functions. The FEM grid for the soil includes a random rough surface having a normal probability density and specified covariance function. The mine is modeled as a homogeneous body of deterministic shape having the thermal properties of TNT. Natural solar insolation and the effects of convective heat transfer are represented by linearized boundary conditions. The behavior over a periodic diurnal cycle is studied by running the simulation to steady state. Finite element solutions for the thermal emissions are combined with reflected radiometric components to predict the signatures seen by an IR camera. Numerical simulations are presented for a representative target, a 25 cm anti-tank mine simulant developed by the US Army. The temporal evolution of the temperature distribution and IR signature are presented for both smooth and rough surfaces.

An image processing technique is described for the detection of miens in RI imagery. The proposed technique is based on a third-order neural network, which processes the output of a wavelet packet transform. The technique is inherently invariant to changes in signature position, rotation and scaling. The well-known memory limitations that arise with higher-order neural networks are addressed by (1) the data compression capabilities of wavelet packets, (2) protections of the image data into a space of similar triangles, and (3) quantization of that 'triangle space'. Using these techniques, image chips of size 28 by 28, which would require 0(10⁹) neural net weights, are processed by a network having 0(10²) weights. ROC curves are presented for mine detection in real and simulated imagery.

Shallowly buried object sand the background soil have different specific heat and heat conductivity. Thus, the thermal alternation over day and night produces a thermal contrast on the soil surface that can be detected by an IR sensor. We present a simple model for the spatio-temporal temperature signature of a buried land mine by means of a 3D Gaussian function. Such a mode is appropriate since both measured data and simulations based on the finite element method show a spatio-temporal behavior that strongly resembles a 3D Gaussian shape. An advantage of modeling the signature as a Gaussian shape is that objects with approximately the same size but with different physical properties can be obtained simply from a scaled version of the original model.

Many present day vehicle-mounted landmine detection systems use IR imagers. Information furnished by these imaging systems usually consists of video and the location of targets within the video. In multisensor systems employing data fusion, there is a need to convert sensor information to a common coordinate system that all sensors share.

Statistically significant sets of buried anti-tank mine and background electro-optic spectral signatures were collected and analyzed by the Veridian ERIM International team under the US Army's Night Vision and Electronic Sensors Directorate Hyperspectral Mine Detection Phenomenology FY98/99 Program as reported last year. Those analyses established predicted buried mine spectral discrimination performance in key practical sensor spectral regions using typical multispectral sensor bandwidths spanning 20 to 200 nm. This year, we report further analyses of selected sets of HMDP data that quantitatively predict performance for two specific cases of joint spectral regions. This work exhibits these initial results and compares the predicted buried mine spectral discrimination performance determined from the joint and the single spectral regions.

Research to assess the feasibility of developing a standoff active or passive optical tripwire detector is discussed. Reflectives of typical tripwires and background materials were measured for UV, VNIR and SWIR wavelengths. A breadboard testbed was developed to obtain images of tripwires against various backgrounds for various geometries and a wide range of VU and VNIR wavelengths. Sample images of simulated and real tripwires in uncluttered environments and against typical cluttered backgrounds were acquired and analyzed. Line detection algorithms were applied to the images to detect tripwires. Although detection was not attempted in real-time, analysis showed that available, cost-effective DSPs could potentially execute those algorithms on the images in real-time. The algorithms successfully detected tripwires in a heavily cluttered background and even have the capability to detect partially obscured wires. To complement the measurements, a spreadsheet model was developed to evaluate the merits of different detectors, sources of illumination, wavebands and geometries for different scenarios. Acceptable signal-to- clutter ratios were found for a number of reasonable passive and active illumination scenarios. The study demonstrated that an optical tripwire detector is feasible in principle.

The diffuse and specular reflections of four representative scatterable anti-personnel landmines have been measured in the UV, visible and IR regions of the electromagnetic spectrum. These results are presented here and are compared to potential backgrounds in which such anti-personnel mines are likely to be sown.

In this paper the landmine detection performance of an IR and a visual light camera both equipped with a polarization filter are compared with the detection performance of these cameras without polarization filters. Sequences of images have been recorded with a rotating polarization filter in front of the cameras.

In this paper we present a method for passive stand off detection of trip wires. The concept described has the ability to detect trip wires camouflaged in a wide variety of natural backgrounds. The system includes an IR camera using the 3-5 micron band and a polarizing filter. A simple software algorithm is used to dewarp the trip wire in the scenes.

Objects hidden under a lossy medium, like soil, can be detected when the boundary between the object and the medium acts as a reflector for incoming microwave radiation. Under typical soil conditions, the maximum depth at which the object can be detected is a few wavelengths. It is therefore advantageous to employ low-frequency receivers. Under a contract from the US Army CECOM-NVESD, TRW has designed and built the Microwave Radiometer Mine Detector; a hand-held man-portable unit, which employs a cold radiometric sky as the illuminating source. The breadboard unit works at 5 Ghz using a direct RF-gain, total-power radiometer. The unit was field-tested at the Army facility at Fort AP Hill during August of 1999. The test yielded a probability of detection of 45 percent and a false alarm rate of 0.11/m².

The performance of passive microwave radiometers in detecting buried landmines is adversely affected by the presence of foreign objects and moisture in the soil. One possible way to increase the signal contrast between the mine and the surrounding soil is to make the soil artificially dry by reducing the moisture to its solid state. This can be accomplished by injecting liquid nitrogen into the top layers of the soil. This paper describes preliminary result of an experiment designed to test this approach under controlled laboratory conditions.

Passive IR sensors have been demonstrated to be effective for detecting surface landmines. For shallow buried landmines, detection rates and false-alarms rates are poorer and highly dependent on environmental conditions such as time of day and cloud cover. An advantage touted by advocates of passive microwave sensors is that their performance does not depend on the time of day and the inherent soil-mine temperature differences. To assess the complementary detection potential of passive microwave sensors, data was collected on a Thomson-Thorne microwave sensor at a test site in England. This sensor was mounted on a scanning rack apparatus and operated at a frequency of 10 Ghz. The test included investigations of both antitank and antipersonnel miens at the surface and to depths of 2 inches. Analysis of the raw data shows that surface and buried targets produce signals that are significantly higher than background clutter. In this paper, we present a brief description of the passive-microwave detection apparatus and the data-collection exercises that were completed. Analysis of the raw sensor data is then presented with an emphasis on comparing signal strengths of mines with signals reflected from the soil in the absence of mines. Particular attention is paid to the effects of varied incident sensor angles, sensor polarizations, and system scan speeds on sensor performance. Signal-to-noise as well as signal-to-clutter ratios are calculated as a function of these different variables.

An analytical study of environmental and clutter effects on microwave radiometers used for the detection of buried objects is presented. To simplify the analysis, it is assumed that the soil/target medium has a constant physical temperature versus depth, so that Kirchhoff's law can be applied to determine emissivities, and a simple layered medium geometry is used to model a buried target. Changes in brightness temperatures which result due to the present of a buried target are illustrated for varying soil dielectric properties, radiometer frequencies, and target depths, and are contrasted with changes in brightness temperatures which can occur when no target is presented due to slight soil moisture or soil temperature variations. Brightness temperature clutter due to a small surface roughness is also analytically modeled, through application of the small slope approximation for the homogeneous medium case and the small perturbation method in the presence of a subsurface layer, and it is shown that surface clutter effects can be mitigated through proper choice of sensor polarization and observation angle. Particular attention is given to the relationship between passive and active microwave sensors; results demonstrate that these two can provide complementary information. Finally, the use of wideband radiometric measurements are discussed as a means for reducing environmental clutter effects and improving detection algorithms.

In this paper, two different data sets which use linear FM incident signals with different bandwidths, namely 40 KHz and 80 KHz, are used for benchmarking. The goal is to study the effects of using larger bandwidth for underwater target classification. The classification system is formed of several subsystems including preprocessing, a subband decomposition suing wavelet packets, linear predictive coding in subbands, feature selection and neural network classifier. The classification performance is demonstrated on ten noisy realizations of the data sets formed by adding synthesized reverberation effects with 12 dB signal-to- reverberation ratio. The ROC and the error location plots for these dat sets are generated. To compare the generalization and robustness of the system on these data sets, the error and classification rate statistics are generated using Monte Carlo simulations on a large set of noisy data. The results point to the fact that the wideband sonar provides better robustness property. Three-aspect fusion is also adopted which yields almost perfect classification performance. These issues will be thoroughly studied and analyzed in this paper.

Given detection opportunities, mine hunting relies heavily upon the availability of effective and efficient classification techniques, particularly when trying to hunt mines in cluttered environments. A wide range of classification techniques is currently available. However, given the absence of a widely accepted quantitative measure of classifier performance, the best system for any particular application is not always obvious.

A new adaptive feature mapping scheme is presented in this paper to cope with environmental and target signature changes in underwater target classification. A wavelet packet-based feature extraction scheme is used in conjunction with the linear prediction coding (LPC) scheme as the front-end processor. The core of the adaptive classification system is the adaptive feature mapping sub- system that minimizes the classification error of the classifier. The extracted feature vector is mapped by the resultant feature mapping matrix in such a way that the mapped version remains invariant to the environmental and sensory changes. The feedback to the adaptation mechanism is provided by a K-nearest neighbor (K-NN) classifier. In order to alleviate problems caused by poorly scaled features, a revised K-NN based on the scaled Euclidean distance was adopted. Two error criteria were used in the adaptive system, one is the least squares (LS) error criterion and the other is 2D sigmoid cost function. Those two criteria were combined together to offer a better performance. The test results on 40KHz sigmoid cost function. Those two criteria were combined together to offer a better performance. The test result on 40KHz linear FM acoustic backscattered data collected for six different objects are presented. The effectiveness of the adaptive system vs. non- adaptive system is demonstrated when the signal-to- reverberation ratio in the environment is varying.

The problem of classification of underwater targets from the acoustic backscattered signals is considered in this paper. A wavelet packet-based feature extraction scheme is used in conjunction with the linear prediction coding scheme as the front-end-processor. Selected features with higher discriminatory power are then fed to a neural network classifier. Several different classification systems are benchmarked in this paper. These include K-nearest neighbor classifier, PNN and SVM. These paradigms are examined on the acoustic backscattered data for both 40 KHz and 80 KHz sonar bandwidth. Performance comparison of these systems with that of the previously used Back-Propagation Neural Network is provided as well.

The very shallow water regions contain much mine-like clutter, which cause high false alarm rates in minehunting sonar systems. This paper present a method of reducing false alarms by the fusion of multiple detection and classification algorithms. The algorithm fusion method is based on the well-known Fisher Discrimination.

Computing Devices Canada, a General Dynamics company, undertakes research in image processing with focus on the automatic recognition of sea mines. This paper present the use of advanced gray-scale morphological filters for this function as applied to side scan sonar imagery. Sea mines in side scan sonar imagery can be characterized by a mine-body and a mine-shadow. Mine-bodies consist of bright regions, relative to the background, with a specific shape and size. Mine-shadows consist of dark regions, relative to the background, with a specific shape and sizes. The shapes and sizes of these regions depend on the mine type, the orientation of the mine, the physical acquisition process of the sonar imagery, and the environment in which the mine is located. Advanced gray-scale morphological filters provide very powerful and robust tools to extract bright and dark regions with low signal to noise ratio in very noisy imagery using geometric constraints such as shape, size and total surface area. For the detection of sea mines we use these morphological filters with the minimum and maximum geometric constraints for the mine-bodies and mine-shadows. The independent detection of mine-bodies and mine-shadows allows the detection of bottom, moored and drifting mines with the same detection algorithm. Consistent mine-body and mine- shadow combinations are resolved into mine like objects.

In shallow-water mine countermeasure, the objective is to find underwater mines with consistently high P_D and low P_FA in diverse environmental conditions. In order to achieve this goal in a cost-effective manner, we are developing a Web-centric virtual prototyping environment that consists of various tools to make algorithm development and performance analysis as seamless as possible. Furthermore, by making algorihrtm toolboxes available for download at the project Web site, we are involving end users in the development cycle in an attempt to improve the utility and functionality of the toolboxes. This approach focuses on community collaboration and uses the Internet as a communications medium, i.e., open-source software development paradigm. We envision the virtual prototyping environment to eventually address the entire operating spectrum form algorithm development to real-time implementation by concatenating complementary toolboxes and hosting various services at a Web data center. The advantages of this approach are more cost-effective algorithm development, facilitation of accurate performance comparison between existing and new algorithms, and minimization of performance ambiguity through the use of tap points and visual information presentation. Perhaps the biggest advantage is that the environment allows researchers to spend more time on creative aspects of algorithm development and less time on mundane parts.

