This PDF file contains the front matter associated with SPIE Proceedings Volume 6953, including the Title Page, Copyright information, Table of Contents, and the Conference Committee listing.

Most studies of the electromagnetic induction (EMI) response of a low-metal landmine buried in soil ignore any influence that the plastic casing may have on such response. In most cases such treatment is adequate since only the metal components of a landmine are expected to contribute to such a response. However, when the landmine is buried in a soil that has significant conductivity and/or magnetic susceptibility, the electromagnetic void created by the casing may have an influence on the EMI response of the landmine. That possibility is investigated using a simple analytical model and an experiment. A sphere is chosen as a simple prototype for the small metal parts in low-metal landmines, and a concentric spherical shell, made of foamed polystyrene, encasing the sphere is used to represent the plastic landmine body. The time-domain EMI response is measured using a purpose-designed system based on a modified Schiebel AN19/2 metal detector. Responses of the metallic sphere, the polystyrene shell and the metal-polystyrene composite target are measured with the targets buried in magnetic soil half-spaces. The particular soil type for which data are presented in this paper is Cambodian "laterite" with dispersive magnetic susceptibility, which serves as a good model for soils that are known to affect the performance of metal detectors. The metal sphere used has a diameter of 0.0254 m and is made of 6061-T6 aluminum, and the polystyrene shell has an outer diameter of 0.15 m. For the specific soil and targets used, theoretical results show that a small effect on the time-domain response is expected from the presence of the polystyrene casing. Experimental results confirm this for the case of the buried polystyrene shell. However the small difference in the example of the composite target is masked by experimental errors.

Recently different time domain (TD) electromagnetic induction sensors have been developed and tested for UXO detection and discrimination. These sensors produce well-located, multi-axis, high-density data. One of such sensors is the Man Portable Vector (MPV) TD sensor build by G&G., Inc., which has two 75-cm diameter transmitter loops and five tri-axial cubic receivers located around the transmitter coils. This sensor produces unprecedented high-fidelity complete vector data sets. To take advantage of these high-quality data, in this paper we adapt the normalized surface magnetic source (NSMS) model to the MPV. 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 3-D object or set of objects. The NSMS approach uses magnetic dipoles distributed on a fictitious closed surface as responding sources for predicting objects' EMI responses. The amplitudes of the NSMS sources are determined from actual measured data, and at the end the total NSMS is used as a discriminator. Usually, discrimination between UXO and non-UXO items is processed by first recovering the buried object's location and orientation using standard non-linear minimization techniques; this is the most time consuming part of the UXO classification process. In order to avoid solving a traditional ill-posed inverse scattering problem, here we adapt to TD-MPV data a recently developed physics-based approach, called (HAP), to estimate a buried object's location and orientation. The approach assumes the target exhibits a dipolar response and uses only three global values: (1) the magnetic field vector H, (2) the vector potential A, and (3) the scalar magnetic potential at a point in space. Of these three global values only the flux of the H field is measurable by the MPV sensor. However, the vector and scalar magnetic potentials can be recovered from measured magnetic field data using a 2D NSMS approach. To demonstrate the applicability of the NSMS and HAP techniques we report the results of a blind-test analysis using multi-axis TD MPV data collected at the U.S. Army's ERDC UXO test site.

This work provides a performance comparison between two frequency-domain electromagnetic induction (EMI) sensors - one quadrupole and one dipole sensor for the detection of subsurface anti-personnel and anti-tank landmines. A summary of the physical differences between the two sensors and from those of other EMI sensors will be discussed. Previously we presented a performance analysis of the dipole sensor for a variety of detection algorithms over data collected at a government test facility indicating robust performance using the dipole sensor. The algorithms considered previously included an energy detector, matched subspace detector and a kNN probability density estimation approach over the features of a four parameter phenomenological model. The current sensor comparison will include, in addition to the previous detection methods, a Random Forests classification algorithm and utilize a larger training data set.

