This PDF file contains the front matter associated with SPIE Proceedings Volume 7303, including the Title Page, Copyright information, Table of Contents, and the Conference Committee listing.

The ability of animals to detect explosives is well documented. Mammalian systems, insects and even single celled organisms have all been studied and in a few cases employed to detect explosives. This paper will describe the potential ability of ants to detect, disperse and possibly neutralize bulk explosives. In spring 2008 a team of DRDC and Itres scientists conducted experiments on detecting surface-laid and buried landmines, improvised explosive devices (IEDs) and their components. Measurements were made using state-of-the-art short wave and thermal infrared hyperspectral imagers mounted on a personnel lift. During one of the early morning measurement sessions, a wispy, long linear trail was seen to emanate several meters from piles of explosives that were situated on the ground. Upon close visual inspection, it was observed that ants had found the piles of explosives and were carrying it to their ant hill, a distance of almost 20 meters from the piles. Initial analysis of the hyperspectral images clearly revealed the trail to the ant hill of explosives, despite being present in quantities not visible to the unaided eye. This paper details these observations and discusses them in the context of landmine and IED detection and neutralization. Possible reasons for such behaviour are presented. A number of questions regarding the behaviour, many pertinent to the use of ants in a counter-landmine/IED role, are presented and possible methods of answering them are discussed. Anecdotal evidence from deminers of detection and destruction of explosives by ants are presented.

Hexamethylene triperoxide diamine (HMTD) is a well known amine peroxide that starts to decompose at ambient temperature. At 40°C HMTD began to break up into volatile pungent compounds, including trimethyl amine. The production of these compounds could be useful for the vapor detection of HMTD by common techniques such as GC-MS and IMS. GC-MS analysis was performed and several volatile amines could be detected including initial reagents, such as hexamine. IMS produced an alarm indicating the presence of other compounds. Open-Air Chemical Ionization (OACI)-Time of Flight (TOF)-Mass Spectroscopy was the most useful technique for the analysis of HMTD. Even at high temperatures (250°C), it was possible to detect the molecular ion at m/z 209.078. Other fragments observed in the mass spectrum were the loss of formaldehyde at m/z 179.069 and the loss of hydrogen peroxide at m/z 145.060. A mixture of 30 ppb of HMTD and triacetone triperoxide (TATP) was successfully analyzed by OACI-TOF-MS, thus demonstrating its feasibility for trace analysis of organic peroxides and related compounds.

A small-scale field column experiment was set up to assess the impact of a native tropical grass (Fimbrystilis Cymosa) on the transport and distribution of high explosives (TNT and DNT). Explosives powders in a membrane were embedded as a point source below 2 inches from the column surface. Three different surfaces were layered on top of the explosives layer: one column with sand, two columns with Fimbrystilis Cymosa, and one column with a mixture of (sand+clay) soil. Hydraulic differences due to surface vegetation which would affect explosives transport were monitored by measuring the amount of infiltrated rain water. For the biogeochemical parameters, explosives concentrations in the infiltrated water were quantified. At the end of the experiment, each column was sacrificed by multiple layers and distribution of explosives concentrations, soil pH, and soil dehydrogenase concentration was quantified from the layers. Plants were also analyzed for explosives concentrations in their leaves and roots.

The transport of Explosive Related Compounds (ERCs) has been studied as part of a research program aiming to the development of chemical sensors for detecting landmines. TNT and its degradation products typically make up the explosive charge in buried mines. The spatial and temporal distribution of concentrations of ERCs depends primarily on the mobility of the water phase since the main chemicals are transported through the liquid phase of the soil (water). This work presents an analytical approach to the description of the transport process. The model is based on the conservation equations applied to the vadose zone and predicts the concentration profiles of water and ERCs as a function of time. Techniques, such as linearization, variable transformations, and perturbation analysis are used in the development of the model. Results agree with experiments and numerical simulations previously reported.

Underwater video has long been known for its murkiness and low contrast. Modern recording techniques only exacerbate the problem because of the lossy compression methods applied to the video signals. Various image enhancement tools have been implemented in software and hardware, but most of these assume that the user is processing uncorrupted video and that the image has a long dynamic range. Simple preparatory steps can be applied before the enhancement to derive a better image. These will be discussed, with some examples shown processed with a Matlab. GUI. A comparison will be made with a commercial video-enhancement unit.

An improved automatic target recognition processing string has been developed. The overall processing string consists of pre-processing, subimage adaptive clutter filtering, normalization, detection, data regularization, feature extraction, optimal subset feature selection, feature orthogonalization and classification processing blocks. The objects that are classified by the 3 distinct ATR strings are fused using the classification confidence values and their expansions as features, and using "summing" or log-likelihood-ratio-test (LLRT) based fusion rules. The utility of the overall processing strings and their fusion was demonstrated with new high-resolution three-frequency band sonar imagery. The ATR processing strings were individually tuned to the corresponding three-frequency band data, making use of the new processing improvement, data regularization; this improvement entails computing the input data mean, clipping the data to a multiple of its mean and scaling it, prior to feature extraction and resulted in a 3:1 reduction in false alarms. Two significant fusion algorithm improvements were made. First, a nonlinear exponential Box-Cox expansion (consisting of raising data to a to-be-determined power) feature LLRT fusion algorithm was developed. Second, a repeated application of a subset Box-Cox feature selection / feature orthogonalization / LLRT fusion block was utilized. It was shown that cascaded Box-Cox feature LLRT fusion of the ATR processing strings outperforms baseline "summing" and single-stage Box-Cox feature LLRT algorithms, yielding significant improvements over the best single ATR processing string results, and providing the capability to correctly call the majority of targets while maintaining a very low false alarm rate.

The underlying physics of low frequency EMI scattering phenomena in underwater environments from highly conducting and permeable metallic objects is analyzed using an approach that combines the method of auxiliary sources and a surface impedance boundary condition. The combined algorithm solves EMI boundary-value problems by representing the electromagnetic fields in each domain of the structure under investigation by a finite linear combination of analytical solutions of the relevant field equations, corresponding to elementary sources situated a small distance away from the boundaries of each domain. Numerical experiments are conducted for homogeneous and multilayer targets of canonical (spheroidal) shapes subject to frequency- or time-domain illumination, as well as for heterogeneous UXO like targets, to demonstrate: (a) how marine environments change EMI sensor performance and associated processing approaches for detecting highly conducting and permeable metallic objects underwater, and (b) what are the EMI sensors detectability limits. Near and far EMI field and induced eddy-current distributions are presented to help gain insight into underwater EMI scattering phenomena. Particularly, the results illustrate coupling effects between the object and its surrounding conductive medium, especially at high frequencies (early times for time-domain sensors). The results also suggest that this coupling depends on the object's material properties, the conductivity of the medium, and the distance between the sensor and the object's center.

Recently, several sensor technologies, such as magnetometers (total-field and gradiometers) and various types of timedomain and frequency-domain electromagnetic induction (EMI) sensors have been developed and applied successfully to land-based subsurface unexploded ordnance (UXO) detection and mapping. Current researchers of underwater UXO detection commonly apply land-based UXO detection technologies directly to underwater scenarios. Since the electric conductivity of water is much higher than that of soil, an object's EMI response underwater should be different than in a dry environment because inside the conducting water low-frequency electromagnetic signals change both in magnitude and phase, particularly at high frequencies where induction numbers (i.e., wavenumbers) are significantly high. In order to fully explore the capabilities and limitations of land-based EMI sensors for underwater UXO detection and discrimination, in this paper we assess the applicability of current EMI forward models by investigating how the electromagnetic parameters of seawater affect the performance of state-of-the-art EMI sensors. The studies are conducted using the Generalized Standardized Excitation Approach. Objects' locations are inverted for using a reduced version of the HAP technique that combines the magnetic field and its gradient. Particular attention is given to understanding how seawater EM parameters or a multilayer conductive background change objects' EMI responses and affect the UXO discrimination process.