We describe here image processing and neural network based algorithms for detection and classification of mines in side-scan sonar imagery, and the results obtained from their application to two distinct image data bases. These algorithms evolved over a period from 1994 to the present, originally at Draper Laboratory, and currently at Alphatech Inc. The mine-detection/classification system is partitioned into an anomaly screening stage followed by a classification stage involving the calculation of features on blobs, and their input into a multilayer perceptron neural network. Particular attention is given to the selection of algorithm parameters, and training data, in order to optimize performance over the aggregate data set.

An advanced, automatic, adaptive clutter suppression, sea mine detection, classification and fusion processing string has been developed and tested with new sonar imagery data. The overall CAD/CAC string includes pre-processing, adaptive clutter filtering (ACF), normalization, detection , features extraction, classification and fusion processing blocks. The ACF is a multi-dimensional adaptive linear FIR filter, optimal in the Least Squares sense, and is applied to low- resolution data. It performs simultaneous background clutter suppression and preservation of an average peak target signature. Following 2D normalization, the detection consists of thresholding, clustering of exceedances and limiting the number of detections. Subsequently, features are extracted from high-resolution input 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 3 different research teams, are fused, using a variety of fusion rules, including logic-based and a novel orthogonal LLRT-base done. The utility of the overall processing string and their fusion was demonstrated with high-resolution side-scan sonar imagery from a difficult shallow water environment. The processing string classification performance was optimized by appropriately selecting a subset of the original feature set. The overall CAD/CAC processing string fusion result in improved mine classification capability, providing up to a four-fold false alarm rate reduction, compared to the best single CAD/CAC processing string results.

Background normalizers are used, among other things, to obtain an estimate of the local signal-to-clutter ratio (SCR), for the cell under test, under 'target present' assumptions. The classical implementation of these normalizers utilize a sliding 'split window' configuration to estimate the standard deviation of the background, with the 'split' or 'gap' in the window aiming to avoid the possibility of target signal leaking into the estimate. Unfortunately, there are tradeoffs between gap size and estimation accuracy: if the gap is too small, target signal leakage will taint the estimate; if the gap is too large, the estimate will not be representative of the background in the vicinity of the target. These drawbacks translate into reduced spatial resolution or lack of detection sensitivity.

SAR processing has revolutionized radar imaging over the past 35 years. SAR techniques are now being applied employed in sonar processing; the enabling technologies being adaptive focusing techniques imported from the SAR world. Synthetic aperture sonar (SAS) produces acoustic images with high resolution that is independent of range. Moreover, within the limits of diffraction, that resolution is also independent of frequency, thus making available a wide engineering trade-space within which sensor performance may be optimized.

Miniature flextensional transducers, called cymbals, have been incorporated into thin, lightweight, large area panels for use a slow frequency acoustic projectors in shallow water. The prototype panels, measuring 100-mm by 100-mm by 6.35-mm thick exhibit a high acoustic output at a relatively low in-water resonance frequency. Furthermore a second resonance frequency that is over an order of magnitude higher suggests that the panel may be used to provide sound output over almost a two decade frequency band. The mass of the unplotted panel is less than 150 grams and the total thickness is 6.35 mm. The cymbal panels are believed to be excellent candidates as acoustic projectors on autonomous and/or unmanned underwater vehicle platforms as well as other shallow water platforms where low frequency, light weight and high acoustic output are desired.

The purpose of this system development is to permit single pass data acquisition and increase the area coverage rate for a mine and obstacle survey of shallow water. The system would be useful in the search for other submerged targets as well. By coupling a high resolution SAR with a submerged acoustic source, a large area can be examined with a single pass of the aircraft that carries the radar. An experiment for the proof of feasibility of Radar Acoustic Hybrid is described. Experiment results are discussed and evaluated. This system is related to a predecessor that employed laser and acoustics. The trade-offs for convenience of use, resolution, and technological risk are considered so that the two system architecture options can be compared.

In this paper, the author presents the recent results of passive Broadband Bionic Sonar System in the detection and identification of underwater targets in background noise, 'acoustic daylight'. Using a resonance detection technique, various underwater objects, cylinders and spheres of different sizes and different material compositions, were detected in acoustic backgrounds noise in Kaneohe Bay, Hawaii. In addition, bottomed and buried 3-inch diameter, stainless steel sphere targets were detected in background noise. The report presents the result of many passive broadband sonar experiments in acoustic background noise.

A huge number of chemical sensors are based on electrochemical measurement methods. Particularly amperometric sensorsystems are employed for the fast detection of pollutants in industry and environment as well as for analytic systems in the medical diagnosis. The large number of different applications of electrochemical sensors is based on the high sensitivity of electrochemical methods and on the wide of possibilities to enhance the selectivity by variation of electrochemical and chemical parameters. Besides this, electrochemical sensorsystems are frequently simple to operate, transportable and cheap. Up to now the electrochemical method of cyclic voltammetry is used only seldom for sensors. Clearly the efficiency of cyclic voltammetry can be seen at the sensorsystem for the detection of nitro- and aminotoluenes in solids and waters as presented here. The potentiodynamic sensors system can be employed for the fast and easy risk estimation of contaminated areas. Because of the high sensitivity of electrochemical methods the detection of chemical substances with a low vapor pressure is possible also. The vapor pressure of TNT at room temperature is 7 ppb for instances. With a special electrochemical set-up we were able to measure TNT approximately 10 cm above a TNT-sample. In addition we were able to estimate TNT in the gaseous phase approximately 10 cm above a real plastic mine. Therefore it seems to be possible to develop an electrochemical mien detection. Moreover, we present that the electrochemical detection of RDX, HMX and chemical warfare agents is also possible.

Sensing the chemical signature emitted from the main charge explosives from buried landmines and unexploded ordnance (UXO) 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. The chemidynamic processes are fairly well understood from many years of agricultural and industrial pollution soil physics research. Due to the unique aspects of the surface soil environment, computational simulation is being used to examen the breadth of conditions that impact chemical signature transport, from the buried location to a ground surface release. To provide confidence in the information provided by simulation modeling, laboratory experiments have been conducted to provide validation of the model under well-constrained laboratory testing conditions. A soil column was constructed with soil moisture monitoring ports, a bottom porous plate to regulate the soil moisture content, and a top plenum to collect the surface flux of explosive chemicals. The humidity of the air flowing through the plenum was set at about 50 percent RH to generate an upward flux of soil moisture. A regulated flux of aqueous phase 2,4-DNT was injected into the soil at about ten percent of the upward water flux. Chemical flux was measured by sampling with solid phase microextraction devices and analysis by gas chromatography/electron capture detection. Data was compared to model results from the T2TNT code, specifically developed to evaluate the buried landmine chemical transport issues. Data and model results compare exceptionally well providing additional confidence in the simulation tool.

The chemical signature form buried landmines/UXO is affected by a number of environmental fate and transport processes in the soil such as vapor-solid and liquid-solid sorption, diffusion, biodegradation, and water movement. For shallow burial depths, land surface processes, such as wind, solar and long-wave radiation, and precipitation play an important role. The impact of these land surface processes has been evaluated for a landmine/UXO buried 5 cm below the surface using actual weather data for an entire year using the T2TNT computer code. The gas-phase concentration of the chemical signatures, which is used by most chemical sensors currently being developed, shows appreciable diurnal variation and minimum seasonal changes due to the change in the weather. The most dramatic variation in the gas-phase concentration occurs immediately after a rainfall following a long dry period. This information will impact the use of chemical sensors by indicating the best times of the day and best times of the year to sense these signatures.

The detection of buried mines is important to both for humanitarian and military strategic de-mining both at home and abroad, and recent efforts in chemical detection show promise for definitive identification of buried miens. The impact of weather has a large effect on the fate and transport of the explosives vapor that these systems sense. In many areas of military conflict, and at Army military training grounds in cold regions, winter weather affects military operations for many months of the year. In cold regions, the presence of freezing ground or a snow cover may provide increased temporary storage of the explosive, potentially leading to opportunities for more optimal sensing conditions later. This paper discusses the result of a controlled laboratory experiment to investigate explosives diffusion through snow, quantitative microscopy measurements of snow microstructure including specific surface, and verifications of our transport model using this data. In experiments measuring 1,3-DNB, 2,4-DNT and 2,4,6-TNT we determined an effective diffusion coefficient of 1.5 X 10^-6 cm²/s from measurements through isothermal sieved snow with equivalent sphere radius of 0.11 mm. Adsorption is a major factor in diffusive transport of these explosives through snow. The data was used to verify our finite element mole of explosives transport. Measurements and model results show close agreement.

Trinitrotoluene (TNT) explosives contain small amounts of dinitrotoluene (DNT). DNT exhibit a higher vapor pressure than TNT which indicates higher concentration of DNT than of TNT in the vapor phase of the explosive. Analysis of soil samples reveal extended information compared to air samples and thereby increases the probability for chemical detection. Detected substances in soil samples are TNT and related compounds. Therefore, sampling of DNT in vapor phase near the ground or soil solid phase may be an efficient approach to detect buried land mines or unexploded ordnance (UXO) containing TNT. Charges of TNT has been placed both in desiccators ane in a set of different types of soil in the laboratory. Analysis of air samples repeatedly taken in desiccators during a period of 299 days shows a perpetually higher concentration of DNT than of TNT. TNT was also placed in outdoor test beds where the presence of DNT in vapor phase near to the ground were confirmed, as well as TNT and related compounds in soil samples. In mine affected areas, air sample near to the ground over buried miens and soil sampling near the same miens were performed.

New chemiselective polymers have been developed to enhance the nitroaromatic sorption properties of coated acoustic wave (AW) devices. The sensitivity and selectivity of polymer-based sensors depends on several factors including the chemiselective coating used, the physical properties of the vapor(s) of interest, the selected transducer, and the operating conditions. Detection limits with the coated SAW sensors, tested under laboratory conditions, are determined to be < 100 parts per trillion for 2,4-dinitrotoluene. A new SAW based chemical vapor detector the NRL p-CAD has been developed with dramatically improved signal kinetics offering T⁹⁵ response times of less than 0.1 second for a wide range of organic compounds including the nerve agent simulant and agent precursor material dimethylmethylphosphonate. In addition, the NRL p-CAD system offers a rapid 2s baseline reset virtually eliminating baseline drift issues associated with changes in temperature and relative humidity. The p-CAD system has been successfully tested in both ground and unmanned aerial vehicle testing.

Our group has been involved in the design and synthesis of ultra-sensitive fluorescence sensory materials for the detection of 2,4,6 trinitrotoluene (TNT) and 2,4 dinitrotoluene (DNT). These schemes make use of a novel energy migration mechanisms to amplify the fluorescence response and have led to systems capable of rapid detection of these analytes at sub part-per-billion levels. In an effort to optimize the amplification and specificity, we have examined the nature of energy migration in our polymers systems because it is inherent in achieving amplification. polarization measurements and energy transfer studies between polymers were conducted in order to evaluate and maximize energy migration and hence TNT sensory response. The correlation of photo physical properties with molecular structure guided the synthesis of novel polymers with more discriminant optical responses. These synthetic efforts have yielded a library of sensory polymers with varying sensitivities to different analytes.

Chemiresistors are fabricated from materials that change their electrical a resistance when exposed to certain species. Composites of soluble polymers with metallic particles have shown remarkable sensitivity to many volatile organic chemicals, depending on the ability of the analyte molecules to swell the polymer matrix. These sensor can be made extremely small, operate at ambient temperatures, and require almost no power to read-out. However, the chemiresistors itself is only a part of a more complex sensor system that delivers chemical information to a user who can act on the information. We present the design, fabrication and performance of a chemiresistors array chip with four different chemiresistors materials, heaters and a temperature sensor. We also show the design and fabrication of an integrated chemical sensor array, where the electronics for measuring each chemiresistors' resistance are on the same chi with the chemiresistors films. The circuit was designed to perform several functions to make the sensor data more useful. The integrated chemiresistors' resistance are on the same chip with the chemiresistors films. The circuit was designed to perform several functions to make the senor dat more useful. The integrated chemiresistors array's small size and low power demand makes it ideal for deployment on a Sandia-developed microrobot platform.