A multi dipole (MD) model is combined with a statistical algorithm called the mixed model to discriminate between objects of interest, such as unexploded ordnance (UXO), and innocuous items. In the multi dipole model (an extended version of the single dipole model), electromagnetic induction (EMI) responses for bodies of revolution (BOR) are approximated with a set of dipoles placed along the axis of symmetry of the objects. The model accurately takes into account the scatterer's heterogeneity along its axis of symmetry and is fast enough to invert digital geophysical data for discrimination purposes in real/near real time. Determining the amplitudes of the multi dipoles is an ill-posed problem that requires regularization. Obtaining the regularization parameters is not straightforward and in many cases is done via impractical supervised approaches. To overcome this problem, in this paper we combine a new statistical approach called the mixed model with the multi dipole model. Mixed modeling (MM) can be viewed as a generalization of the empirical Bayesian approach. It assumes that the forward model is not perfect: i.e., the model parameters (the amplitudes of the responding multi magnetic dipoles) contain random noise with zero mean and constant variance. Based on these assumptions, the method derives the regularization parameter from the variance of the least square error between the model and actual data using standard linear regression. Numerical results are presented to illustrate the theoretical basis and practical realization of the combined MD-mixed model (MD-MM) algorithm for UXO discrimination under real field conditions. In addition, a new condensed algorithm for determining the location and orientation of buried objects is introduced and tested against the ESTCP pilot discrimination study dynamic data set.

The high definition impedance imaging (HDII) Electroscan algorithm casts the error norm problem into the interior of the region and iteratively minimizes the difference norm calculated between solutions achieved for applied currents (i.e. the Neumann problem) and the solution achieved for the measured voltages (i.e. the Dirichlet problem) - in the electrical-excitation case. This results in very sparse matrices instead of densely-packed Jacobian matrices. Minimization of the error yields a three-dimensional image of the conductivity distribution. The paper presents a rapid, sparse-matrix methodology for high definition admittivity imaging involving a very large number of voxels. It is a least-square algorithm, simultaneously involving all excitations, and it is error resilient and well-conditioned. The solution iterative procedure is accelerated by a variety of means such as: solution of mutually-constrained, three-dimensional field equations; successive point-iterative overrelaxation; multi-acceleration factors; measurements at a multiplicity of electrodes; and excitation modification for image enhancement. Laboratory, field, and simulation case studies are presented. Spatially restricted-region and open-region solutions are compared. Signal-source modeling is not required. Conductivity and, generally, admittivity values are able to be determined. And so, the imaging process has diagnostic capability. It is applicable to non-contact standoff excitations, e.g. magnetic fields, microwave/radar, sonic and elasticity wave excitations.

Korea is one of the heavily mined countries in the world. The demand for mine detection and clearance techniques has always been high in South Korea. In support of this, a new project on ground penetrating radar (GPR) for landmine detection has been launched in South Korea. The GPR under development is an ultra wideband sensor system that requires high-resolution imaging of buried targets and database construction based on target signals in various ground conditions. For initial experiments, a simple GPR has been built using a resistive vee dipole antenna and a vector network analyzer. The GPR is scanned over a sand tank with an area of 2.5m × 2.5m and a depth of 1.5m, which is used for target burial. During the first stage of the project, the data obtained by scanning the GPR antenna over a target are processed to evaluate various radar signal waveforms, performance of various antennas, and other system configurations. Based on the evaluation, an advanced GPR system will be built and used to construct the database during the second stage of the project. A description for motivation for the GPR project, overview of the GPR project, experiment setup, and initial experiment results are presented in this paper.