Raytheon has extensively processed high-resolution sidescan sonar images with its CAD/CAC algorithms to provide classification of targets in a variety of shallow underwater environments. The Raytheon CAD/CAC algorithm is based on non-linear image segmentation into highlight, shadow, and background regions, followed by extraction, association, and scoring of features from candidate highlight and shadow regions of interest (ROIs). The targets are classified by thresholding an overall classification score, which is formed by summing the individual feature scores. The algorithm performance is measured in terms of probability of correct classification as a function of false alarm rate, and is determined by both the choice of classification features and the manner in which the classifier rates and combines these features to form its overall score. In general, the algorithm performs very reliably against targets that exhibit "strong" highlight and shadow regions in the sonar image- i.e., both the highlight echo and its associated shadow region from the target are distinct relative to the ambient background. However, many real-world undersea environments can produce sonar images in which a significant percentage of the targets exhibit either "weak" highlight or shadow regions in the sonar image. The challenge of achieving robust performance in these environments has traditionally been addressed by modifying the individual feature scoring algorithms to optimize the separation between the corresponding highlight or shadow feature scores of targets and non-targets. This study examines an alternate approach that employs principles of Fisher fusion to determine a set of optimal weighting coefficients that are applied to the individual feature scores before summing to form the overall classification score. The results demonstrate improved performance of the CAD/CAC algorithm on at-sea data sets.

Detection and identification of vehicles obscured by forest canopy is a particularly challenging military problem. Imaging techniques, e.g. laser radar imaging a target through gaps in foliage, require extensive data, making this approach processing-intensive and time-consuming. A new method for standoff detection of a vehicle obscured under forest canopy by remotely sensing the vibration of foliage with a laser Doppler vibrometer (LDV) has been proposed. The method uses the effect of the vehicle engine creating sound waves, which then travel through the air and then couple into tree leaves, causing them to vibrate. The presence of a vehicle can be determined by the spectrum of the leaves' vibrations. Experimental study has shown that vibration velocity of leaves excited by sound from a vehicle is high enough to be reliably detected with a LDV. The vibrations of leaves excited with simulated vehicle acoustic stimuli and a real vehicle were successfully measured with a LDV in the laboratory and in an outdoor environment. The effect of wind on measurements have been studied and discussed in the current work.

Using Laser Doppler vibrometry (LDV) to find buried land mines has been shown to have a high probability of detection coupled with a low probability of false alarms. Previous work has shown that is it possible to scan a square meter in 20 seconds, but this method requires that discrete areas be scanned. This limits the use of LDVs for land mine detection to a confirmation role. The current work at the University of Mississippi has been to explore ways to increase the speed of scanning to allow the sensor to move down the road at speed. One approach has been to look at the feasibility of using multiple beams to look at the same spot, time division multiplexing, in order to build a time history over small ground segments as each beam passes over the spot. The composite velocity signature built from each beam will provide a long enough time series to obtain the necessary frequency resolution.

A synthetic aperture acoustic approach is used as a standoff method to assess material properties of a typical cinder block, referred to as a concrete masonry unit (CMU), and a variety of CMU surrogates. The objective is to identify anomalies in CMU wall surfaces. The acoustic specular return and phase change across the blocks are the fundamental measurements of interest. The CMU surrogates are created from commercially available closed cell expanding foam. Results from three test articles are presented that show potentially exploitable differences in terms of acoustic magnitude and acoustic phase response between the surrogates and typical CMUs. The test articles are; a typical CMU, a foam block, and a foam block with an embedded steel object. All test articles are similar in size and shape, and both foam blocks are covered in grout so that surface appearance closely matches that of a CMU. The results show that each of the test articles has characteristics that may be used for discrimination and anomaly detection.

The focus of this paper is a review of methods and algorithms for human motion detection in the presence of nonstationary environmental background noise. Human footstep forces on the ground/floor generate periodic broadband seismic and sound signals envelopes with two characteristic times, T₁ (the footstep repetition time, which is equal to the time of the whole body periodic vibrations) and T₂ (the footstep duration time, which is equal to the time interval for a single footstep from "heel strike" to "toe slap and weight transfer"). Human body motions due to walking are periodic movements of a multiple-degrees-of-freedom mechanical system with a specific cadence frequency equal to 1/T₁. For a walking human, the cadence frequencies for the appendages are the same and lie below 3 Hz. Simultaneously collecting footstep seismic, ultrasonic, and Doppler signals of human motion enhance the capability to detect humans in quiet and noisy environments. The common denominator of in the use of these orthogonal sensors (seismic, ultrasonic, Doppler) is a signal-processing algorithm package that allows detection of human-specific time-frequency signatures and discriminates them using a distinct cadence frequency from signals produced by other moving and stationary objects (e.g. vehicular and animal signatures). It has been experimentally shown that human cadence frequencies for seismic, passive ultrasonic, and Doppler motion signatures are equivalent and temporally stable.

This paper looks at depth estimation techniques using electromagnetic induction (EMI) metal detectors. Four algorithms are considered. The first utilizes a vertical gradient sensor configuration. The second is a dual frequency approach. The third makes use of dipole and quadrapole receiver configurations. The fourth looks at coils of different sizes. Each algorithm is described along with its associated sensor. Two figures of merit ultimately define algorithm/sensor performance. The first is the depth of penetration obtainable. (That is, the maximum detection depth obtainable.) This describes the performance of the method to achieve detection of deep targets. The second is the achievable statistical depth resolution. This resolution describes the precision with which depth can be estimated. In this paper depth of penetration and statistical depth resolution are qualitatively determined for each sensor/algorithm. A scientific method is used to make these assessments. A field test was conducted using 2 lanes with emplaced UXO. The first lane contains 155 shells at increasing depths from 0" to 48". The second is more realistic containing objects of varying size. The first lane is used for algorithm training purposes, while the second is used for testing. The metal detectors used in this study are the: Geonics EM61, Geophex GEM5, Minelab STMR II, and the Vallon VMV16.

Recently, new generation, relatively sophisticated, ultra wideband EMI sensors with novel waveforms and multi-axis or vector receivers, have been developed which operate either in the time domain or in the frequency domain. Among these emerging technologies is the Time-domain Electromagnetic Multi-sensor Tower Array Detection System (TEMTADS). The system consists of 25 transmit/receive pairs arranged in a 5 × 5 grid, each with a square 35-cm diameter transmitter coil and a concentric square 25-cm receiver coil. The sensor activates the transmitter loops in sequence, and for each transmitter all receivers receive, measuring the complete transient response over a wide dynamic time range going approximately from 100 μs to 25 ms and distributed in 123 time gates. Thus it provides 625 data points at each location, without the need for a relative positioning system due to its fixed geometry. The combination of spatial diversity in the measurements and well-located sensor positions offers unprecedented data quality for discrimination processing algorithms. To take advantage of the data diversity that this instrument provides, we will use both of the following in an analysis of data acquired with the TEMTADS at Aberdeen Proving Ground (APG) in 2008: (1) advanced, physically complete EMI forward models such as the normalized surface magnetic source (NSMS) model and (2) a data-inversion scheme that uses the newly developed HAP method to estimate the location of a target. Initially the applicability of the NSMS and HAP algorithms to TEMTADS data sets are demonstrated by comparing the modeled data to test-stand and calibration data, and then the APG blind discrimination studies are conducted using as discrimination parameters the total NSMS and principal axes of the induced magnetic polarizability tensor for each target. The classification is done on the extracted feature vector via statistical classification tools.