Thin films of carbon black-organic polymer composites have been deposited across two metallic leads, with swelling- induced resistance changes of the films signaling the presence of vapors. To identify and classify vapors, arrays of such vapor sensing elements have been constructed. Each element contained a different organic polymer as the insulating phase. The differing gas-solid partition coefficients for the various polymers of the detector array produced a pattern of resistance changes that was used to classify vapors and vapor mixtures. The performance of this system towards DNT, the predominant signature in the vapor phase above land miens, has been evaluated in detail, with robust detection demonstrated in the laboratory in less than 5 s in air at DNT levels in the low ppb range.

Differences between the electron attachment reactions of thermal electrons and representive classes of organic molecules with high electron affinities were used to selectively identify nitroaromatics. The investigations showed that the reactions of thermal electrons with nitroaromatics lead to the formation of products with very low electron affinities. By contrast, other analytes with high electron affinities such as polyhalogenated organics, lead to products with high electron affinities. This difference was exploited to differentiate between nitroaromatics and polychlorinated organics with a monitoring device consisting of a tandem arrangement of two electron capture detectors connected in series with an electron capture detectors connected in series with an electron attachment reactor. The tandem ECD arrangement was used for selective determination of nitroaromatics vapors in the presence of interfering compounds at ppb and sub ppb concentrations.

MEMS technology was used to fabricate bimetallic cantilever sensor for detecting the TNT and DNT residue found in mien fields. A number of experiments yielded reproducible result for the detection of pure 2,4-Dinitrotoluene nanogram particles. A few experiments were performed unsuccessfully to detect explosives directly from soil by placing it on the cantilevers. Alternatively, DNT and TNT were extracted from the soil using acetone and subsequently letting acetone to leave DNT and TNT as a residue. This residue has been placed on the cantilever for detection that yielded very uncertain result. This residua contains a number of other materials, which changes the physical properties of the residue considerable making it unfit for detection using microcantilevers. T

We report on field test results conducted during 1999 in Bosnia and at the Army Mine Training School, Fort Leonard Wood, MO, on a ne prototype landmine detection system. In all test, non-metallic, anti-personnel (AP) and anti-tank (AT) landmines were detected via the NQR explosive signature with a probability of detection of 100 percent. The initial false alarm rate for the AP mine test was < 5 percent and was reduced to zero by a subsequent remeasurement. The test included typical burial depths and a variety of ground and weather conditions. In addition, the system can tolerate very high levels of metallic clutter and has repeatedly achieved zero false alarm rate when scanning for buried explosives at an EOD test range.

Quadrupole resonance (QR) is a technique that discriminates mines from clutter by exploiting unique properties of explosives, rather than the attributes of the mine that exist in many forms of anthropic clutter. After exciting the explosive with a properly designed electromagnetic-induction (EMI) system, one attempts to sense late-time spin echoes, which are characterized by radiation at particular frequencies. It is this narrow-band radiation that indicates the present of explosives, since this effect is not seen in most clutter, both natural and anthropic. However, explosives detection via QR is complicated by several practical issues. First, the late-time radiation is often very weak, particularly for TNT, and therefore the signal- to-noise ratio must be high for extracting the QR response. Further, the frequency at which the radiation occurs is often a strong function of the background environment, and therefore in practice the QR radiation frequency is not known a priori. Also, at frequencies of interest, there is a significant amount of background radiation, which induces radio frequency interference (RFI). In addition, the response properties of the system are sensitive to the height of the sensor above the ground, and the QR sensor effectively becomes 'de-tuned'. Finally, present QR systems cannot detect the explosive in metal-cased mines, thus the system and associated signal processing must be extended to also operate as a metal detector. Previously, we have shown that adaptive noise cancellation techniques, in particular, the least-mean-square algorithm, provide an effective means of RFI mitigation and can dramatically improve QR detection. In this paper we discuss several signal processing tools we have developed to further enhance the utility of QR explosives detection. In particular, with regard to the uncertainties concerning the background environment and sensor height, we explore statistical signal processing strategies to rigorously account for the inherent variability in these parameters.

A commercially available simulated land mine and several custom-made plastic simulants were examined at the University of Florida for their suitability in Lateral Migration Radiography (LMR) land mine detection. In 1997 x- ray LMR was used in the detection measurements of 12 actual antitank and antipersonnel miens. The resulting images indicated that not only were differences in composition between the explosive/casing and soil important, but that internal air volumes greatly increased not only the detectability, but also the discernability of actual mines. This paper explores the use of simulant mines that have internal features, including voids, in lieu of solid simulant miens for use in LMR measurements in the laboratory. Typical commercially available simulated mines have been developed for other detection methods such as those based on E and M technologies. A comparison of LMR images from these simulants and the LMR images of real mines demonstrated that commercially available simulant mines would fail when used with the LMR x-ray detection method. In contrast, simulated mines that we have fabricated with a plastic that has approximately the same electron density of TNT yield LMR images that are consistent with LMR images of actual mines.

At first glance, the surface of the earth appears as a relatively uniform solid surface. Seismic velocities for the earth reported in the literature are in the range of 1500 m/s and stated densities for the surface are near 3 g/m³. The big difference between the impedance of air and the surface of the earth suggests that airborne sound impinging on the surface should be efficiently reflected. During the early 1970s, personnel from Waterways Experiment Station found that geophones planted below the surface of the earth responded well to sound from aircraft. Measurements of ground motion with geophones and the signal form microphones buried in the soil as a function of soil type, depth, and frequency were conducted over a period of several years. The result of the experiments was recognition of how the porosity of soils affects the acoustic impedance of the surface and the acoustic to seismic coupling. The application of Biot theory to air filled soil pores allowed us to understand acoustic to seismic coupling in detail and enabled us to use acoustic measurements to determine soil properties. Determination of soil properties such as flow resistance, porosity and tortuosity, form acoustic measurements compare well to those determined from more conventional, non-acoustic methods. One interesting result of the measurements and theory was a confirmation of the local reaction description typically used for the impedance of soils.

If the ground is assumed to be rigid and porous, then classical porous material models may be used. Parameters such as flow resistivity, porosity, tortuosity, viscous and thermal characteristic dimensions or power size distribution are found to be necessary for the acoustic description of rigid porous materials. Typical values of ground flow resistivity are relatively high and permit simple approximations. The expressions for wave number, attention and surface impedance that result from this type of approach are presented and discussed. If the ground may be treated as an impedance plane then classical theory may be used to describe the field due to a point source above the ground. The results of this theory are described. It enables measurements made at a receiver close to the ground of the sound spectrum due to a nearby broad-band point source to be used to deduce the ground impedance. The free field spectrum due to the source or the field received at another vertically-separated receiver is used as a reference. The influences of layering within the ground and small-scale roughness on the surface are discussed together with the area of the ground involved for a given source-receiver geometry. Results of boundary element modeling of the influence of a buried object on excess attenuation spectra and the feasibility of using acoustic impedance deduction for mine detection are reviewed.

Measurements of a soil surface using a laser Doppler vibrometer are associated with the vibration velocity of the solid particles. Therefore, to model these measurements the deformation of the solid granular frame must be described. To properly account for the coupling of sound into the earth's surface it must be modeled as a porous medium. One model described wave propagation through porous materials with a deformable framework was developed by Biot. Poro- elastic material, described by Biot, can support two dilatational waves and one rotational wave. The dilatational waves are usually referred to as fast, or type I, waves and slow, or type II, waves. These waves deform both the solid and fluid components as they propagation. An overview of the Biot poro-elastic model is presented. Laboratory measurements on an air-filled unconsolidated packing of sand, are discussed to illustrate the predicted behavior of poro-elastic materials. The sand was excited using an acoustic wave from an air-borne source. The transmitted waves were detected using geophones and microphones buried within the sand. These measurements are compared to those using a mechanical shaker in contact with the surface.

The ratio of the surface soil particle velocity to the surface acoustic pressure is termed the acoustic to seismic transfer function. Measurements of this transfer function typically show several maximum and minimum in the frequency range between 50-500 Hz. The magnitude of this transfer function can be explained in light of the porous nature of the ground surface .The ground is modeled as a poro-elastic layer overlying a non-porous substrate. The boundary conditions at the air/porous soil and the porous soil/non- porous substrate interfaces are applied to setup the acoustic-to-seismic coupling problem. In the porous layer, up an downing going Biot Type I, II compressional and shear plane waves are allowed. In the non-porous elastic substrate down going compressional and shear plane waves are allowed. Using the Biot characteristics equations and these boundary conditions the steady state frequency dependent acoustic to seismic transfer function is calculated. Layer depths, Type I, and shear wave speeds are determined from a shallow seismic refraction survey. Soil density, air porosity and permeability are determined from other measurements. The calculated transfer functions are compared to that measured on several outdoor grounds.

During the early 1980s, the phenomenon of acoustic-to- seismic coupling was used to detect buried objects or mines. In these early measurements, large 2 Hz geophones measured the low frequency normal component of the soil particle velocity over buried targets. Several different, naturally- occurring ground types were studied in these measurement, including grass-covered ground; bare, sandy soil surfaces; and 'dirt' roads. Since the large geophone averages the particle velocity over the area of the sensor case, acoustic-to-seismic transfer function measurements were made with new, smaller-sized geophones. Higher frequency measurements were made using accelerometers. 3D maps of the surface particle velocity were made using measured seismic/acoustic transfer function data. Recognizing the need for a non-contact sensor and the need to investigate the geophone/soil coupling effect in the acoustic-to-seismic transfer function, additional measurements were made using a laser Doppler vibrometer (LDV). This paper explains the acoustic-to-seismic coupling mine detection measurement technique using both geophones and an LDV. The early measurements of the acoustic-to-seismic coupling transfer function for mine-like targets are discussed as well as some recent measurements using a LDV.

In March of 1999, a research team from the University of Mississippi brought its data acquisition system consisting of an acoustic/seismic laser Doppler vibrometer (LDV) mine detection sensor, to Fort A P Hill in Virginia. The purpose was to collect data over a variety of miens and to participate in a blind test. IN the blind test, the mine detection apparatus was brought to several 1-m by 1-m areas included a mix of mines, blank spots., and clutter spots as determined from prior test. The data collected over each of these spots was visualized in real time, an a mine/no mine decision was made. The resultant probability of detection was 95 percent with a false-alarm rate (FAR) of 0.03 m^-3. We present a description of the test and a detailed analysis of the data collected by the University of Mississippi in the mine lanes at AP Hill. With knowledge of the baseline, we compute target and clutter statistics, including signal-to-clutter ratios for various categories of mine types and mine depths. We examine detection trends as a function of frequency. Applying image-processing techniques to the data, features such as size and shape are extracted, and the resultant feature-level target and clutter histograms are used to improve performance. The expected performance with a without feature is compared to the demonstrated performance.

A recent blind test and two data collections at the US Army mien test lanes at Ft AP Hill have demonstrated the great potential for the use of acoustic technology to detect buried land mines. The acoustic system built by the University of Mississippi under a contract with the Night Vision and Electronic Sensors Directorate demonstrated a very high probability of detection, a very low false alarm rate, extremely good location accuracy, and significant standoff potential. A large number of papers are being presented at this conference that deal with various specific aspects of this program. This paper will present a broad but technical overview of this program. We will describe the capabilities of this approach and the areas in which improvements are being addressed. We will discuss briefly fusion with additional sensors, which will illustrate the manner in which acoustic technology can be integrate with other sensor to form a viable and robust mine detection system. We will present the present Army requirements and operational concepts that would meet these requirements.

A hybrid technique has been developed that uses both electromagnetic and elastic waves in a synergistic manner to detect buried land mines. The system consists of a moving electromagnetic radar and a stationary elastic-wave source. The source generates elastic waves in the earth. These waves interact with the buried mine and cause both the mine and the earth to be displaced. Because the mechanical properties of the mine are different from those of the earth, the displacements in the region of interaction are distinct from those associated with the free-field propagation of the waves. The radar is used to detect displacement and, thus, the mine. Initial investigations have demonstrated the feasibility of this scheme under controlled conditions. The current experimental effort if focused on understanding and overcoming the issues associated with using the system in field conditions.