Femtosecond laser induced breakdown spectroscopy (LIBS) has been shown to be sensitive to a variety of ERC's (explosive-related compounds) deposited on substrates. In LIBS, surface material is excited by a high-powered laser pulse forming a plasma. The optical emission from this plasma is collected and spectrally analyzed to determine the surface species entrained in the excitation event. The detection of explosive related compounds in the field presents many challenges, one of these being the wide variety of materials surfaces that might be covered with ERC's. Results from femtosecond and nanosecond LIBS of ERC's of metal, glass, and polymer substrates show that the optical properties of the substrate play a large role in the observed emission. Results indicate that nanosecond LIBS of ERC's on metal surfaces yield strong atomic emission while nanosecond LIBS of ERC's on glass results in some molecular emission. Molecular emission is also present in femtosecond LIBS spectra of ERC's on all surfaces but is particularly strong for metal substrates. In particular emission from the CN molecular fragment could provide a means to understand the effect of the substrate on the excitation event in nitroaromatic compounds since it is present in both nanosecond LIBS spectra of the TNT/glass system and femtosecond LIBS spectra of the TNT/Al system. The origins of this CN molecular fragment are currently being studied since fragmentation and reaction processes in LIBS events are not fully understood at this time.

Explosively formed projectiles (EFP) are a major problem in terrorism and asymmetrical warfare. EFPs are often triggered by ordinary infrared motion detectors. A potential weak link is that such electronics are not hardened to ionizing radiation and can latch-up or enter other inoperative states after exposure to a single short event of ionizing radiation. While these can often be repaired with a power restart, they also can produce shorts and permanent damage. A problem of course is that we do not want to add radiation exposure to the long list of war related hazards. Biological systems are highly sensitive to integrated dosage but show no particular sensitivity to short pulses. There may be a way to generate short pulsed subsoil radiation to deactivate concealed electronics without introducing radiation hazards to military personnel and civilian bystanders. Electron beams of 30 MeV that can be produced by portable linear accelerators (linacs) propagate >20 m in air and 10-12 cm in soil. X-radiation is produced by bremsstrahlung and occurs subsoil beneath the point of impact and is mostly forward directed. Linacs 1.5 m long can produce 66 MWatt pulses of subsoil x-radiation 1 microsecond or less in duration. Untested as yet, such a device could be mounted on a robotic vehicle that precedes a military convoy and deactivates any concealed electronics within 10-20 meters on either side of the road.

The development of Improvised Explosive Devices (IED's) by insurgents in Colombia is characterized by a quick response to counter IED measures. Many current IED's do not contain any metal parts and can have any shape or form. Due to the low metal content or the absence of any metal parts, sensors based on metal detection are not useful anymore. Due to the wide variety of sizes, shapes, and enclosure materials of current IED's, one and two-dimensional GPR sensors using a "library" of known shapes as well as acoustic sensors using material characteristic frequencies have become ineffective. Therefore, the Colombian experience strongly suggests that chemical sensors are the way for IED detection in soils since they do not depend on IED metal content, size, or shape but only on the presence of explosives, a necessary ingredient for any IED. Promising recently developed chemical sensors make use of semiconducting organic polymers (SOPs) such as FIDO and laser-induced breakdown spectroscopy (LIBS). Once an explosive has been detected, the IED needs to be identified and located. Therefore, there is a need for three-dimensional high resolution scans for identification of all subsoil features including rocks, roots, and IED's. The recently developed 3D-GPR (Ground Penetrating Radar) can map all features of the subsoil with a spatial resolution of about 2 cm or less. The objectives of this contribution are to inform about the IED problem in Colombia and how novel technologies may contribute to humanitarian IED clearance under humid tropical conditions.

In this paper we present results of experimental validation of a new methodology for anti-personnel mine (APM) detection for humanitarian demining, proposed by the authors and previously validated only by simulation. The technique is based on local heating and sensing by contactless thermometers (pyrometers). A large sand box (2.6m³) has been realized and fitted with a cart moving on rails and holding instrumentation. Accurate mine surrogates have been hidden in the sand together with confounders. Preliminary measurements are consistent with simulations and prove validity of the approach.