In this paper a physically complete model called the Normalized Surface Magnetic Source (NSMS) model is applied to data collected using the Berkeley UXO Discriminator time-domain sensor. The sensor has three pairs of rectangular transmitters and eight pairs of receivers that measure gradients of scattered fields. The system is cart-based and produces well-located EMI data sets. In order to take advantage of this high quality data the NSMS technique is utilized for the BUD instrument. The NSMS is a very simple and robust technique for predicting the EMI responses of various objects. The technique is applicable to any combination of magnetic or electromagnetic induction data for any arbitrary homogeneous or heterogeneous 3D object or set of objects. The NSMS approach uses magnetic dipoles, distributed on a fictitious closed surface, as responding sources for predicting an object's EMI response. The amplitudes of the NSMS sources are determined from actual measured data and the resulting total NSMS is used as a discriminant. To demonstrate the applicability of the NSMS technique, we compare actual and predicted data for various UXO. The data were collected at Yuma Proving Ground UXO sites by personnel from the University of California, Berkeley.

Recently the SERDP/ESTCP office under the UXO Discrimination Pilot Study Program acquired high-density data over hundreds of targets using time-domain EM-63 sensor at Camp Sibert. The data were inverted and analyzed by various research groups using a simple dipole model approach and different classification tools. The studies demonstrated high discrimination probability with a low false-alarm rate. However in order to further improve discrimination between UXO and non-UXO items a better understanding is needed of the limits of current and emerging processing approaches. In this paper, the simple dipole model and a physically complete model called the normalized surface magnetic source (NSMS) the Camp Sibert data sets. The simple, infinitesimal dipole representation is by far the most widely employed model for UXO modeling. In this model, one approximates a target's response when excited by a primary (transmitted) field using an induced infinitesimal dipole (in turn described by a single magnetic polarizability matrix). The greatest advantage of the dipole model is that it is simple and imposes low computation costs. However, researchers have recently begun to realize the limitations of the simple dipole model as an inherently coarse description of the EMI behavior of complex, heterogeneous targets like UXO. To address these limitations, here the NSMS is employed as a more powerful forward model for data inversion and object discrimination. This method is extremely fast and equally applicable to the time or frequency domains. The object's location and orientation are estimated by using a standard nonlinear inversion-scattering approach. The discrimination performance between the dipole and NSMS models are conducted by investigating model fidelity and data density issues, positional accuracy and geological noise effects.

During the last decade, the safety regulations of the airports have been set to a new level. As the number of passengers is constantly increasing, yet effective but quick security control at checkpoints sets great requirements to the 21st century security systems. In this paper, we shall introduce a novel metal detector concept that enables not only to detect but also to classify hidden items, though their orientation and accurate location are unknown. Our new prototype walk-through metal detector generates mutually orthogonal homogeneous magnetic fields so that the measured dipole moments allow classification of even the smallest of the items with high degree of accuracy in real-time. Invariant to different rotations of an object, the classification is based on eigenvalues of the polarizability tensor that incorporate information about the item (size, shape, orientation etc.); as a further novelty, we treat the eigenvalues as time series. In our laboratory settings, no assumptions concerning the typical place, where an item is likely situated, are made. In that case, 90 % of the dangerous and harmless items, including knives, guns, gun parts, belts etc. according to a security organisation, are correctly classified. Made misclassifications are explained by too similar electromagnetic properties of the items in question. The theoretical treatment and simulations are verified via empirical tests conducted using a robotic arm and our prototype system. In the future, the state-of-the-art system is likely to speed-up the security controls significantly with improved safety.

Soils that exhibit strong Viscous Remanent Magnetization (VRM) have a major effect on time- and frequency-domain data collected by electromagnetic induction (EMI) sensors. Small scale topography in the form of bumps or troughs will also distort the EM signal due to UXO. If these components of "geologic noise" are not adequately accounted for in the inversion process, then the ability to carry out discrimination will be marginalized. Our long-term goal is to include these effects into the inversion but the chosen methodology depends upon some crucial issues. Foremost, we need to be certain that we can numerically compute the effects of complex magnetic susceptibility and topography that would be encountered in field surveys. Second, we need to investigate whether there is significant electromagnetic interaction between the UXO and its host material or whether the signals are additive. If the total signal can be adequately represented by the superposition of the two individual signals (ie the field of a UXO in free space, and the effect of a conductive host with topography and complex magnetic susceptibility) then there are many avenues by which data can be preprocessed to remove contaminating effects, or by which joint inversion of UXO and host parameters can be carried out. In this paper we concentrate upon the issues of modeling and the possibility of additivity. We first validate our EM numerical modeling code for halfspaces having VRM. We then show that EM interaction between the host and a compact metallic object is minimal for a specific example which is typical of a buried ordnance in a highly magnetic soil such as on Kaho'olawe, Hawaii. We also model soil responses for simple variations of surface roughness including both a single bump and a single trench and compare those results with field data acquired over similar environments.

In UXO contaminated sites, there are often cases in which two or more targets are likely close together and the electromagnetic induction sensors record overlapping signals contributed from each individual target. It is important to develop inversion techniques that have the ability to recover parameters for each object so that effective discrimination can be performed. The multi-object inversion problem is numerically challenging because of the increased number of parameters to be found and because of the additional nonlinearity and non-uniqueness. An inversion algorithm is easily trapped in a local minimum of the objective function that is being minimized. To tackle these problems we exploit the fact that, based on an equivalent magnetic dipole model, the measured electromagnetic induction signals are nonlinear functions of locations and orientations of equivalent dipoles and linear functions of their polarizations. Based on these conditions, we separate model parameters into nonlinear parts (source locations and orientations) and linear parts (source polarizations) and proceed sequentially. We propose a selected multi-start nonlinear procedure to first localize multiple sources and then get the estimated polarization tensor matrix for each item through a subsequent or a nested linear inverse problem. It follows that the orientations of the objects are estimated from the computed tensor matrix. The resultant parameter set is input to a complete nonlinear inversion where all of the dipole parameters are estimated. The overall process can be automated and thus efficiently carried out both in terms of human interaction and numerical computation time. We validate the technique using synthetic and field data.

Clutter is the bane of electromagnetic induction (EMI) surveying for subsurface unexploded ordnance (UXO) under realistic circumstances. Relatively small near-surface metallic items can still produce significant signals simply because they are much closer to the sensor than the larger underlying target of interest. Based on measured, fully multi-static, scalar data at some typical elevation above the ground, one may infer a surface layer of equivalent sources that will produce that data. Without having to locate or characterize the actual targets, one can use these equivalent sources to predict complete vector field data that would be obtained at any elevation equal to or greater than that of the original data. Such computational upward continuation (UC) of signals successfully suppressed clutter in field data. This was even the case when the local clutter signal was significantly stronger than that of the broader underlying UXO response and was embedded directly within it. The success of the approach is directly tied to the fact that it relies on the governing physics.

The detection of unexploded ordnance (UXO) in the presence of a discrete and large clutter is here investigated in the electromagnetic induction regime using a Newton method with no a priori information on the position or the strength of each object. The problem is formulated as a cost-function minimization on the difference in magnetic fields between the measured or synthetic data and their corresponding predictions. Both a bistatic and a monostatic operating modes are considered and applied to various geometrical configurations such as targets in close proximity or on top of each other. Measurement data from the TEMTADS sensor in a two-object configuration are also analyzed. The results illustrate the accuracy of the method in many situations, but also point out at some current limitations for which further improvements are suggested.