A 3D finite-difference model for elastic waves in the ground has been developed and implemented. The model is used to investigate the interaction of elastic waves with buried land miens. When elastic waves interact with a buried mine, a strong resonance occurs at the mine location. The resonance can be used to enhance the mine's signature and to distinguish the mine from clutter. Results are presented for a single mine buried in the ground and several miens in the presence of clutter. The predictions of the numerical model are in fairly good agreement with experimental results.

A near-field beamforming array is investigated for use in a radar system that is part of a hybrid elastic/electromagnetic technique for detecting land miens. The radar is used to measure the displacement of the surface of the earth and land miens due to elastic waves in the earth. The beamforming array is used to obtain a sufficiently small spatial resolution for the measurement of the displacement while allowing an adequate standoff distance for the radar. Both theoretical and experimental models are developed to investigate the viability of the beamforming array.

This acoustic mine detection system uses an acoustic array of hydrophones embedded within a unique fluid-coupling structure that deforms to the ground contours and has an acoustic impedance comparable to that of the ground to facilitate energy transfer and eliminate losses at the air ground interface. Broadband and impulsive acoustic array techniques are used to localize buried objects and interpret the buried object's surrounding. The goal of this system is a low-cost, hand-held mine detector that rolls or slides across the ground, suitable for a soldier ti inspect and clear a two-foot wide path. The array contains sensor and sound source, which send out various acoustic waveforms and analyzes the returning echoes and emissions to determine if an object buried below the surface has affected the propagating sound. The sensor on the array remain in a fixed linear geometry hovering over the ground to facilitate beamforming while eliminating the huge losses associated with coupling airborne sounds to the ground. Reflections at material discontinuities, as well as mine shape, materials, and depth contribute to the variations of the induced and resultant sound field. Preliminary data is presented that shows detections of underground objects, and a discussion of future efforts, to include further processing of the data introduced in this report.

The use of acoustic-to-seismic coupling to detect buried landmines has been successfully demonstrated over the past year. The technique uses a laser Doppler vibrometer (LDV) to measure the velocity of the ground vibration as it is being sonified. As it is currently implemented, the LDV scans individual points on the ground. The technique shows much promise, but it is slow when compared to some other techniques. This work investigates the feasibility of acquiring data with the LDV as the beam moves continuously across the ground. Simple models were developed and experiments were performed to explain the cause of this noises. These result are presented and the feasibility of the approach is discussed. It has been shown that this approach is possible, but that the continuous scanning process introduces noise into the data.

Probability as logic is used to estimate the surface velocity of a patch of soil driven by an incident acoustic wave. The data used by the estimation procedure is obtained from a laser Doppler vibrometer (LDV). The output of the LDV is an intermediate-frequency carrier that is frequency- modulated by the soil surface velocity. Additionally, the LDV output is amplitude modulated by an undesirable variation in the returned laser signal due to dynamic optical speckle. The effect of the amplitude modulated by an undesirable variation in the returned laser signal due to dynamical optical speckle. The effect of the amplitude modulation on the estimate of the soil surface velocity is illustrated with results obtained using the Markov chain Monte Carlo method.

An acoustics-based system has recently proved successful at detecting buried land mines. The present paper describes the use of this land mine detection system to discern shapes of buried objects. Steel plate targets of three shapes were used: circle, square, and equilateral triangle, each buried in sand with their major surface horizontal. In each case, for certain frequency bands, when a color-scaled spatial distribution of particle velocity amplitude is displayed in real time, the target shape is clearly visible. Calculations are made using a simplistic theoretical model in an effort to understand the frequency dependence of the experimental result. For each target, wave scattering is crudely mimicked by calculating the radiant pattern in an infinite fluid from a simple source distribution of the same shape as the target and visualizing its interference with plane incident wave. Limited qualitative understanding of experimental result is obtained with this crude mode, but the need for a more realistic scattering calculation is indicated.

Mechanisms for the production of acoustic energy in soil by pulsed CO₂ laser excitation of the surface are reported. When the laser pulse in unfocused with a spot size about 1 cm in diameter, a single narrow acoustic pulse is observed with a spectral content near the detector limit of 100 kHz and a velocity of 255 m/s, close to the speed of sound in air. Whenthe laser is focused to a spot size on the order of 1 mm diameter, the audible acoustic intensity in greatly increased and we observe a second broad acoustic feature. This feature has a much lower frequency and velocity. We have tentatively identified the fast mode as a normal compressive mode and the slow mode as a Biot slow-wave. A study of visible light emission when the focused CO₂ laser beam strikes the sand surface indicates ionized nitrogen, oxygen, and silicon are present. This implies that the mechanism for sound production with the focused beam involves ionization by the optical electric field, expansion, and subsequence collapse of the air. The mechanisms for sound production by the unfocused beam, which produces better imaging of underground objects, appears to be quite different.

Acoustic sensing shows promise for the detection of buried landmines. One of us has previously demonstrated successful imaging of mine simulants buried at depths from the surface to 15 centimeters, using speakers and a laser vibrometer, which collects spectral data at low frequencies. The strength of the method is in the contrast between the porous soil and the nonporous mine, while the limitations are the strong attenuation of the probing acoustic wave and coupling of the sound directly into the vibrometer.

Acoustic waves can be a viable tool for the detection and identification of land mines and unexploded ordnance. 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 model such acoustic attention, we propose a comprehensive model to simulate the acoustic wave interactions with land mines and soils based on the Biot theory for poroelastic media. The finite-difference time-domain method is then used to solve the Biot equations. For the truncation of the computational domain in the FDTD method, we extend the acoustic and elastic perfectly matched layer (PML) to poroelastic media. Numerical experiments show that, with only 10 cells of PML medium, a high attenuation of about 50 dB can be achieved for outgoing waves. The numerical model is validated by comparison with analytical solutions. Unlike the pure elastic wave mode, this efficient PML-FDTD model for poroelastic media incorporates the interactions of waves and the fluid-saturated variation with offset in three different ground media: dry sand, fully water saturated sand and partly water saturated sand. The interaction of elastic wave with a plastic mine buried in dry sand ins simulated. The results show that the surface wave is significantly influenced by the existence of a mine-like object. The diffraction of the surface wave can serve as an acoustic target signature.

The idea of using acoustically induced Doppler spectra as a means for target detection and identification is introduced. To demonstrate feasibility of such a technique, an analytical solution for the calculation of the bistatic scattered Doppler spectrum from an acoustically excite, vibrating dielectric circular cylinder is presented. In this paper, the incident plane wave is assumed to be polarized along the axis of the cylinder is presented. In this paper, the incident plane wave is assumed to be polarized along the axis of the cylinder. A perturbation method is developed to calculate the electromagnetic scattering from a slightly deformed and inhomogeneous dielectric cylinder. Then, assuming the vibration frequency is much smaller than the frequency of the incident electromagnetic wave, a closed form expression for the time-frequency response of the bistatic scattered field is obtained. The solution for acoustic scattering from a solid elastic cylinder is applied to give the displacement on the surface as well as the compression and dilation within the cylinder. Both the surface displacement and the variation in material density within the cylinder contribute to the Doppler component of the of the electromagnetic scattered field. Results indicate that the scattered Doppler frequencies correspond to the mechanical vibration frequencies of the cylinder, and the sidelobe Doppler spectrum level is, to the first order, linearly proportional to the degree of deformation and is a function of bistatic angle. Moreover, the deformation in the cylinder, and thus the Doppler sidelobe level, only becomes sizeable near frequencies of normal modes of free vibration in the cylinder. These resonant frequencies are found to depend only on the object properties and are independent of the surrounding medium. Utilizing the information in the scattered Doppler spectrum could provide an effective means of buried object identification, where acoustic waves are used to excite the mechanical resonances of a buried object.

We have developed, using both partial differential equation approaches and integral equation formulations, a precise method to invert acoustic or electromagnetic scattering data from macroscopic concealed objects. Our approach makes use of the ideas associated with our exact solution of partial differential equations as described in our paper where we were able to collapse the number of equations by elimination of transcendentals therefore preserving the absolute mathematical precision inherent in the partial differential equation formulation. Our mathematical method, as a consequence, has not encountered the traditional loss of precision when inverting the scattered data. The unrestricted wavelength range allows us to penetrate any material which may surround the object and differentiate between the object and the media. For this reason we have applied our inversion scheme to landmine detection as we can penetrate and differentiate under both wet and dry conditions. Also, we are able to account, under certain conditions, for dielectric nonlinearities of material in the concealed object. Therefore, we are able to build in density dependent false colors a 3D grid representative of both the media and of the embedded object including the internal structure of the object. We have surveyed the literature on the subject of recovery of physical location of concealed objects and we have found that most of the present applications such as land mine detection, and we have found that most of the present applications have shortcomings due to the physical changes that are present in the surrounding media or the discontinuities of physical properties of the media. For all the above reasons we believe that we may have the most versatile and mathematically precise approach to the solution of this problem.

To acquire detection performance required for an operational system for the detection of anti-personnel landmines, it is necessary to use multiple sensor and sensor-fusion techniques. This paper describes five decision-level sensor- fusion techniques and their common optimization method. The performance of the sensor-fusion techniques is evaluated by means of Receiver Operator Characteristics curves. These techniques are tested on an outdoor test facility. Three of four test lanes of this facility are used as training set and the fourth is used as evaluation set. The detection performance of naive Bayes, Dempster-Shafer, voting and linear discriminant are very similar on both the training and the evaluation set. This is probably caused by the flexibility of the sensor-fusion techniques resulting into similar optimal solutions independent of the fusion technique.

A method is described to improve the performance of sensor fusion algorithms. Data sets available for training fusion algorithms are often smaller than described, since the sensor suite used for data acquisition is always limited by the slowest, least reliable sensor. In addition, the fusion process expands the dimension of the data, which increases the requirement for training data. By using structural risk minimization, a technique of statistical learning theory, a classifier of optimal complexity can be obtained, leading to improved performance. A technique for jointly optimizing the local decision thresholds is also described for hard- decision fusion. The procedure is demonstrated for EMI, GPR and MWIR data acquired at the US Army mine lanes at Fort AP Hill, VA, Site 71A. It is shown that fusion of features, soft decisions, and hard decisions each yield improved performance with respect to the individual sensors. Fusion decreases the overall error rate from roughly 20 percent for the best single sensor to roughly 10 percent for the best fused result.

Traditional algorithms for UXO remediation experience severe difficulties distinguishing buried targets from anthropic clutter, and in most cases UXO items are found among extensive surface clutter and shrapnel from ordnance operations. These problems render site mediation a very slow, labor intensive, and efficient process. While sensors have improved significantly over the past several years in their ability to detect conducting and/or permeable targets, reduction of the false alarm rate has proven to be a significantly more challenging problem. Our work has focused on the development of optimal signal processing algorithms that rigorously incorporate the underlying physics characteristics of the sensor and the anticipated UXO target in order to address the false alarm issue. In this paper, we describe several techniques for discriminating targets from clutter that have been applied to data obtained with the Multi-sensor Towed Array Detection System (MTADS) that has been developed by the Naval Research Laboratory. MTADS includes both EMI and magnetometer sensors. We describe a variety of signal processing techniques which incorporate physics-based models that have been applied to the data measured by MTADS during field demonstrations. We will compare and contrast the performance of the various algorithms as well as discussing tradeoffs, such as training requirements. The result of this analysis quantify the utility of fusing magnetometer and EMI dat. For example, the JPG-IV test, at the False Positive level obtained by NRL, one of our algorithms achieved a 13 percent improvement in True Positive level over the algorithm traditionally used for processing MTADS data.

The MH/K) Close in Detector (CID) system employs ground penetrating radar (GPR), forward looking IR and metal detectors that are individually and collectively processed to generate automatic target detections. The detection preprocessing is initially being accomplished on a sensor by sensor basis. For the IR sensor, we apply digital filtering techniques and morphology to detect the mines and a separate filter to characterize the background level. For preprocessing the GPR returns, we apply an inverse Fourier transform to the complex frequency return signal to obtain depth information and apply digital filtering techniques to remove fixed pattern noise and provide contrast enhancement. Metal detector returns are preprocessed using a distance measure of the return compared with the averaged background. A representative data set is extracted from the preprocessed data for each of the sensor types. Other MH/K team efforts for ATR and fusion development include TRW, BAE and Sandia National Laboratories.