The EC DELVE Support Action project has analyzed the bottlenecks in the transfer of Humanitarian Demining (HD) technology from technology development to the use in the field, and drawn some lessons learned, basing itself on the assessment of the European Humanitarian Demining Research and Technology Development (RTD) situation from early 1990 until 2006. The situation at the European level was analyzed with emphasis on activities sponsored by the European Commission (EC). This was also done for four European countries and Japan, with emphasis on national activities. The developments in HD during the last 10 years underline the fact that in a number of cases demining related developments have been terminated or at least put on hold. The study also showed that the funding provided by the EC under the Framework Program for RTD has led directly to the creation of an extensive portfolio of Humanitarian Demining technology development projects. The latter provided a range of research and supporting measures addressing the critical issues identified as a result of the regulatory policies developed in the field of Humanitarian Demining over the last ten years. However, the range of instruments available to the EC to finance the necessary research and development were limited, to pre-competitive research. The EC had no tools or programs to directly fund actual product development. As a first consequence, the EC funding program for development of technology for Humanitarian Demining unfortunately proved to be largely unsuitable for the small-scale development needed in a field where there is only a very limited market. As a second consequence, most of the research has been demonstrator-oriented. Moreover, the timeframe for RTD in Humanitarian Demining has not been sufficiently synchronized with the timeframe of the EC policies and regulations. The separation of the Mine Action and RTD funding streams in the EC did also negatively affect the take-up of new technologies. As a conclusion, creating coherence between: (1) the EC policy based on political decisions, (2) RTD, testing and industrialization of equipment, and (3) timely deployment, requires a new way of coordinated thinking: "end-to-end planning" has to be supported by a well organized and coordinated organizational structure involving different DGs and even extending beyond the EU. This was not the case for Mine Action, but appears today to be the case for Environmental Risk Management.

An improved automatic target recognition (ATR) processing string has been developed. The overall processing string consists of pre-processing, subimage adaptive clutter filtering (SACF), normalization, detection, data regularization, feature extraction, optimal subset feature selection, feature orthogonalization and classification processing blocks. A new improvement was made to the processing string, data regularization, which entails computing the input data mean, clipping the data to a multiple of its mean and scaling it, prior to feature extraction. The classified objects of 3 distinct 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, which resulted in a 3:1 reduction in false alarms. Two significant fusion algorithm improvements were made. First, a nonlinear 2nd order (Volterra) feature LLRT fusion algorithm was developed. Second, a repeated application of a subset Volterra feature selection / feature orthogonalization / LLRT fusion block was utilized. It was shown that cascaded Volterra feature LLRT fusion of the ATR processing strings outperforms baseline summing and single-stage Volterra 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.

In this paper a new coherence-based feature extraction method for sonar imagery generated from two disparate sonar systems is developed. Canonical correlation analysis (CCA) is employed to identify coherent information from co-registered regions of interest (ROI's) that contain target activities, while at the same time extract useful coherent features from both images. The extracted features can be used for simultaneous detection and classification of target and non-target objects in the sonar images. In this study, a side-scan sonar that provides high resolution images with good target definition and a broadband sonar that generates low resolution images, but with reduced background clutter. The optimum Neyman-Pearson detector will be presented and then extended to the dual sensor platform scenarios. Test results of the proposed methods on a dual sonar imagery data set provided by the Naval Surface Warfare Center (NSWC) Panama City, FL will be presented. This database contains co-registered pair of images over the same target field with varying degree of detection difficulty and bottom clutter. The effectiveness of CCA as the optimum detection tool is demonstrated in terms of probability of detection and false alarm rate.

Single-point statistical properties of envelope-detected data such as signal returns from synthetic aperture radar and sonar have traditionally been modeled via the Rayleigh distribution and more recently by the K-distribution. Two-dimensional correlations that occur in textured non-Gaussian imagery are more difficult to model and estimate than Gaussian textures due to the nonlinear transformations of the time series data that occur during envelope detection. In this research, textured sonar imagery is modeled by a correlated K-distribution. The correlated K-distribution is explained via the compound representation of the one-dimensional K-distribution probability density function. After demonstrating the model utility using synthetically generated imagery, model parameters are estimated from a set of textured sonar images using a nonlinear least-squares fit algorithm. Results are discussed with regard to texture segmentation applications.