A broadband quadrapole electromagnetic induction (EMI) array with one transmitter and three receiver coils is built for detecting buried metallic targets. In this paper, it is shown that the locations of multiple metallic targets including their depth and cross-range position can be estimated accurately with the EMI array using an orthogonal matching pursuit (OMP) approach. Conventional OMP approaches use measurement dictionaries generated for each possible target space point which results in huge dictionaries for the 3D location problem. This paper exploits the inherent shifting properties of the scanning system to reduce the size of the dictionary used in OMP and to lower the computation cost for possibly a real-time EMI location estimation system. The method is tested on both simulated and experimental data collected over metal spheres at different depths and accurate location estimates were obtained. This method allows EMI to be used as a pre-screener and results in valuable location estimates that could be used by a multi-modal GPR or other sensor for enhanced operation.

The Man Portable Vector (MPV) instrument is a time-domain handheld electromagnetic induction (EMI) instrument with five vector receivers and subcentimeter positioning accuracy. For cued interrogations, the MPV is designed to discriminate unexploded ordnance (UXO) from non-UXO using models ranging from the simple dipole model to physically complete models such as the Normalized Surface Magnetic Source (NSMS) method. The MPV acquires both EMI data and position at a 10Hz sampling rate resulting in 150 data points per second at each of a user selectable number time channels (typically 30-90) starting at 100 microseconds. Several factors might limit the usefulness of this data under real world conditions including an excess of usable data, noise in the position data, and insufficient coverage of anomalies. In this paper, we investigate the impact these factors have on the accuracy of discrimination results based on both static and dynamic MPV data. We investigate the effect of using only a subset of the data along with averaging techniques to reduce the amount of MPV data from a single anomaly. In addition, we inject various levels of noise into the position of the MPV in order to gauge the robustness of the discrimination results. Data is also selectively considered based on number of receivers and vector component(s). Results suggest that remarkably few data points are required for accurate discrimination results and that the vector receivers and low hardware noise of the MPV lead to robust results even with sparse data or noisy positional data.

The GEM-3D⁺ sensor developed by Geophex, Ltd. is a new incarnation of their widely known GEM-3. The sensor provides the analyst with all three vector components of the secondary magnetic field over a wide range of frequencies. The GEM-3D⁺ features an innovative "beacon-based" positioning system that provides a full description of its location and orientation at every point without requiring any on-sensor hardware beyond an electronic compass. This enhances the usefulness of the instrument for dynamic surveying, This paper presents some methods and results related to UXO identification using the GEM-3D⁺. Our analyses exploit data provided by the sensor in both grid-based and dynamic measurements to characterize different objects, including metal spheres and actual UXO. For the data analysis we alternate between the dipole model and the more rigorous standardized excitation approach. We review some ill-conditioning issues encountered with the latter model and the different approaches that we use to overcome them. In applications, the availability of horizontal field components in the data allow us to identify UXO vs. non-UXO items while minimizing the nonnegligible effects of ground response.

For development of electromagnetic induction (EMI) sensors for landmine detection, a testing facility has been established for automated measurements of typical targets with both individual sensors and arrays of sensors. A six-degree of freedom positioner has been built with five automated axes (three translational stages and two rotational stages) and one manual axis for target characterizations with no metal within the measurement volume. Translational stages utilize commercially-available linear positioner hardware. Rotational stages have been customized using nonmetallic components to position the targets within the measurement volume. EMI sensors are held fixed in one location while the positioner orients the targets and moves them along a prescribed path through the region surrounding the sensor. The automated movement is computer-controlled and data are acquired continuously. Data are presented from three-dimensional scans of targets at various orientations. Typical targets include shell casings, wire loops, ball bearings, and landmines.

Magnetic properties of soils have adverse effects on metal detectors, particularly hampering operations during clearance of landmines and unexploded ordnance. Although there is well established research in soil magnetism and modeling electromagnetic induction systems these have tended to exist in disparate disciplines. Hence, a workshop was organized to bring together researchers, academics, stakeholders and manufacturers to discuss key priorities for research and technology in a unique multidisciplinary environment. Key knowledge gaps identified include limited information on the spatial heterogeneity of soil magnetic properties in 2D and 3D, whether current models describing soil responses are appropriate for all soils and the need for compensation mechanisms in detectors to be improved. Several priorities were identified that would maximize future developments for multidisciplinary research in soil magnetism and detector technology. These include acquiring well constrained empirical data on soil electromagnetic properties and detector response over the frequency range of detectors; development of predictive models of soil magnetic properties; investigating variability of soil magnetic properties in two and three dimensions across a range of scales. Improved communication between disciplines is key to effective targeting and realization of research priorities. Possible platforms include a multidisciplinary pilot study at an appropriate site and the development of an online repository to assist dissemination of results and information.

Soil moisture conditions influence practically all aspects of Army activities and are increasingly affecting its systems and operations. Regional distributions of high resolution soil moisture data will provide critical information on operational mobility, performance of landmine and UXO sensors, and meteorological conditions at the km scale. The objective of this study is to calibrate RADARSAT-2 surface soil moisture estimates with field measurements in the semi-arid Middle Rio Grande Valley of New Mexico. RADARSAT-2 was launched in December 2007 and is the first SAR sensor to offer an operational quad-polarization mode. This mode allows to generate soil moisture (and cm-scale surface roughness) maps from single data sets. Future combination of such maps into time series will lead to further accuracy enhancement through additional exploitation of soil moisture evolution constraints. We present RADARSAT-2 soil moisture maps, field soil moisture measurements, and soil moisture maps derived from optical imagery. In addition, future work is proposed that may contribute to enhanced algorithms for soil moisture mapping using RADARSAT-2.

Soil moisture conditions influence practically all aspects of Army activities and are increasingly affecting its systems and operations. Regional distributions of high resolution soil moisture data will provide critical information on operational mobility, penetration, and performance of landmine and UXO sensors. The US Army Corps of Engineers (USACE) developed the Gridded Surface/Subsurface Hydrologic Analysis (GSSHA), which is a grid-based two-dimensional hydrologic model that has been effectively applied to predict soil moisture conditions. GSSHA computes evapotranspiration (ET) using the Penman-Monteith equation. However, lack of reliable spatially-distributed meteorological data, particularly in denied areas, makes it difficult to reliably predict regional ET and soil moisture distributions. SEBAL is a remote sensing algorithm that computes spatio-temporal patterns of ET using a surface energy balance approach. SEBAL has been widely accepted and tested throughout the world against lysimeter, eddy-covariance and other field measurements. SEBAL estimated ET has shown good consistency and agreement for irrigated fields, rangelands and arid riparian areas. The main objective of this research is to demonstrate improved GSSHA soil moisture and hydrological predictions using SEBAL estimates of ET. Initial results show that the use of SEBAL ET and soil moisture estimates improves the ability of GSSHA to predict regional soil moisture distributions, and reduces uncertainty in runoff predictions.

The small-scale variability of physical soil properties has a negative influence on ground exploration with physical sensors. This particularly holds true for small target objects like landmines. Studies were carried out to determine magnetic susceptibility, electric conductivity and dielectric permittivity of natural soils. The spatial variability of the field data is quantitatively characterised by means of geostatistical analysis. We present field measurements on different soils types in Germany and on former minefields in Mozambique. The spatial distribution of magnetic susceptibility is governed by the mineral composition of the soil and its stone content. The correlation lengths are in the range of a few meters. In contrast, electric conductivity and permittivity is mainly determined by soil moisture. Due to the small-scale variability of topsoil water content, these two electric properties often feature very small correlation lengths in the range of decimeters. By way of example, the influence of soil variability on landmine detection is illustrated for radar sensors. Geostatistical simulation techniques are used to generate random soil models which are used for realistic finite-differences (FD) calculations of electromagnetic wave propagation. Permittivity variations appear to have a greater influence on radar detector performance than conductivity variations and can mask the signals from the mines.