This paper quantifies the mine detection performance by fusing ground penetrating radars and a metal detector. Specifically, the fusion scheme used in this paper is ANDing different sensors with high probability of detection regardless of the false alarm rate. As the false alarms are random, and each sensor processes detected objects differently to produce high probability of detection, fusion by ANDing eliminates the majority of false alarms, and hopefully maintains the high probability of detection based on the mutually exclusive property of the sensor being fused. This paper uses data collected with different GPR's of Vehicular Mounted Mine Detection ATD systems and a handheld metal detector at Aberdeen Testing Center, Maryland and Socorro, New Mexico test sites. The total number of mines encountered and area coverage are approximately 450 and 13000m², respectively.

In this paper, a statistical technique for multi-sensor data fusion is applied to countermine problems. The fusion method has previously been shown to be a powerful tool in SAR/ATR applications. An earlier study, using data from the separate x, y and z coils of the EM61-3D metal detector collected at the Seabee site at Ft. Carson, demonstrated that the method shows promise for countermine applications as well. This paper briefly reviews the mathematical foundation of the fusion technique and then a new application to multi-sensor data for the Mine Hunter/Killer system is discussed at length. The sensor suite includes a ground-penetrating radar, metal detectors, and an IR camera; data were collected at Fort AP Hill, VA. The results of applying the probabilistic fusion method to these multi-sensor data are present and analyzed in detail for the first time. The advantages and limitations of using this fusion approach for countermine applications are discussed.

In this paper, an architecture for multisensor data fusion and evidence accumulation for landmine detection and discrimination is presented. Evidential and discriminatory information about the buried object such as shape, size, depth, and material, chemical or electromagnetic properties is obtained from different sensor and sensor algorithms. A streamlined assimilation of these varied information from dissimilar and non-homogenous sensor and sensor algorithms is presented. Information theory based pre-processing of the data and subsequent unsupervised clustering using Dignet architecture is used to capture the underlying structure of the information available from different sensors. Sensor information is categorized into type, size, depth, and position data channels. Each sensor may provide one or more of this information. Type data channel provides any relevant discriminatory characteristics of the buried object. A supervised feed-forward neural network is used to learn the causality between the cluster information and the evidence of a given class of the buried object. Size, depth and phenomenology input are used as control gating input for the neural network mapping. The supervisory feedback is provided by the output of the global sensor fusion system and accommodates both autonomous and human assisted learning. Dempster-Shafer evidential reasoning is used to accumulate different evidence from sensor channels and thus to detect and discriminate between different types of buried landmine and clutter. Performance of fusion architecture and Dempster-Shafer reasoning is studied using simulated data. For the simulated data noisy images of regular and irregular shapes of different objects are produced. Fourier descriptor, moment invariant and Matlab shape features are used to define the shape information of the objects. Evidence accumulation is done using shape and size information form each of the algorithms.

Mine signature measurements have been performed at the experimental facilities of the Joint Research Center of the European Commission, including far-field radar measurements in the European Microwave Signature Laboratory, thermal IR measurements in the Radiometry Laboratory and the European Goniometric Facility, and metal detector measurements. The targets were reference objects, army training surrogate anti-personnel mines, US Army simulants and mine-like objects. The environments were sand and soil with various moisture levels, gravel, mixtures of gravel and soil with and without vegetation cover.

This paper present the quantitative detection performance of buried mine reflectance signature sin various soils and burial durations. The spectral signatures including the distribution representing class separation between mines and background is performed. The quantified detection performance is based on single hypothesis test using the distance measure and the thresholding method. This paper uses a subset of the data collected under the Night Vision and Electronic Sensors Directorate sponsored ERIM Hyperspectral Mine Detection Phenomenology data collections. The dat set used was collected with an Analytical Spectral Devices Field Spectrometer at Ft. Carson, and contains about 700 mine, and 600 background signatures with hundreds of bands extending from .35 to 2.5 micrometers .

Many statistical signal processing approaches to target detection and classification assume the measurement is corrupted by independent, identically distributed white Gaussian noise. This common assumption often results in simpler, and less computationally, intense, mathematically realization for the processor. However, in many instances it is not clear if this assumptions regarding the statistics of the noise is valid. In this paper, the effects of assuming i.i.d. white Gaussian noise on the performance of likelihood ratio detectors and maximum likelihood classifiers implemented in a non-Gaussian noise environment are discussed. If the assumptions regarding the noise distribution are accurate, the resulting likelihood ratio detector and classifier are optimal. However, if those assumptions are inaccurate, performance may be degraded. We present simulation result illustrating the effects of mismatch between the assumed and actual noise distributions on detection and classification performance for likelihood ratio processors derived under several assumptions regarding the noise distribution. Specifically, target detection and classification utilizing electromagnetic induction sensors is considered.

Recent research sponsored by the Army, Navy and DARPA has significantly advanced the sensor technologies for mine detection. Several innovative sensor systems have been developed and prototypes were built to investigate their performance in practice. Most of the research has been focused on hardware design. However, in order for the systems to be in wide use instead of in limited use by a small group of well-trained experts, an automatic process for mine detection is needed to make the final decision process on mine vs. no mine easier and more straightforward. In this paper, we describe an automatic mine detection process consisting of three stage, (1) signal enhancement, (2) pixel-level mine detection, and (3) object-level mine detection. The final output of the system is a confidence measure that quantifies the presence of a mine. The resulting system was applied to real data collected using radar and acoustic technologies.

A new hybrid algorithm, based on combining the decorrelating and packing qualitites of Principal Component (PC) analysis and the shape extracting and filtering properties of Mathematical Morphology, is investigated in the frame-work of land mien detection. The new method is similar in spirit to the MM-MNF algorithm, which is based on a linear pre- filter, followed by a morphological multispectral detection component (MM). The new filter (PC-MM), has a similar concatenated structure, and addresses some of the weaknesses inherent in the linear component of the MM-MNF algorithm; namely, the susceptibility of the MNF transform to clutter inhomogeneity, as well as to variation sin clutter covariance estimation. The PC-MM algorithm addresses the stationarity problem by solely operating on image peaks extracted by a morphological top-hat transform. Therefore, the algorithm is much less susceptible to the present of different textural regions. Subsequently, the peaks in the extracted multispectral top-het image are projected into uncorrelated bands using the principal component (PC) transform. Due to the packing property of the PC transform, the target markers are typically found in the first and second bands in the PC transformed image. The targets are then detected using a variant of the morphological detection scheme. The new method provides a fast and satisfactory first-pass detection result, for images of different clutter homogeneities and target types. The extracted targets, from the first pass, are then issued to improve the detection result in a subsequent iteration, by updating covariance estimates of relevant filter variables.

A differential Global Position System (DGPS) is one of the capabilities that has been used in Countermine testing and operation by the vehicular mounted mine detector system to record the locations of the detected mines. During the system's data acquisition, the DGPS may incur the measurement errors due to measurement noises and other sources of inaccuracy that result in mis-positioning of the detected mines. The position errors, however, could be reduced if the Kalman filtering algorithm is implemented in the system. To optimally reduce these errors, the fixed- interval smoother could also be used if a small interval of processing time is allowed during the countermine testing or operation via the use of standoff radar or the forward looking IR sensor. These improvements will enhance ground registration of the candidate mines. Thus, data of the same mine that have been collected from multiple sensor or from the multiple looks by a single sensor can be associated correctly. The correct data association will provide better data fusion that improves the probability of detection and the false alarm rate. Kalman filtering and fixed-interval smoothing algorithms are developed and implemented, and the results of their applications to real test data are presented.

We numerically examine subsurface sensing via an ultra- wideband ground penetrating radar system. The target is assumed to reside under a randomly rough air-ground interface ,and is illuminated by a pulsed plane wave. The underlying wave physics is addressed through application of the multiresolution time-domain algorithm. The scattered time-domain fields are parametrized as a random process and an optimal detection scheme is formulated, accounting for the clutter and target signature statistics. Detector performance is evaluated via receiver operating characteristics, for variable sensor parameters and for several rough-surface statistical models.

We consider angle-dependent scattering form a general target, for which the scattered signal is a non-stationary function of the target-sensor orientation. A statistical model is presented for the wavelet coefficients of such a signal, in which the angular non-stationary is characterized by an 'outer' hidden Markov model. The statistics of the wavelet coefficients, within a state of the outer HMM, are characterized by a second, 'inner' HMM, exploiting the tree structure of the wavelet decomposition. This dual-HMM construct is demonstrated by considering multi-aspect target identification using measured acoustic scattering data.

The finite difference time domain technique, FDTD, is used to calculate the scattered field din the n era zone from 1D random rough surfaces. Different statistics for the random surface will be assumed in this work. First, the random rough surface will be characterized by one-scale roughness with Gaussian distribution for the heights and Gaussian auto-correlation function. In the second part, the surface will be assumed to have two-scale roughness with the same Gaussian statistics as before. The statistics of the scattered fields are calculated in this work using Monte Carlo simulations. Numerical results comparing scattered fields from one-scale roughness and two-scale roughness are shown. The results obtained indicate that the distortion in the scattered signals is primarily due to the small-scale roughness while the two-scale roughness causes more time delay In the scattered signals. Different rough surface parameters will be used to quantify their effect on the statistics of scattered signals.

An ultra-wideband frequency modulated continuous wave (FMCW) radar was used to detect plastic anti-personnel (AP) mines. The study was conducted sign AP mines that were flush buried in a test bed prepared form crushed gravel. The ultra- wideband radar resolved the signals reflected from the top and bottom surfaces of the AP mines. It was not possible to detect these mines using the surface reflection as a detection threshold because of the high ground clutter. However, the ability to detect these mines improved greatly when the sub-surface reflections were used as a detection threshold. The study demonstrated that an ultra-wideband FMCW radar can be used to reject the ground clutter and to detect the AP mines buried at shallow depth.

Recent development of wideband, high-resolution SAR technology has shown that detecting buried targets over large open areas may be possible. Ground clutter and soil type are tow limiting factor influencing the practicality of using wideband SAR for wide-area target detection. In particular, the presence of strong ground clutter because of the unevenness, roughness or inconsistency of the soil itself may limit the radar's capability to resolve the target from the clutter. Likewise, the soil material properties can also play a major tole. The incident wave may experience significant attenuation as the wave penetrates lossy soil. In an attempt to more fully characterize this problem, fully polarimetric ultra-wideband measurements have been taken by the US Army Research Laboratory's SAR at test sites in Yuma, Arizona, and Elgin Air Force Base, Florida. SAR images have been generated for above-ground and subsurface unexploded ordnance targets, including 155-mm shells. Additionally, a full-wave method of moments (MoM) model has been developed for the electromagnetic scattering from these same targets, accounting for the lossy nature and frequency dependency of the various soils. An approximate model based on phys9cal optics (PO) has also been developed. The efficacy of using PO in lieu of the MoM to generate the electromagnetic scattering data is examined. We compare SAR images from the measured data with images produced by the MoM and PO simulations by using a standard back-projection technique.

The complex dielectric constant of the soil surrounding a land mien and its contrast with the dielectric constant of the landmine are critical to the effectiveness of ground penetrating radar (GPR) for landmine detection. These parameters affect the velocity and attenuation of the radar signal as well as the strength of the reflection form the mine. The dielectric properties of the soil depend on the soil texture and bulk density as well as the soil water content. In previous work, we have simulated the unsaturated water flow around a landmine. In this paper we summarize a collection of models that can be used to predict the dielectric constant, velocity of the GPR signal, attenuation, and reflection coefficient form soil type and soil water content. These models have been integrated into a MATLAB software package. Using these models, we can determine whether or not field conditions are appropriate for use of GPR. Under dry conditions, the soil water content may be too low for good GPR performance. If the soil is too dry, we can select an appropriate level of soil water content and design a watering scheme to bring the soil water content up the desired level. We present a case study in which a soil watering scheme was designed, simulated, and the performed at a field site.