This paper describes our attempts to model sea bottom textures in high-frequency synthetic aperture sonar imagery using a Gaussian Markov random field. A least-squares estimation technique is first used to estimate the model parameters of the down-sampled grey-scale sonar images. To qualitatively measure estimation results, a fast sampling algorithm is then used to synthesize the sea bottom textures of a fourth-order Gaussian Markov random field which is then compared with the original sonar image. A total of four types of sea floor texture are used in the case study. Results show that the 4th order GMRF model mimics patchy sandy textures and sand ripple, but does not reproduce more complex textures exhibited by coral and rock formations.

There are approximately one million acres of underwater lands at Department of Defense (DOD) and Department of Energy (DOE) sites that are highly contaminated with unexploded ordnance (UXO) and land mines. The detection and disposal of Underwater Military Munitions are more expensive than excavating the same targets on land. Electromagnetic induction (EMI) sensing has emerged as one of the most promising technologies for underwater detection. In order to explore the full potential of various EMI sensing technologies for underwater detection and discrimination, to achieve a high (~100%) probability of detection, and to distinguish UXO from non-UXO items accurately and reliably, first the underlying physics of EM scattering phenomena in underwater environments needs to be investigated in great detail. This can be achieved by using an accurate 3D numerical code, such as the combined method of auxiliary sources and thin skin depth approximation (MAS/TSA), the pseudospectral time-domain technique, finite element methods or other approaches. This paper utilizes the combined MAS/TSA, originally developed for detection and discrimination of highly conducting and permeable metallic objects placed in an environment with zero or negligible conductivity. Here, first the theoretical basis of the MAS/TSA is presented for metallic objects placed in an electrically conductive environment. Then numerical experiments are conducted for homogeneous targets of canonical (spheroidal) shapes subject to frequency- or time-domain illumination. The results illustrate coupling effects between the object and its surrounding conductive medium, particularly at high frequencies (early times for time-domain sensors), and the way this coupling depends on the distance between the sensor and the object's center.

Airborne EO imagery, including wideband, hyperspectral, and multispectral modalities, has greatly enhanced the ability of the ISR community to detect and classify various targets of interest from long standoff distances and with large area coverage rates. The surf zone is a dynamic environment that presents physical and operational challenges to effective remote sensing with optical systems. In response to these challenges, BAE Systems has developed the Tactical Multi-spectral (TACMSI) system. The system includes a VNIR six-band multispectral sensor and all other hardware that is used to acquire, store and process imagery, navigation, and supporting metadata on the airborne platform. In conjunction with the hardware, BAE Systems has innovative data processing methods that exploit the inherent capabilities of multi-look framing imagery to essentially remove the overlying clutter or obscuration to enable EO visualization of the objects of interest.

The detection of buried land mines in soil is a well-studied problem; many existing technologies are designed and optimized for performance in different soil types. Research on mine detection in shallow water environments such as beaches, however, is much less developed. Electrical impedance tomography (EIT) shows promise for this application. EIT uses current-stimulating and voltage-recording electrode pairs to measure trans-impedances in the volume directly beneath the electrode array, which sits flat over the ground surface. The trans-impedances are used to construct a conductivity profile of the volume. Non-metallic and metallic explosives appear as perturbations in the conductivity profile, and their location and size can be estimated. Lab testing has yielded promising results using a submerged array positioned over a sand bed. The instrument has also successfully detected surrogate mines in a traditional soil environment during field trials. Resolution of the detector is roughly half the pitch of electrodes in the array. In underwater lab testing, non-conducting targets buried in the sand are detected at a depth of 1.5 times the electrode pitch with the array positioned up to one electrode pitch above the sand bed. Results will be presented for metallic and non-metallic targets of various shapes and sizes.

An intelligent robotic system can be distinguished from other machines by its ability to sense, learn, and react to its environment despite various task uncertainties. One of the most powerful sensing modality for robotic system is vision as it enables the robot to see its environment, recognize objects around it and interact with objects to accomplish its task. This paper discusses vision enabling techniques that allows a robot to detect, characterize, classify, and discriminate UneXploded Ordnance (UXO) from clutters in unstructured environments. A soft-computing approach is proposed and validated via indoor and outdoor experiments to measure its performance efficiency and effectiveness in correctly detection and classifying UXO vs. XO and other clutter. The proposed technique has many potential applications for military, homeland security, law enforcement, and in particular, environment UXO remediation and clean-up operations.