Soil magnetic properties can seriously impede the performance of metal detectors used in landmine clearance operations. For a proper planning of clearance operations pre-existing information on soil magnetic susceptibility can be helpful. In this study we briefly introduce a classification system to assess soil magnetic susceptibilities from geoscientific maps. The classification system is based on susceptibility measurements conducted on archived lateritic soil samples from 15 tropical countries. The system is applied to a soil map of Angola, resulting in a map that depicts soil magnetic susceptibilities as a worst case scenario. An additional layer depicting the surveyed mine affected communities in Angola is added to the map, which demonstrates that a large number of those are located in areas where soil is expected to impede metal detector performance severely.

A diffractive optics lens based longwave infrared hyperspectral imager has been used to collect laboratory and outdoor field test data. The imager uses a specially designed diffractive optics Ge lens with a 320×256 HgCdTe focal plane array (FPA) cooled with a Sterling-cooler. The imager operates in 8-10.5 μm (long wave IR, LWIR) spectral region and an image cube with 50 to 200 bands can be acquired rapidly. Spectral images at different wavelengths are obtained by moving the lens along its optical axis. An f/2.38 diffractive lens is used with a focal length of 70 mm at 8 μm. The IFOV is 0.57 mrad which corresponds to an FOV of 10.48°. The spectral resolution of the imager is 0.034 μm at 9 μm. The pixel size is 40×40 μm² in the FPA. In post processing of image cube data contributions due to wavelengths other than the focused one are removed and a correction to account for the change in magnification due to the motion of the lens is applied to each spectral image. A brief description of the imager, data collection and analysis to characterize the performance of the imager will be presented in this paper.

A significant amount of background data was collected as part of May 2005 tests at an arid site for airborne minefield detection. An extensive library of the target chips for MSI (four bands) and MWIR sensors for false alarms and mines was created from this data collection, as discussed in another paper in the same proceeding. In this paper we present some results from the analysis of this background data to determine spectral and shape characteristics of different types of false alarms. Particularly, a set of spectral features is identified that can be used for effective false alarm rejection for the benefit of airborne minefield detection programs. A reasonable separation between vegetation and non-vegetation (like rocks) is shown for Normalized Difference Vegetation Index (NDVI) type metrics. Also, a reasonable separation is shown between different types of false alarms at a given time using Color Contrast feature. The spatial distribution of different types of false alarms, as seen in available airborne background data, is also evaluated and discussed. Such spatial analysis is of interest from the perspective of minefield level detection and analysis. The paper is concluded with a discussion on future directions for this effort.

This work is concerned with buried landmines detection by long wave infrared images obtained during the heating or cooling of the soil and a segmentation process of the images. The segmentation process is performed by means of a local fractal dimension analysis (LFD) as a feature descriptor. We use two different LFD estimators, box-counting dimension (BC), and differential box counting dimension (DBC). These features are computed in a per pixel basis, and the set of features is clusterized by means of the K-means method. This segmentation technique produces outstanding results, with low computational cost.

A significant amount of background airborne data was collected as part of May 2005 tests for airborne minefield detection at an arid site. The locations of false alarms which occurred consistently during different runs, were identified and geo-referenced by MultiSensor Science LLC. Ground truth information, which included pictures, type qualifiers and some hyperspectral data for these identified false alarm locations, was surveyed by ERDC-WES. This collection of background data, and subsequent survey of the false alarm locations, is unique in that it is likely the first such airborne data collection with ground truthed and documented false alarm locations. A library of signatures for different sources of these false alarms was extracted in the form of image chips and organized into a self-contained database by Missouri SandT. The library contains target chips from airborne mid wave infrared (MWIR) and multispectral imaging (MSI) sensors, representing data for different days, different times of day and different altitudes. Target chips for different surface mines were also added to the database. This database of the target signatures is expected to facilitate evaluation of spectral and shape characteristics of the false alarms, to achieve better false alarm mitigation and improve mine and minefield detection for airborne applications. The aim of this paper is to review and summarize the data collection procedure used, present the currently available database of target chips and make some recommendations regarding future data collections.

In this study, identification of the different metallic objects with various burial depths was considered. Metal Detector (MD) and Ground Penetrating Radar (GPR) were used to obtain metallic content and dielectric characteristic of the buried objects. Discriminative features were determined and calculated for data set. Six features were selected for metal detector and one for Ground Penetrating Radar. Twenty classification algorithms were examined to obtain the best classification method, for this data set. A Meta learner algorithm completed the classification process with 100% performance.

The successful detection and discrimination of mines is very difficult in areas of high soil mineralization. In these areas, the soil can make a significant contribution to the received signal that causes false detections or masks the true mine response. To address this problem, Minelab has developed a continuous wave (CW) multi-frequency digital detector (MFDD). It transmits four frequencies (between 1 kHz and 45 kHz) and each has a high dynamic range that approaches 120 dB. The mineralized soil with high magnetic susceptibility affects the characteristics of the sensor-head, in particular the inductance of the transmitting and receiving windings. These in turn affect the front-end electronics and measuring circuits and can lead to excessive ground noise that makes detection difficult. Minelab has modeled the effect that the soil has on the sensor-head and developed methods to monitor these effects. By having a well calibrated detector, which is demonstrated by the tight agreement of raw ground signals with theoretical ground models, the tasks of ground balance and discrimination become much more reliable and robust.

Tohoku University, Japan is developing a new hand-held land mine detection dual-sensor (ALIS) which is equipped with a metal detector and a GPR. ALIS is equipped with a sensor tracking system, which can record the GPR and Metal detector signal with its location. The Migration processing drastically increases the quality of the imaging of the buried objects.Evaluation test of ALIS has been conducted several test sites. Tests in real mine fields in Croatia has been conducted between December 2007 and April 2008. Under different soil and environment conditions, ALIS worked well. Then ALIS evaluation test started in Cambodia in February 2009 and we could find discrimination capability of ALIS in test lanes, and we are planning to start evaluation test in real mine fields in Cambodia.

The Nemesis detection system has been developed to provide an efficient and reliable unmanned, multi-sensor, groundbased platform to detect and mark landmines. The detection system consists of two detection sensor arrays: a Ground Penetrating Synthetic Aperture Radar (GPSAR) developed by Planning Systems, Inc. (PSI) and an electromagnetic induction (EMI) sensor array developed by Minelab Electronics, PTY. Limited. Under direction of the Night Vision and Electronic Sensors Directorate (NVESD), overseas testing was performed at Kampong Chhnang Test Center (KCTC), Cambodia, from May 12-30, 2008. Test objectives included: evaluation of detection performance, demonstration of real-time visualization and alarm generation, and evaluation of system operational efficiency. Testing was performed on five sensor test lanes, each consisting of a unique soil mixture and three off-road lanes which include curves, overgrowth, potholes, and non-uniform lane geometry. In this paper, we outline the test objectives, procedures, results, and lessons learned from overseas testing. We also describe the current state of the system, and plans for future enhancements and modifications including clutter rejection and feature-level fusion.