In order to detect anti-tank mines in noisy backgrounds, we have developed a ground penetrating SAR. The system operates over the frequency band 500 MHz to 1.8 GHZ. Our GPSAR system uses multiple transmit and receive antennas to acquire stepped-frequency data at 26 cross-track focal locations each separated by 1.38 inches. System motion is used to achieve along track data sampling. Multiple radar channels and high-speed radio frequency switching techniques are used to accelerate the data acquisition process, thereby increasing the system scan rate. Synthetic aperture, nearfield beamforming techniques are used to reduce clutter. The system is optimized for mine detection but is also capable of detecting deeper objects. Test against actual miens on US Army mine lanes indicate that the system can detect both plastic and metallic anti tank mines as well as anti-personnel mines. Images and analysis of data from these test are presented.

Ground-reflected clutter is often a performance-limiting factor in ground-penetrating radar detection of near-surface targets including anti-personnel mines. When a down-looking antenna is scanned across the surface this reflection produces a strong band in the image, which obscures shallow targets. Imperfections in the system impulse response can produce similar bands. Radar images of buried targets can be degraded by these forms of clutter.

An ideal sensor for mine detection would not only provide sufficient information to detect subsurface objects, but also to identify the object as a mine or a clutter object. Without this ability, the sensor will suffer from an unacceptably high false alarm rate in a high clutter environment. This capability to discriminate can be accomplished by taking into account the fact that most landmines have vertical planes of symmetry that are not found in other natural objects. There are specific scattering responses from a symmetric object that are independent of size, shape, dielectric or target depth. A target identification technique has been developed to exploit this feature using sets of bistatic and multipolarimetric data collected by GPR. The data sets are grouped into various subarrays and synthesized into a subsurface image using a 3D FDTD-based Multipolarimetric Reverse-Time Migration imaging techniques. These subarrays will be combined to from multiple mirrored elements to detect the presence of vertical symmetry. A measurement method was developed to represent numerically the level of symmetry associated with the subsurface object. These sets of numerical values, together with the total scattering response, can be used to from a false color image that locates and identifies the presence of a mine. Simulated results demonstrate the potential of this method.

This paper describes a novel approach to detecting shape features of non-metallic anti-personnel land mine using certain attributes of wavelet packets and wavelets in general. Data sources included 2GHz and 10 GHz pulse ground penetrating radar (GPR) data, and 6 GHz stepped-frequency GPR data, from laboratory measurements. Targets are chosen to include rocks and other non-lethal clutter which normally present false alarms to the GPR. The GPR signals are first de-noised with an adaptive wavelet packet de-noising method, then the shape features of the APL are determined form exploiting the regularity properties of wavelets. Results are shown which indicate that the method may be applied to pulse GPR processing, but not for frequency-stepped GPRs. Preprocessing the GPR signal for clutter reduction may not always be necessary, thus simplifying the detection process.

We apply high-dimensional analysis of variance (HANOVA) and sequential probability ratio test (SPRT) to detect buried land mines from array ground penetrating radar (GPR) measurements. The GPR array surveys a region of interest in a progressive manner starting at a known position and moving step by step in a fixed direction. Our detection method consists of two stages. Because, at each stop of the array the path lengths are different from every transmitter/receiver pair to a mien target, there exists statistically significant difference among received signals when a mine target is presented. Thus, the first step in our processing consists of a HANOVA test to detect this statistical difference at each stop. HANOVA does not incorporate new data as the GPR array moves down-track. So secondly, we resort to a sequential probability ratio test to look for changes in the HANOVA statistics as the array proceeds down track. The SPRT allows for real-time processing as anew data are obtained by the GPR array. Finally, real sensor data are processed to verify the method.

In this paper, we examine the conical spiral antenna (CSA) for use in ground-penetrating radars (GPR). When this antenna is isolated in free space, its performance, e.g., input impedance, is nearly frequency independent over a broad range of frequencies, and it has a nearly unidirectional far-field pattern with circular polarization. An antenna with these characteristics would be useful for application in GPRs. However, in a GPR the antenna is placed close to the ground, and the near field of the antenna in the air and ground is of primary importance, not the far field. It is not clear that the desirable characteristics of the CSA in free space, mentioned above, will be preserve din this new situation. This is the motivation for the research described in this paper.

To date, most of the vehicular-mounted mine detection systems employing ground-penetrating radar are down looking in the sense that the array of radar antennas is approximately 1-m forward of the vehicle and pointed straight down. Advantages of systems that are able to look forward of the vehicle by more than 10 m include the ability to make detections at greater stand-off distances and to use mulitpe looks at targets to discriminate mines from clutter. Data collected by Jaycor's forward-looking ground- penetrating radar (FLGPR) system provides a means by which these advantages can be assessed. In February 1999, Jaycor took, its FLGPR to the antitank (AT) mine lanes at Socorro, New Mexico. Jaycor made several excursions over simulated roads that contained a mix of metal- and plastic-cased AT mines on the surface and buried up to 4 in.

In this paper, we investigate the feasibility and effectiveness of using forward-imaging ultra-wideband radar technology for vehicular-based mine detection. A synthetic aperture is formed by moving the radar form side to side while the vehicle moves forward. In addition, the changing depression angle to the image area from the forward motion of the radar allows for depth resolution. Consequently, we generate 3D images of buried mines using a back propagation approach and a method-of-moments technique to model the backscatter from buried M20 mines. We developed software to simulate different imaging scenarios in order to define system limitations and optimize parameter values. For example, we generate imagery using different radar paths, and determine the effects of side-to-side sweep rate and vehicle speed on 3D resolution and overall system performance. Simulation results will be presented along with system design recommendations.

In this paper, we report on the high-resolution ground- penetrating radar system designed, built, and deployed by SRI under contrast to the US Army Night Vision and Electronic Sensors Directorate at Fort Belvoir. This fully polarimetric, 300 to 3000 MHz stepped frequency radar is configured to act as a forward-looking synthetic aperture system with resolution approaching 5 cm. This test bed is being sued 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. This radar has been used to collect surface and buried antitank mine data at the government test sties at Fort AP Hill, Virginia, and very recently at the Yuma Proving Ground, Arizona. Images formed from these dat sets, using different frequency bands and polarizations, will be presented. Preliminary result, including signal-to-clutter ratios for the miens, clutter statistics, and mine probability density functions will be presented.

This paper proposes the use of a linear prediction technique in the frequency domain for landmine detection. A clutter vector sample is modeled by a linear prediction model, where the current clutter vector sample can be expressed as a linear combination of the past few vector samples plus a random component. The detector first computes the Maximum Likelihood estimate of the prediction coefficients and then generates the prediction error. The detector decides the current sample is from landmine if the prediction error is large. Subband processing is also proposed to further improve the performance of the detector. Detection results are provided on measured data collected from a variety of geographical locations. The data sets contain over 2300 mine encounters.

Previous results with Hidden Markov models showed that they could be used to perform reliable classification between mines and background/clutter under a variety of conditions. Since the, new features have been defined and continuous models have been implemented. In this paper, new results are presented for applying them to calibration lane GPR data obtained during the vehicle mounted mine detection (VMMD) Advanced Technology Demonstrations. Morphological Neural Networks can be trained to perform feature extraction and detection simultaneously. Generalizing these networks to incorporate Choquet Integrals provides the added capability of robustness and improved feature learning. These features can provide complementary information compared to those generate by humans. Result of applying these networks to calibration lane GPR data from the VMMD Advanced Technology Demonstrations are provided. Combinations of the various methodologies with previously developed algorithms are also evaluated.

Many researchers and system developers have proposed exploiting synthetic aperture processing to enhance the spatial resolution of subsurface radar imaging for mine and UXO detection. In this paper, we examine the 3D spatial impulse response associated with alternative geometries for synthetic aperture data collection. Several alternative radar geometries have been chosen for examination corresponding to both forward-looking as well as down- looking configurations.

A novel handheld time-domain array GPR antipersonnel mine detection system prototype has been developed. Using an offset paraboloidal reflector antenna to collimate rays form an ultra-wideband feed, the transmitted microwave impulse is concentrated forward, in front of the antenna structure. The resulting wave is a non-uniform plane wave over the portion of ground be investigated, and is incident at 45 degrees to normal. As such, much of the ground reflect wave is directed further forward, away from the operator, the reflector, and the receiving antennas, thereby reducing clutter. However, the wave transmitted into the ground, which interacts with the target, tends to have significant backscatter returning toward the receiving antennas. These receiving antennas are configured in a 2 by 2 array to provide spatial focusing in both along and cross-track directions. This is accomplished by measuring and comparing the backscattered signal at each receiver in the narrow time window between the times when the ground reflected wave passes the receiver and before this wave re-reflects from the reflector components. 2D FDTD simulation of this parabolic reflector transmitter indicates that it generates a beam with a non-uniform planar wavefront, which scatters form rough ground primarily in the forward direction. The wave transmitted into the ground is also planar, propagating at the angle of refraction, and scattering fairly isotropically from a small penetrable target. This system has been built and tested at LLNL, using a very narrow pulse shape. LLNL's Micro-Impulse Radar (MIR) and custom-built wideband antenna elements operate in the 1.5 to 5 GHz range. One particular advantage of using the MIR module is its low cost: an important feature for mine detectors used in developing countries. Preliminary measured data indicates that the surface clutter is indeed reduced relative to the target signal, and that small non-metallic anti-personnel mines can be reliably detected at burial depths as shallow as 1 inch in both dry.

In previous papers the authors reported work on a 3D frequency domain simulation of bistatic GPR scenarios involving signals from buried mines and clutter due to random permittivity fluctuations in the soil, roughness of the air-ground interface, and distributions of rocks. The analysis is based on Born approximations. The emphasis in these papers was on simulation of a focused array radar, which is a multi static system and hence its simulation requires a large number of runs. Simulation of a multi static GPR system places a high premium on speed, which necessitates some loss of accuracy. Work is currently underway on validation of this code through comparison with experimental results and with result obtained with numerical codes that can achieve great accuracy with very long running times. In the work reported in this paper, results obtained with our code, requiring only minutes of running time, are compared with result of a 3D FDFD code, which requires many hours of CPU time for the same case.

Two common techniques proposed for detection of landmines are ground-penetrating radar (GPR.) and IR imagery. Because of the wide diversity of mines, the clutter which is encountered in minefields, and variation caused by the ground surface, the task of interpreting GPR. signals is daunting. Likewise, variations in thermal properties of soil, solar heating, clutter, and surface irregularities lead to limited performance for IR imaging systems.

The Ultra-Wideband detection of plastic land mines buried in lossy, dielectric soils is simulated design a new semi- analytic mode matching (SAMM) algorithm. Here, we apply SAMM to the 3D canonical problem of finding the nonspecular reflection of an obliquely-incident plane wave ona lossy dielectric half-space containing a small, shallowly-varying convex-shaped mines buried under modestly rough ground. In the SAMM algorithm, the frequency-dependent scattered fields are constructed form moderately rough ground. In the SAMM algorithm, the frequency-dependent scattered fields are constructed form moderately low-order modal superpositions of spherical waves, each satisfying the Helmholtz equation in its respective material. By least squares fitting, mode coefficients are found which optimally match all boundary conditions at designated points along the boundary surfaces. Spherical wave expansions are chosen at multiple coordinate centers so that small numbers of modes are needed to given convergent results.

Ground Penetrating Radar (GPR) has been found to be a promising technology for detecting landmines. Landmines reflect electromagnetic energy in many directions, causing the GPR antenna to pick up these reflections at multiple positions with respect to the landmine. This decreases the resolution of the GPR image. Reflection seismology, which images the earth's subsurface using elastic rather than electromagnetic waves, experiences similar difficulties. Migration is a technique used in reflection seismology to address this problem. This paper present an adaption of the Gazdag phase-shift migration algorithm for use on GPR data in the context of landmine detection. The possible benefits of migrating GPR dat are more precisely locating mines and improved detection of mines with weak signatures. Landmine detection requires a real-time approach whereas reflection seismology performs migration after the data are acquired. The different scenarios of usage have significant implications on the form of candidate migration algorithms. The adapted algorithm was applied to GPR data acquired during test of the Mine Hunter/Killer system at Fort AP Hill. The results presented herein demonstrate the resolution enhancement potential of the proposed algorithm.

Imaging of land mines using signal of a light-weight GPR is considered as an inverse problem for a Helmholtz-like equation. This equation is derived from Maxwell's system. The inverse problem consists in recovery of electrical permittivity and conductivity of a target(s) using multi- frequency measurements of the back-reflected signal. A novel method of solution is proposed. A crucial advantage of this algorithm over many traditional ones is that it avoids entirely the problem of local minima, since it does not use a least squares cost functional.