The Engineering Research and Development Center participated in several field programs, mainly in desert areas, using ground-based and airborne thermal imagers and radiometers to investigate the thermal signatures of disturbed and undisturbed soils, including disturbed soils over buried munitions. Analysis of the thermal imagery indicates the thermal temperature difference between the disturbed and undisturbed soil varies diurnally. The thermal temperature differences have similar diurnal patterns for the different field programs and different environmental conditions. This paper presents the analysis of the field measurements and model simulations used to quantify the observed thermal temperature differences.

Over the past five years, advances have been made in the spectral detection of surface mines under minefield detection programs at the U. S. Army RDECOM CERDEC Night Vision and Electronic Sensors Directorate (NVESD). The problem of detecting surface land mines ranges from the relatively simple, the detection of large anti-vehicle mines on bare soil, to the very difficult, the detection of anti-personnel mines in thick vegetation. While spatial and spectral approaches can be applied to the detection of surface mines, spatial-only detection requires many pixels-on-target such that the mine is actually imaged and shape-based features can be exploited. This method is unreliable in vegetated areas because only part of the mine may be exposed, while spectral detection is possible without the mine being resolved. At NVESD, hyperspectral and multi-spectral sensors throughout the reflection and thermal spectral regimes have been applied to the mine detection problem. Data has been collected on mines in forest and desert regions and algorithms have been developed both to detect the mines as anomalies and to detect the mines based on their spectral signature. In addition to the detection of individual mines, algorithms have been developed to exploit the similarities of mines in a minefield to improve their detection probability. In this paper, the types of spectral data collected over the past five years will be summarized along with the advances in algorithm development.

Georgia Tech recently initiated a weathering effects measurement program to monitor the optical properties of several common building materials. A set of common building materials were placed outdoors and optical property measurements made over a series of weeks to assess the impact of exposure on these properties. Both reflectivity and emissivity measurements were made. Materials in this program included aluminum flashing, plastic sheets, bricks, roof shingles, and tarps. This paper will discuss the measurement approach, experimental setup, and present preliminary results from the optical property measurements.

Often in hyperspectral overhead land mine imagery, there exists clutter with similar spatial and spectral characteristics to those of land mines. However groups of clutter features are rarely related spatially in the same way that groups of mines are related. For this reason, recognition of field patterns in overhead land mine imagery is critical to the detection of mine fields. The material presented here addresses means by which to spatially sample overhead hyperspectral imagery for the accentuation of mine field patterns. Our initial approach is to assume that the mines are laid out in a particular field pattern. We then search for spectral anomalies that are spatially distributed according to such a pattern. For this purpose, we utilize an RX detector with locally estimated mean and covariance matrix. We then use the pattern to predict the locations of additional mines. These locations provide us with search regions for the use of a second anomaly detector, in this case we use an anomaly detector based upon an eigenspace separation transform. Examples are provided using LWIR imagery.

The US Army's RDECOM CERDEC Night Vision and Electronic Sensors Directorate (NVESD), Countermine Division is evaluating the compressibility of airborne multi-spectral imagery for mine and minefield detection application. Of particular interest is to assess the highest image data compression rate that can be afforded without the loss of image quality for war fighters in the loop and performance of near real time mine detection algorithm. The JPEG-2000 compression standard is used to perform data compression. Both lossless and lossy compressions are considered. A multi-spectral anomaly detector such as RX (Reed & Xiaoli), which is widely used as a core algorithm baseline in airborne mine and minefield detection on different mine types, minefields, and terrains to identify potential individual targets, is used to compare the mine detection performance. This paper presents the compression scheme and compares detection performance results between compressed and uncompressed imagery for various level of compressions. The compression efficiency is evaluated and its dependence upon different backgrounds and other factors are documented and presented using multi-spectral data.