The extensive use of improvised explosive devices (IED) by irregular armed groups in Colombia has posed a threat, not only to the Colombian regular army but to the civilian population. It is expected that in future years, after an eventual cease fire, humanitarian missions for IED clearance will be fundamental to secure safe transit of people and goods, particularly in Colombian rural areas. The clandestine nature of IEDs preparation in Colombia yielded a rather diverse nature of these irregular weapons. Although, some have metal parts such as nails and shrapnel, others are metal-free devices in which detonation is obtained by chemical means. Despite this variability in IEDs design, one thing that is common in IEDs preparation is the use of significant amounts of ANFO (ammonium nitrate (AN), fuel oil (FO)) in their construction. The goal of this work was to identify AN and FO in soils using LIBS. Experiments showed the ability of LIBS to identify the presence of AN based on H_α (656.3 nm) and H_β (486.1 nm) emission lines. It was not possible to identify FO mixed with soils in the spectral windows studied. FO caused a reduction on spark intensity on the samples. This would represent a challenge for identification of chemical compounds in wet soils. Potential interferences with fertilizers are discussed as well.

The ability to interrogate objects buried in soil and ascertain their chemical composition in-situ would be an important capability enhancement for both military and humanitarian demining. Laser Induced Breakdown Spectroscopy (LIBS) is a simple spark spectrochemical technique using a pulsed laser. Recent developments in broadband and man-portable LIBS provide the capability for the real-time detection at very high sensitivity of all elements in any target material because all chemical elements emit in the 200-940 nm spectral region. This technological advance offers a unique potential for the development of a rugged and reliable man-portable or robot-deployable chemical sensor that would be capable of both in-situ point probing and chemical sensing for landmine detection. Broadband LIBS data was acquired under laboratory conditions for more than a dozen different types of anti-personnel and anti-tank landmine casings from four countries plus a set of antitank landmine simulants. Subsequently, a statistical classification technique (partial least squares discriminant analysis, PLS-DA) was used to discriminate landmine casings from the simulants and to assign "unknown" spectra to a mine type based upon a library classification approach. Overall, a correct classification success of 99.0% was achieved, with a misclassification rate of only 1.8%. This performance illustrates the potential that LIBS has to be developed into a field-deployable device that could be utilized as a confirmatory sensor in landmine detection. The operational concept envisioned is a small LIBS system that is either man-portable or robot-deployed in which a micro-laser is contained in the handle of a deminer's probe, with laser light delivered and collected through an optical fiber in the tapered tip of the probe. In such a configuration, chemical analysis could be readily accomplished by LIBS after touching the buried object that one is interested in identifying and using real-time statistical signal processing techniques to accomplish "mine/no-mine" discrimination and even object identification if a material library could be constructed prior to analysis.

We report on a prototype Laser Induced Breakdown Spectroscopy (LIBS) Deminers' Probe used to identify underground objects. We have built a prototype, and are in the process of developing a more advanced LIBS based Deminer' s Probe used to prod objects underground, and then sense them by creating a micro-plasma plume of the surface material and analyzing the spectrum of the emitted light to identify the object. It is expected that the Deminer will be able to eliminate many false positives, which consume most of the Deminers' time. SARA Fiber-Optics coupled LIBS system consists in a probe that can be inserted into the ground to provide a path for both the laser beam to the target, and for the micro-plasma plume fluorescence from the target to a spectrometer or spectrometers for analysis. The probe is closely modeled after the conventional Deminers' probe, resembling a saber. We have demonstrated that this simple system is capable of producing remarkably different spectra from different materials. Our next steps are to add a number of features to the Deminers' Probe. These include: a new optical configuration to increase the irradiance and fluence created by the pulsed laser at the target, a multiple channel fluorescence reception system that can increase the amount of light delivered to the spectrometers, a fluidic system to clear the detritus away from the probe tip, and a complete operational/control and readout system for the Deminer to use. Mine-lane tests are planned to be performed in the later part of 2009, or shortly thereafter.

Recent advances in Laser-Induced breakdown spectroscopy (LIBS), Raman spectroscopy, and other spectroscopic approaches have increased interest in the application of spectroscopy to detection of explosives along with other chemical-signature identification tasks. However most existing spectroscopic data collection techniques require manual interaction with data files including data manipulation using multiple pieces of software and different file formats, time-consuming feature-selection, and model re-generation. Not only do these steps reduce analytic efficiency and slow the progress of research in spectroscopy, but they also inhibit real-time use of the systems by end-users. In this work we present a graphical user interface designed to increase efficiency for spectroscopic data collection, feature selection, classifier development, and testing. We present a software architecture that provides enough flexibility to handle data from many different spectroscopic sensors. We also discuss feature-level and model-level software components that allow for the features and classification approaches to be manipulated interactively, and we present a simple and intuitive testing screen suitable for an end user to make decisions in the field with out requiring a "human in the loop" for processing.

Objects buried in unimproved surfaces can be inferred from the disturbance of the soil above them. We have found for mines emplaced according to U.S. military doctrine in clay-rich soils, that imaging at visible, shortwave infrared, and thermal infrared are effective at different times under various illumination conditions, and that these techniques can be synergistic. Complementary visible - thermal infrared laboratory spectral measurements show that grain size differences associated with disturbed soils can make them more reflective or emissive than undisturbed soils. However, the field measurements demonstrate that grain size effects are not significant under passive visible and shortwave infrared illumination. Instead, shortwave infrared (1.55 - 1.7 μm) imaging, in particular, is effective because the roughened disturbed soil casts a pattern of shadows under a wide range of illumination conditions that are also emphasized by a background of undisturbed soil possessing few contrast variations.

Molecular recognition is an important aspect of any biosensor system. Due to increased stability in a variety of environmental conditions, molecular imprinted polymer (MIP) technology is an attractive alternative to biological-based recognition. This is particularly true in the case of improvised explosive device detection, in which the sensor must be capable of detecting trace amounts of airborne nitroaromatic compounds. In an effort to create a sensor for detection of explosive devices via nitroaromatic vapor, MIPs have been deployed as a molecular recognition tool in a fluorescence-based optical biosensor. These devices are easily scalable to a very small size, and are also robust and durable. To achieve such a sensor scheme, polymer microparticles synthesized using methacrylic acid monomer and imprinted with a 2,4-dinitrotoluene (DNT) template were fabricated. These microparticles were then conjugated with green CdSe/ZnS quantum dots, creating fluorescent MIP microparticles. When exposed to the DNT template, rebinding occurred between the DNT and the imprinted sites of the MIP microparticles. This brings the nitroaromatic DNT into close proximity to the quantum dots, allowing the DNT to accept electrons from the fluorescent species, thereby quenching the fluorescence of the quantum dot. Results indicate that this novel method for synthesizing fluorescent MIPs and their integration into an optical biosensor produced observable fluorescence quenching upon exposure to DNT.

The SALSA camera from Bossa Nova Technologies uses an electronically rotated polarization filter to measure four states of polarization nearly simultaneously. Initial imagery results are presented, with an investigation of polarization invariants, as affected by illumination, sensing geometry, and atmospheric effects. Applications for the system as it is being developed are in change detection and surface characterization. Preliminary results indicate an ability to distinguish new from old asphalt and disturbed earth from undisturbed earth with some image processing.

In this paper a measuring complex of C-, K_u and K_a-band, multi-polarization, combined scatterometer-radiometer systems, their structures and operational features, measuring platforms and calibration facilities are presented. As well as the results of preliminary, spatio-temporally collocated, multi-frequency and multi-polarization measurements of bare soil microwave reflective (radar backscattering coefficient) and emissive (brightness temperature) characteristics angular dependences at 5,6GHz and 15GHz are presented, under various soil moisture and air temperature conditions.