The Mine Hunter/Killer Advanced Technology Demonstrator (MH/K ATD) is a US Army program that will demonstrate the current capabilities of technologies for a route clearance mission. The first part of the paper will focus on the War fighter requirements and the utility of the MH/K ATD to the GSTAMIDS Block 1 program. Then, the critical parameters of rate of advance, probability of detection, false alarm rate, probability of kill, and time to neutralize will be discussed. The development of advanced multi-sensor data fusion techniques will be covered here. Next, system integration will be discussed, centering on the challenges of vehicle interface, teleoperation, mine overpass capability, accurate target location and tracking, and autonomous control, including an automated neutralizer delivery mechanisms. Issues for transitioning the MH/K technologies and capabilities to the GSTAMIDS Block 1 program will be discussed.

The Close-in Detection (CID) System is the vehicle-mounted multisensor landmine detection system for the Army CECOM Night Vision Electronic Sensors Directorate (NVESD) Mine Hunter/Killer (MH/K) Program. The CID System is being developed by BAE Systems in San Diego, CA. TRW Systems and Information Technology Group in Arlington, VA and a team of specialists for ERIM, E-OIR, SNL, and APL/JHU support NVESD in the development, analysis and testing of the CID and associated signal and data processing. The CID System includes tow down-looking sensor arrays: a ground- penetrating radar (GPR) array, and a set of Electro-Magnetic Induction (EMI) coils for metal detection. These arrays span a 3-meter wide swath in front of a high mobility, multipurpose wheeled vehicle. The system also includes a forward looking IR imaging system mounted on the roof of the vehicle and covering a swath of the road ahead of the vehicle. Signals from each sensor are processed separately to detect and localize objects of interest. Features of candidate objects are integrated in a processor that uses them to discriminates between anti-tank miens and clutter. Mine locations are passed to the neutralization subsystem of MH/K. This paper reviews the design of the sensors and signal processing of the CID system and gives examples and analysis of recent test results at the NVESD mine lanes. The strengths and weaknesses of each sensor are discussed, and the application of multisensor fusion is illustrated.

A statistical performance analysis of the USMC Coastal Battlefield Reconnaissance and Analysis (COBRA) Minefield Detection (MFD) Model has been performed in support of the COBRA ATD Program under execution by the Naval Surface Warfare Center/Dahlgren Division/Coastal Systems Station . This analysis uses the Veridian ERIM International MFD model from the COBRA Sensor Performance Evaluation and Computational Tools for Research Analysis modeling toolbox and a collection of multispectral mine detection algorithm response distributions for mines and minelike clutter objects. These mine detection response distributions were generated form actual COBRA ATD test missions over littoral zone minefields. This analysis serves to validate both the utility and effectiveness of the COBRA MFD Model as a predictive MFD performance too. COBRA ATD minefield detection model algorithm performance results based on a simulate baseline minefield detection scenario are presented, as well as result of a MFD model algorithm parametric sensitivity study.

Assessing the performance of mine-detection systems usually means calculating probability of detection (P_d and a false-alarm rate (FAR). relying on these measures of performance is a consequence of the way in which mine detection systems are tested. Most advanced technology demonstrations of mine detection systems require the participating contractors to provide the testing agency with a set of alarms, or declarations, that correspond to locations on the ground where a mine is suspected to be buried. Superimposing these alarms with the ground truth, or baseline, allows one to compute the P_d and the FAR, but does not give insight into issues such as signal-to-noise ratios or signal-to-clutter ratios. With knowledge of S/N and S/C ratios, expected performance can be compared with demonstrated performance to determine how sensor sensitivity affects overall performance. In addition, S/C ratios provide a means to judge relative performance, but P_d and FAR alone can be ambiguous.

An overview of the UK MOD Applied Research Program for Land Mine Detection. The Defense Evaluation and Research Agency carries out and manages the whole of the UK MOD's Counter Minewarfare Applied Research program both within its own laboratories and in partnership with industrial and academic research organizations. This paper will address two specific counter Minewarfare programs, the Advanced Hand Held Mine Detector which started in April 1995 and the Mine Detection Neutralization and Route Marking System which started in April 1997. Both are multi-sensor systems, incorporating between the metal detection, ground penetrating radar, nuclear quadrupole resonance, ultra-wideband radar, and polarized thermal imaging.

A new multispectral camera response model has been developed in support of the US Marine Corps (USMC) Coastal Battlefield Reconnaissance and Analysis (COBRA) Advanced Technology Demonstration (ATD) Program. This analytical model accurately estimates response form five Xybion intensified IMC 201 multispectral cameras used for COBRA ATD airborne minefield detection. The camera model design is based on a series of camera response curves which were generated through optical laboratory test performed by the Naval Surface Warfare Center, Dahlgren Division, Coastal Systems Station (CSS). Data fitting techniques were applied to these measured response curves to obtain nonlinear expressions which estimates digitized camera output as a function of irradiance, intensifier gain, and exposure. This COBRA Camera Response Model was proven to be very accurate, stable over a wide range of parameters, analytically invertible, and relatively simple. This practical camera model was subsequently incorporated into the COBRA sensor performance evaluation and computational tools for research analysis modeling toolbox in order to enhance COBRA modeling and simulation capabilities. Details of the camera model design and comparisons of modeled response to measured experimental data are presented.

In this paper we suggest the use of the forward-looking geometry with the Radar mounted in the front of a vehicle. In order to assess the feasibility of this measurement geometry, we have performed an extensive series of measurements in an anechoic chamber. Results show that the forward looking geometry is appropriate for the detection of buried landmines.

The Coastal Battlefield Reconnaissance and Analysis)COBRA) system described here was a Marine Corps Advanced Technology Demonstration (ATD) development consisting of an unmanned aerial vehicle (UAV) airborne multispectral video sensor system and ground station which processes the multispectral video data to automatically detect minefields along the flight path. After successful completion of the ATD, the residual COBRA ATD system participated in the Joint Countermine (JCM) Advanced Concept Technology Demonstration (ACTD) Demo I held at Camp Lejeune, North Carolina in conjunction with JTFX97 and Demo II held in Stephenville, Newfoundland in conjunction with MARCOT98. These exercises demonstrated the COBRA ATD system in an operational environment, detecting minefields that included several different mine types in widely varying backgrounds. The COBRA system performed superbly during these demonstrations, detecting mines under water, in the surf zone, on the beach, and inland, and has transitioned to an acquisition program. This paper describes the COBRA operation and performance results for these demonstrations, which represent the first demonstrated capability for remote tactical minefield detection from a UAV. The successful COBRA technologies and techniques demonstrated for tactical UAV minefield detection in the Joint Countermine Advanced Concept Technology Demonstrations have formed the technical foundation for future developments in Marine Corps, Navy, and Army tactical remote airborne mine detection systems.

JMDT is a Navy/Marine Corps 6.2 Exploratory Development program that is closely coordinated with the 6.4 COBRA acquisition program. The objective of the program is to develop innovative science and technology to enhance future mine detection capabilities. The objective of the program is to develop innovative science and technology to enhance future mine detection capabilities. Prior to transition to acquisition, the COBRA ATD was extremely successful in demonstrating a passive airborne multispectral video sensor system operating in the tactical Pioneer unmanned aerial vehicle (UAV), combined with an integrated ground station subsystem to detect and locate minefields from surf zone to inland areas. JMDT is investigating advanced technology solutions for future enhancements in mine field detection capability beyond the current COBRA ATD demonstrated capabilities. JMDT has recently been delivered next- generation, innovative hardware which was specified by the Coastal System Station and developed under contract. This hardware includes an agile-tuning multispectral, polarimetric, digital video camera and advanced multi wavelength laser illumination technologies to extend the same sorts of multispectral detections from a UAV into the night and over shallow water and other difficult littoral regions. One of these illumination devices is an ultra- compact, highly-efficient near-IR laser diode array. The other is a multi-wavelength range-gateable laser. Additionally, in conjunction with this new technology, algorithm enhancements are being developed in JMDT for future naval capabilities which will outperform the already impressive record of automatic detection of minefields demonstrated by the COBAR ATD.

The methods of delay nonlinear system exploration for minelike objects detection have been developed by means of analytical and scheme construction of delay discrete transformation. Conditions of stability for proposed transformation have been defined from eignevalues analysis for corresponding Jacobian-matrix of delay functional operator. The evolution of the object under investigation is considered as dissipativity described by the vector field divergency. The corresponding numerical simulations confirmed the reliability of proposed methods.

The development of locally topological analysis resulting in reduction of required computer resources and experimental data has been proposed. Calculating local nonuniformity for obtained topological sequence provides a high reliability and accuracy of statistical parameters definition for the attractor under investigation and for complex system diagnostics. Segment variance exploration by partition into both fixed length regions and variable left boundary segments with adaptive transformations allows to increase accuracy of obtained results. The generalized model of adaptive transformations formalism has been represented. Numerical experiments with chaotic time series obtained from delay differential equation confirm the validity of developed methods.

Recently, we have successfully applied the Elliptic Systems Method to inverse problems in laser medical imaging applications. As part of applying this method to mine detection, accurate and fast algorithms are required for solving the forward problem to generate data for the inverse problem. Results for the 2D forward problem using GMRES method are compared with 1D transmission line models. Simulation result for miens and clutter are provided for both methods. The comparison with 1D results suggests that GMRES is an effective approach to modeling the forward problem in mine detection. In addition, the contrast between results for mines and clutter provide useful signal features for initial screening between mines and clutter.

A model of slow-flowing-in time gravitational compression of a spheroidal body has been considered. The pressure inside a spheroidal body has been calculated at mechanical unstable equilibrium, and the internal energy connected with it has been shown to be 3 times less than the potential energy of the gravitating spheroidal body. Time equations have been derived for a slow-flowing gravitational compression of spheroidal body in the vicinity of unstable equilibrium state. Also, has been deduced a thermodynamic relation whereby the gravitational thermodynamic potential has been determined. Two thirds of gravitational potential energy have been shown to be consumed in the substance mass transfer inside the spheroidal body due to the gravitational thermodynamic potential. The density of the mass flow has been introduced for the slow-gravitational spheroidal body.

Detection of various subsurface objects, anti-infantry mines and anti-personnel land miens in particular, hampers representation of the boundary line 'air soil'. The image of the in boundary line 'air soil' is the most intensive, since the reflection coefficient from this line is proportional to a gradient of refraction index. If this image to be suppressed the problem of subsurface an object detecting is essentially facilitated.

In this paper we discuss an improved algorithm for sensor- specific data processing and unsupervised data clustering for landmine discrimination. Pre-processor and data- clustering modules forma central part of modular sensor fusion architecture for landmine detection and discrimination. The dynamic unsupervised clustering algorithm is based on Dignet clustering. The self-organizing capability of Dignet is based on the idea of competitive generation and elimination of attraction wells. The center, width and depth characterize each attraction well. The Dignet architecture assumes prior knowledge of the data characteristics in the form of predefine well width. In this paper some modifications to Dignet architecture are presented in order to make Dignet truly self-organizing and data independent clustering algorithm. Information theoretic per-processing is used to capture underlying statistical properties of the sensor data which in turn is used to define important parameter for Dignet clustering such as similarity metrics, initial cluster width etc. The width of the cluster is also adapted online so that a fixed width is not enforced. A suitable procedure for online merge and clean operations is defined to re-organize the cluster development. A concept of dual width is employed to satisfy the competing requirements of compact clusters and high coverage of the data space. The performance of the improved clustering algorithm is compared with base-line Dignet algorithm using simulated data.

An automated, rapid-cycling vapor concentrator and sample introduction device was designed and evaluated. The device consists of an inert deactivated fused silica capillary sampling loop. The temperature of the loop was manipulated through contact with a cold plate or a hot plate, maintained at pre-selected temperatures with a thermoelectric cooler and heating cartridge, respectively. The position of the loop was controlled with a stepper motor under microprocessor control. The low mass of the loop permit its rapid cooling and heating. This permits efficient trapping of adsorptive vapors such as the nitroaromatics from the air stream and also allows rapid and quantitative transfer of the trapped analytes to the detection system. The use of at thermoelectric cooler permits variable trapping temperatures and increased sampling selectivity without the use of cumbersome cryogenic fluids. Chemically inert sampling train surfaces prevent analyte loss due to irreversible adsorption and cross contamination between samples. The device was evaluated for rapid analysis of nitroaromatic and chlorinated aromatic vapors from air stream at trace concentrations with a selective electron capture detection system. Trapping efficiencies of > 95 percent can be readily obtained with the device for nitroaromatics at ppb and sub ppb concentrations.