A man-portable Magnetic Scalar Triangulation and Ranging ("MagSTAR") technology for Detection, Localization and Classification (DLC) of unexploded ordnance (UXO) has been developed by Naval Surface Warfare Center Panama City Division (NSWC PCD) with support from the Strategic Environmental Research and Development Program (SERDP). Proof of principle of the MagSTAR concept and its unique advantages for real-time, high-mobility magnetic sensing applications have been demonstrated by field tests of a prototype man-portable MagSTAR sensor. The prototype comprises: a) An array of fluxgate magnetometers configured as a multi-tensor gradiometer, b) A GPS-synchronized signal processing system. c) Unique STAR algorithms for point-by-point, standoff DLC of magnetic targets. This paper outlines details of: i) MagSTAR theory, ii) Design and construction of the prototype sensor, iii) Signal processing algorithms recently developed to improve the technology's target-discrimination accuracy, iv) Results of field tests of the portable gradiometer system against magnetic dipole targets. The results demonstrate that the MagSTAR technology is capable of very accurate, high-speed localization of magnetic targets at standoff distances of several meters. These advantages could readily be transitioned to a wide range of defense, security and sensing applications to provide faster and more effective DLC of UXO and buried mines.

High definition impedance imaging (HDII) is applicable, from d.c. upward for electrical, sonic and elasticity signal excitations. At low frequencies, great depth is achievable in contrast to that provided by radar without HDII. The HDII solution process results in a very large and sparse matrix system and associated algorithms provide convergence with few iterations and high image definition. The methodology solves the three-dimensional image solution rather that by solving in slices. HDII image quality results from the number of linearly independent equations resulting from the number of electrodes and linearly independent measurements that are obtained. To construct a standoff (i.e. contactless) system, the three-dimensional vector Helmholtz equation, i.e. the formulation used in antennal analysis, may be employed. To do this, the same basic HDII imaging algorithm, as used for the contact case, is employed for standoff imaging. Over determination can permit significantly refined image quality.

In this paper the structure and operational features of C-band, multi-polarization, combined scatterometer-radiometer system and the results of preliminary, spatio-temporally collocated measurements of bare soil and land snow cover microwave reflective (radar backscattering coefficient) and emissive (brightness temperature) characteristics angular dependences at ~5.6GHz are presented.

We present a magnetic detection system based on superconducting gradiometric sensors (SQUID gradiometers). The system provides a unique fast mapping of large areas with a high resolution of the magnetic field gradient as well as the local position. A main part of this work is the localization and classification of magnetic objects in the ground by automatic interpretation of geomagnetic field gradients, measured by the SQUID system. In accordance with specific features the field is decomposed into segments, which allow inferences to possible objects in the ground. The global consideration of object describing properties and their optimization using error minimization methods allows the reconstruction of superimposed features and detection of buried objects. The analysis system of measured geomagnetic fields works fully automatically. By a given surface of area-measured gradients the algorithm determines within numerical limits the absolute position of objects including depth with sub-pixel accuracy and allows an arbitrary position and attitude of sources. Several SQUID gradiometer data sets were used to show the applicability of the analysis algorithm.

The use of broadband techniques in the Terahertz (THz) region, has allowed us to probe the effectiveness of far-infrared waves to assess the penetration depths required for anti-personnel mine detection. Using THz-Time Domain Spectroscopy (THz-TDS) based methods we have probed the effective penetration depth of broadband THz pulses through soil samples collected locally in Ankara, Turkey. It was found that certain frequencies within the spectral range from 0.01 to 1THz were severely attenuated by both absorption and scattering effects dependent on the geometry and size of the particles inside the soil samples. For the most extreme case, the rocky soil sample exhibited severe attenuation of the frequencies above 0.1 THz. The measurements were repeated for wetted soil samples as well. From these results we have gained insight into the boundaries imposed by soil conditions on the detection of mines buried near and far beneath the surface. We found that the effective penetration depth is strongly affected by scattering effects and the dominant scattering mechanism is explained by Mie scattering for the soil samples in consideration rather than Rayleigh scattering.

Current ground penetrating radar algorithms for landmine detection require accurate estimates of the location of the air/ground interface to maintain high levels of performance. However, the presence of surface clutter, natural soil roughness, and antenna motion lead to uncertainty in these estimates. Previous work on improving estimates of the location of the air/ground interface have focused on one-dimensional filtering techniques to localize the air/ground interface. In this work, we propose an algorithm for interface localization using a 2- D Gaussian Markov random field (GMRF). The GMRF provides a statistical model of the surface structure, which enables the application of statistical optimization techniques. In this work, the ground location is inferred using iterated conditional modes (ICM) optimization which maximizes the conditional pseudo-likelihood of the GMRF at a point, conditioned on its neighbors. To illustrate the efficacy of the proposed interface localization approach, pre-screener performance with and without the proposed ground localization algorithm is compared. We show that accurate localization of the air/ground interface provides the potential for future performance improvements.

The Edge Histogram Detector (EHD) is a landmine detection algorithm for sensor data generated by ground penetrating radar (GPR). It uses edge histograms for feature extraction and a possibilistic K-Nearest Neighbors (K-NN) rule for confidence assignment. To reduce the computational complexity of the EHD and improve its generalization, the K-NN classifier uses few prototypes that can capture the variations of the signatures within each class. Each of these prototypes is assigned a label in the class of mines and a label in the class of clutter to capture its degree of sharing among these classes. The EHD has been tested extensively. It has demonstrated excellent performance on large real world data sets, and has been implemented in real time versions in hand-held and vehicle mounted GPR. In this paper, we propose two modifications to the EHD to improve its performance and adaptability. First, instead of using a fixed threshold to decide if the edge at a certain location is strong enough, we use an adaptive threshold that is learned from the background surrounding the target. This modification makes the EHD more adaptive to different terrains and to mines buried at different depths. Second, we introduce an additional training component that tunes the prototype features and labels to different environments. Results on large and diverse GPR data collections show that the proposed adaptive EHD outperforms the baseline EHD. We also show that the edge threshold can vary significantly according to the edge type, alarm depth, and soil conditions.

A high range resolution ground penetrating radar signal is processed to convert the A-scan data into a binary valued string in which a one represents the location of an impedance change and a zero otherwise. For non-metallic landmines it has been shown that this pattern is unique and can be used to discriminate among landmines and clutter. The discrimination method is based on regular languages which consist of the binarized sequences produced by various landmines. Methods have been developed to automatically create language recognizers which not only recognize a landmine's characteristic string, but also variations of those strings.

We propose a landmine detection algorithm that uses a mixture of discrete hidden Markov models. We hypothesize that the data are generated by K models. These different models reflect the fact that mines and clutter objects have different characteristics depending on the mine type, soil and weather conditions, and burial depth. Model identification could be achieved through clustering in the parameters space or in the feature space. However, this approach is inappropriate as it is not trivial to define a meaningful distance metric for model parameters or sequence comparison. Our proposed approach is based on clustering in the log-likelihood space, and has two main steps. First, one HMM is fit to each of the R individual sequence. For each fitted model, we evaluate the log-likelihood of each sequence. This will result in an R×R log-likelihood distance matrix that will be partitioned into K groups using a hierarchical clustering algorithm. In the second step, we pool the sequences, according to which cluster they belong, into K groups, and we fit one HMM to each group. The mixture of these K HMMs would be used to build a descriptive model of the data. An artificial neural networks is then used to fuse the output of the K models. Results on large and diverse Ground Penetrating Radar data collections show that the proposed method can identify meaningful and coherent HMM models that describe different properties of the data. Each HMM models a group of alarm signatures that share common attributes such as clutter, mine type, and burial depth. Our initial experiments have also indicated that the proposed mixture model outperform the baseline HMM that uses one model for the mine and one model for the background.