The goal of the waterjet-based mine location and identification project is to investigate the use of waterjets to locate and differentiate buried objects. When a buried object is struck with a high-pressure waterjet, the impact will cause characteristic vibrations in the object depending on the object's shape and composition. These vibrations will be transferred to the ground and then to the water stream that is hitting the object. Some of these vibrations will also be transferred to the air via the narrow channel the waterjet cuts in the ground.

IR imaging has been used for landmine detection and discrimination by exploiting the variations in temperature profile on the surface, which may be induced by natural phenomenon such as diurnal cycles or using artificial means such as heated waterjets. While the former method has, in general, not been able to reliably detect and discriminate for small antipersonnel mines, the latter suffers from poor response time. Our previous research has shown that, for waterjet induced thermal images, it takes approximately 15 minutes for the profile of the buried object before it is available on the surface. In this paper we explore the possibility of using thermal profile induced by a single heated water jet when viewed directly into the hole created by the waterjet. A heated waterjet, as it penetrates the ground cover, also digs a hole through which the heat radiates out. The spatial and temporal variation of the heat profile in and around the hole has shown to be rich in information about the buried object. Moreover, the response is much faster when compared to the conduction of heat through the soil to the surface. This paper will present the basic phenomenology and characterize such thermal images induced by single heated waterjet. The spatial and temporal variations are used to detect the presence of an object and its material type. Some possibility to measure the depth of the buried object is also explored.

A series of forward-looking ground penetrating radar measurements were taken that bracketed the emplacement of mines at a controlled test site. Fully polarimetric data were collected over a broad rang of frequencies at several times that spanned placement of the mines. These data have been used as input to chagne-detection processing both to examine the overall improvement in signal-to-clutter ratio provided by change detection and to begin assessment of the effects of weather and other relatively long-term variables on its effectiveness. Measurements have been performed initially with the system in a fixed position in order to determine base signal-to-clutter levels given ideal image registration. Under these controlled conditions, change detection is shown to be very effective over the short term in an application such as this. Changes in surface texture are also evident following the emplacement of mines. Morphological closing is shown to be a useful technique for delimiting such regions.

A method for combining the outputs of three different computer aided detection/computer aided classification (CAD/CAC) algorithms is presented and applied to a set of sidescan sonar data taken in the very shallow water environment, where the CAD/CAC algorithms are each tuned to detect mine-like objects. The fusion center receives from each algorithm the planar image coordinates and a confidence factor associated with individual CAD/CAC contacts, and produces fused classification reports of the mine-like objects. Since the three CAD/CAC algorithms use very different approaches, we make the reasonable assumption that valid classifications are nearby each other and false alarms occur randomly in the image. The resultant geometric clustering eliminates most of the false alarms while maintaining a high level of correct classification performance. Our unique fusion algorithm takes a constrained optimization approach, which minimizes the total number of false alarms over the clustering distance and cluster confidence factor thresholds for a given probability of correct classification. Resultant receiver operating characteristics show a significant reduction in the number of false contacts: the false alarm rate from any individual CAD/CAC algorithm is reduced by a factor of four or greater through the optimized data fusion processing.

The explosive charge within a landmine is the source for a mixture of chemical vapors that form a distinctive 'chemical signature' indicative of a landmine. The concentration of these compounds in the air over landmines is extremely low, well below the minimum detection limits of most field- portable chemical sensors. Described in this paper is a man- portable landmine detection system that has for the first time demonstrated the ability to detect landmines by direct sensing of the vapors of signature compounds in the air over landmines. The system utilizes fluorescent polymers developed by collaborators at the MIT. The sensor can detect ultra-trace concentrations of TNT vapor and other nitroaromatic compounds found in many landmine explosives. Thin films of the polymers exhibit intense fluorescence, but when exposed to vapors of nitroaromatic explosives the intensity of the light emitted from the films decreases. A single molecule of TNT binding to a receptor site quenches the fluorescence from many polymer repeat units, increasing the sensitivity by orders of magnitude. A sensor prototype has been develop that response in near real-time to low femtogram quantities of nitroaromatic explosives. The prototype is portable, lightweight, has low power consumption, is simple to operate, and is relatively inexpensive. Simultaneous field testing of the sensor and experienced canine landmine detection teams was recently completed. Although the testing was limited in scope, the performance of the senor met or exceeded that of the canines against buried landmines.

This paper describes the theory and practice of ground penetrating radar (GPR) clutter characterization and removal. Clutter and target parametric and non-parametric modeling methods are described and results of these methods on laboratory data are presented. Data were collected at the Technische Universitaet Ilmenau using a 6 GHZ frequency- stepped GPR. Targets were chosen to include rocks and other non-lethal clutter which normally present false targets to the GPR. Results indicate a quantifiable improvement in target class discrimination using the clutter reduction methods over standard mean background removal methods.

The main disadvantage of applying IR thermal images for detection of buried mines, is the present of various false indications in thermograms together with strong influence of the environmental conditions for final results. A simple use of IRT equipment with better temperature resolution would not help in distinguishing mines, since noise does not come form camera but from the soil surface. The purpose of this paper is to present the phenomenology of the potential soil temperature gradients and distributions on the surface of the soil induced by both natural sources and buried mine. The aim of presented models is to help in recognizing the peculiarities of signal and noises depending on such parameters as: time and space variability of moisture and density of soil, buried mine and soil features and environmental conditions. Numerous examples of simulations and thermographic measurements are presented. Measurements were made for field and laboratory stand-ups, using methodologies typical for 'single-shot' measurements as well as analyses of transient processes based on sequence of thermograms. This paper shows chosen limitations of the thermal methodology efficiency.

The purpose of the Explosives Fate and Transport (EF and T) experiments is to define in detail the accessible trace chemical signature produced by the explosives contained in buried landmines. We intend to determine the partitioning, composition, and quantity of explosive related chemicals which emanate form different kinds of landmines buried in multiple soil types and exposed to various climatic events. We are also developing a computer model that will enable us to predict the composition and quantity of ERC under a much wider range of environmental conditions than we are able to measure experimentally.

The EPFL and the VUB have been investigating the response of metal detectors within the framework of humanitarian demining research activities, in particular frequency domain systems. A simple circuit model has bene looked at first, followed by the analysis of a more completed model. As has also been stressed before, this analysis indicates the possibility of identifying some metallic objects. In addition the phase shift of the received signal turns out to be a continuous, monotonically decreasing function of the object size; this leads to the idea of imposing a 'phase threshold' in order to reduce the amount of detected clutter. This discrimination-based approach is less ambitious than object identification, but is likely to be more robust and to work when looking for metallic objects of a certain size, e.g. non minimum-metal mines or UXO. A first series of measurements was therefore carried out using a commercially available, differential tow frequent metal detector, the Foerster MINEX 2FD. The detector's internal signals have been recorded in a laboratory setup along linear scans varying different object parameters for several representative objects. The collection of data as a function of movement enables the possibility of analyzing the data in the complex plane, and makes it possible to exploit global object properties. Some representative results are presented and the limits of such discrimination/identification approaches briefly outlined.

The result form field-tests using a Stepped-Frequency Ground Penetrating Radar (SF-GPR) and promising antenna and air- ground deembedding methods for a SF-GPR is presented. A monostatic S-band rectangular waveguide antenna was used in the field-tests. The advantages of the SF-GPR, e.g., amplitude and phase information in the SF-GPR signal, is used to deembed the characteristics of the antenna. We propose a new air-to-ground interface deembedding technique based on Principal Component Analysis which enables enhancement of the SF-GPR signal from buried objects, e.g., anti-personal landmines. The methods are successfully evaluate on field-test data obtained from measurements on a large-scale in-door test field.

A standard pulsed induction metal. detector is used to image buried metallic objects by scanning an area of interest. It is shown that, under specific hypotheses, the output image is the result of the convolution of a target function with a kernel depending on the incident magnetic field. Several hypotheses are considered, leading to different kernel shapes and different interpretations of the target function. As the detector imaging function is a low-pass filter, shape's details spread out and the resulting raw image are blurred, Since a high-pass restoration filter must be used to deconvolve the raw images, care must be taken to avoid a strong amplification of noise. The imaging filter is computed using a numerical simulation of the incident magnetic field. Finally, the restoration filter is computed using the Wiener approach. Results are shown for a couple of metallic pieces.

In this paper, the time domain modeling of an indoor impulse UWB GPR systems, built in the scope of the HUDEM project, is presented. For an impulse UWB system, a time-domain modeling is an obvious choice. We explain how the antennas can be characterized by their normalized impulse response. By considering the antenna as a convolution operator, we get a mechanism for modeling the whole radar system as a cascade of linear response, which gives a lot of advantages and possible application. In our research it is used to express the radar range equation in the time-domain, to optimize the antenna configuration and to calculate the point-spread function of the UWB GPR at a given depth. The point-spread function can be used for migration by deconvolving it from the collected data. In this way the migration method takes into account the characteristics of the radar system. Finally, results of this migration method on data obtained by our UWB GPR system are shown.

In this paper, ideas for modeling humanitarian mine detection sensors and their combination within Dempster- Shafer framework are presented. Reasons for choosing this framework are pointed out, taking into account specificity and sensitivity of the problem. This work is done in the scope of the HUDEM project, where three promising and complementary sensors are investigated, so detail analysis is performed in case of fusing the data from them. A way for including in the model influence of various factors on sensor and their result ins discussed as well and will be further analyzed in the future. The application of the approach proposed in this paper is illustrated on the case of sensing metallic objects, but it is possible to modify it for other situations.

In this presentation, we compare the gain in performance offered by combing the result of a ground-penetrating radar, an electromagnetic induction metal detector, and a magnetometer (MAG) against the performance offered by any one of these sensors alone on the problem of buried mine and unexploded ordnance detection. Using the community-wide DARPA background clutter data set, we characterize the single-channel performance of each of these detectors, describing the preprocessing and detection processing used for each. We then combine the sensor results, using a variety of binary decision-level Boolean methods. A performance gain was observed as a two-to-threefold reduction in the false alarm rate, operating at an 80 percent probability of detection, for 'majority voting', which was the best of the combining methods.

This paper explores new imaging approaches for detection of buried plastic mines using data from a ground penetrating radar array. The first approach is based on fine-resolution imaging of the perrnittivity contrast in the region of interest. We develop a model relating the collected data to the contrast generated by buried objects, using ray optics to represent the air/soil interface and Borrt weak scattering approximations. We use this model to develop inversion algorithms for image formation, using alternative regularization approaches to overcome modeling error and ill-posedness due to limited sensing geometries. The second approach is based on tomographic curve evolution techniques, which use object boundaries and a set of contrast coefficients to represent the underlying perrnittivity contrast field. The inversion method seeks to extract less information than a full image, concentrating on accurate estimation of object boundaries. The problem of determining the reduced parameters is formulated as a non-linear estimation problem, which is solved iteratively using level set techniques. The algorithms are tested on data generated by nonlinear finite difference time domain electromagnetic simulations, under conditions involving different ground roughness and object geometries.

Sensor and algorithm fusing is playing an increasing role in many application domains. As detection and recognition problems become more complex and costly, it is apparent that no single source of information can provide the ultimate solution. However, complementary information can be derived from multiple sources. Given a set of outputs form constituent source there are many frameworks within which to combine the pieces into a more definitive answer. A more fundamental question, though, is the following. If we know the general characteristics of a set of sensors, can we predict the value added by fusing their outputs together. Correspondingly, can we specify the needed characteristics of a new sensor/algorithm to add to an existing suite to gain a desired improvement in performance. These questions are difficult and, of course, coupled to the fusion framework. In this paper, we consider these questions in the context of fuzzy set theory, taking a step towards an answer. In particular, we look at a quantitative analysis of sensor system fusion of landmine detection locations. We develop new tools to examine the performance of detection position errors, modeled by vectors of fuzzy sets, in a simulation environment. The approach is shown with general data obtained from an advanced technology demonstration.