This paper considers the use of data from a wideband electromagnetic induction (EMI) sensor in a prescreener for a landmine detection system employing both ground-penetrating radar (GPR) and EMI sensors. The paper looks at a unique EMI prescreening strategy based on the use of prototypes derived from a training set of landmines. We show that this prescreener is robust to a wide range of induced energy levels in sensed objects. We also compare properties of the receiver operating characteristics (ROC) curve of this prescreener on a varied collection of targets to the properties of a GPR prescreener, identifying performance difference with respect to target object classes.

In previous work, a sensor management framework has been developed that manages a suite of sensors in a search for static targets within a grid of cells. This framework has been studied for binary, non-binary, and correlated sensor observations, and the sensor manager was found to outperform a direct search technique with each of these different types of observations. Uncertainty modeling for both binary and non-binary observations has also been studied. In this paper, a new observation model is introduced that is motivated by the physics of static target detection problems such as landmine detection and unexploded ordnance (UXO) discrimination. The new observation model naturally accommodates correlated sensor observations and models both the correlation that occurs between observations made by different sensors and the correlation that occurs between observations made by the same sensor. Uncertainty modeling is also implicitly incorporated into the observation model because the underlying parameters of the target and clutter cells are allowed to vary and are not assumed to be constant across target cells and across clutter cells. Sensor management is then performed by maximizing the expected information gain that is made with each new sensor observation. The performance of the sensor manager is examined through performance evaluation with real data from the UXO discrimination application. It is demonstrated that the sensor manager is able to provide comparable detection performance to a direct search strategy using fewer sensor observations than direct search. It is also demonstrated that the sensor manager is able to ignore features that are uninformative to the discrimination problem.

We present a novel method for improving landmine detection with ground-penetrating radar (GPR) by utilizing a priori knowledge of environmental conditions to facilitate algorithm training. The goal of Context-Dependent Feature Selection (CDFS) is to mitigate performance degradation caused by environmental factors. CDFS operates on GPR data by first identifying its environmental context, and then fuses the decisions of several classifiers trained on context-dependent subsets of features. CDFS was evaluated on GPR data collected at several distinct sites under a variety of weather conditions. Results show that using prior environmental knowledge in this fashion has the potential to improve landmine detection.

We present a local method for fusing the results of several landmine detectors using Ground Penetrating Radar (GPR) and Wideband Electro-Magnetic Induction (WEMI) sensors. The detectors considered include Edge Histogram Descriptor (EHD), Hidden Markov Models (HMM), and Spectral Correlation Feature (SCF) for the GPR sensor, and a feature-based classifier for the metal detector. The above detectors use different types of features and different classification methods. Our approach, called Context Extraction for Local Fusion with Feature Discrimination(CELF-FD), is a local approach that adapts the fusion method to different regions of the feature space. It is based on a novel objective function that combines context identification and multi-algorithm fusion criteria into a joint objective function. The context identification component thrives to partition the input feature space into clusters and identify the relevant features within each cluster. The fusion component thrives to learns the optimal fusion parameters within each cluster. Results on large and diverse GPR and WEMI data collections show that the proposed method can identify meaningful and coherent clusters and that these clusters require different fusion parameters. Our initial experiments have also indicated that CELF-FD outperforms the original CELF algorithm and all individual detectors.

In-situ trace detection of explosive compounds such as RDX, TNT, and ammonium nitrate, is an important problem for the detection of IEDs and IED precursors. Spectroscopic techniques such as LIBS and Raman have shown promise for the detection of residues of explosive compounds on surfaces from standoff distances. Individually, both LIBS and Raman techniques suffer from various limitations, e.g., their robustness and reliability suffers due to variations in peak strengths and locations. However, the orthogonal nature of the spectral and compositional information provided by these techniques makes them suitable candidates for the use of sensor fusion to improve the overall detection performance. In this paper, we utilize peak energies in a region by fitting Lorentzian or Gaussian peaks around the location of interest. The ratios of peak energies are used for discrimination, in order to normalize the effect of changes in overall signal strength. Two data fusion techniques are discussed in this paper. Multi-spot fusion is performed on a set of independent samples from the same region based on the maximum likelihood formulation. Furthermore, the results from LIBS and Raman sensors are fused using linear discriminators. Improved detection performance with significantly reduced false alarm rates is reported using fusion techniques on data collected for sponsor demonstration at Fort Leonard Wood.

Forward-looking ground-penetrating radar (FLGPR) has received a significant amount of attention for use in explosive hazards detection. A drawback to FLGPR is that it is sensitive to not only explosive hazards but also to benign objects, which results in an excessive number of false detections. This paper presents our analysis of the explosive hazards detection system developed by Planning Systems Inc (PSI). The PSI system combines FLGPR with an infrared (IR) camera. We present an FLGPR target detection algorithm that leverages the multiple observations aspect of FLGPR. The FLGPR target detections are then projected into the IR imagery. A Mahalanobis-metric classifier is then used to reduce the number of false detections. We show that our proposed FLGPR target detection algorithm, coupled with our IR-based false alarm reduction method, is effective at detecting explosive hazards while reducing the number of false alarms.

In this paper we investigate how shape/spectral similarity of the mine signature and the minefield like spatial distribution can be exploited simultaneously to improve the performance for patterned minefield detection. The minefield decision is based on the detected targets obtained by an anomaly detector, such as the RX algorithm in the image of a given field segment. Spectral, shape or texture features at the target locations are used to model the likelihood of the targets to be potential mines. The spatial characteristic of the patterned minefield is captured by the expected distribution of nearest neighbor distances of the detected mine locations. The false alarms in the minefield are assumed to constitute a Poisson point process. The overall minefield detection problem for a given segment is formulated as a Markov marked point process (MMPP). Minefield decision is formulated under binary hypothesis testing using maximum log-likelihood ratio. A quadratic complexity algorithm is developed and used to maximize the log-likelihood ratio. A procedure based on expectation maximization is evaluated for estimating unknown parameters like mine-level probability of detection and mine-to-mine separation. The patterned minefield detection performance under this MMPP formulation is compared to baseline algorithms using simulated data.

Rapid detection of landmines and explosive hazards is a critical issue for modern military operations. Due to the varied nature of the objects of interest and the complexity of the surrounding, one approach is to utilize the superior recognition capabilities of the human brain in the detection process. We are developing frameworks and algorithms to fuse image data from multiple sources and to provide cuing capability for a human-in-the-loop detection system.

The detection of weak scattering plastic landmines using ground penetrating radar (GPR) is a challenging task. This paper presents a few enhancements to the previously proposed subspace spectral correlation feature (SCF) to improve the detection of weak scattering plastic landmines. Preliminary results indicate that the improved subspace SCF technique is able to reduce the false alarm rate by 28% compared to the original subspace SCF at 90% probability of detection, in a data collection obtained from three different test sites that contains about 460 weak-scattering plastic anti-tank landmines. Fusion results with edge histogram descriptor (EHD) confirm the performance improvement resulted from enhanced subspace SCF.

In this study discrimination of two different metallic object classes were studied, utilizing Ground Penetrating Radar (GPR). Feature sets of both classes have almost the same information for both Metal Detector (MD) and GPR data. There were no evident features those are easily discriminate classes. Background removal has been applied to original B-Scan data and then a normalization process was performed. Image thresholding was applied to segment B-Scan GPR images. So, main hyperbolic shape of buried object reflection was extracted and then a morphological process was performed optionally. Templates of each class representatives have been obtained and they were searched whether they match with true class or not. Two data sets were examined experimentally. Actually they were obtained in different time and burial for the same objects. Considerably high discrimination performance was obtained which was not possible by using individual Metal Detector data.