Mine fields are often distinguishable in overhead hyperspectral LWIR imagery due to the spatial pattern in which the mines are laid. Recognition of these field patterns in overhead landmine imagery shows promise for enhancing the ability to detect mine fields. However, before one can search for a field pattern in an image, it is necessary to determine the orientation and size of the pattern within the image, should it exist. We present a method for determining likely scales and orientation for grids of landmines. The approach is to consider pairs of interest points and then look for patterns in the slopes of the lines connecting them. The dominant slope then determines an orientation angle. Next, we look for patterns in the distances between pairs of points that have a slope close to the orientation angle. An application to detecting mine fields via recognition of patterns of features in hyperspectral LWIR imagery is given.

Soil properties have a significant impact in the observed responses of various sensors for mine detection. Ground penetrating radar (GPR) is an important sensor for mine detection. The performance GPR is largely governed by the soil moisture content. Characterizing the spatial and temporal changes in the dielectric properties of soil surrounding the landmines represents a major challenge for radar evaluation studies. Laboratory and field studies are currently in progress to better document the effect of soil moisture variability on radar sensing of buried landmines. These studies are conducted using commercially available GPRs operating at 400 MHz and 1.5 GHz. The study site is a government mine test facility with various anti-tank (AT) and anti-personnel (AP) mines buried at different depths. The test lanes at this facility are grass-covered and the sub-surface root system plays an important role in modulating the soil properties. Our goal is to investigate the seasonal changes in soil processes at this site and to document how these processes impact the radar signatures of landmines.

Soil properties make a significant impact in the observed responses of various sensors for subsurface target detection. Ground penetrating radars (GPRs) have been extensively researched as a tool for subsurface target detection. A key soil parameter of interest for evaluating GPR performance is the soil attenuation rate. The information about the soil attenuation rate coupled with target properties (size, shape, material properties and depth of burial) can be used to estimate the effectiveness of radar sensors in a particular soil environment. Radar attenuation in desert soil is of interest in today's political and military climate. Laboratory measurements of desert soil attenuation were conducted using samples collected from a desert in Southwestern United States and in Iraq. These measurements were made in a coaxial waveguide over the frequency ranging from 250 MHz to 4 GHz. The soil grain size distribution, mineralogy, moisture and salinity were also measured. This report describes the experimental procedure and presents the radar attenuation rates observed in desert soils. The results show that the soluble salt content is an important parameter affecting the attenuation behavior of desert soils.

A thorough understanding of thermal soil regimes is critical information for a wide variety of disciplines and engineering applications as well as for the evaluation of potential and limitations of thermal and optical sensors. In this study we have developed a procedure for the evaluation of global thermal soil regimes. First, pedotransfer functions are used to derive thermal soil properties (volumetric soil heat capacity and thermal conductivity) from readily available soil data on texture, bulk density, and organic carbon. Next, the average annual soil temperature is derived from the average annual air temperature. Then, the thermal top boundaries are derived either for well-watered sites using the daily and annual air temperature amplitudes as proxies for the daily and annual soil surface temperature amplitudes or for a wide range of environmental conditions using the model HYDRUS1D. A thorough validation of the proposed procedure is needed for the quantification of the probability with which soil thermal regimes can be predicted.

One of the Department of Defense's most pressing environmental problems is the efficient detection and identification of unexploded ordnance (UXO). In regions of highly magnetic soils, magnetic and electromagnetic sensors often detect anomalies that are of geologic origin, adding significantly to remediation costs. In order to develop predictive models for magnetic susceptibility, it is crucial to understand modes of formation and the spatial distribution of different iron oxides. Most rock types contain iron and their magnetic susceptibility is determined by the amount and form of iron oxides present. When rocks weather, the amount and form of the oxides change, producing concomitant changes in magnetic susceptibility. The type of iron oxide found in the weathered rock or regolith is a function of the duration and intensity of weathering, as well as the original content of iron in the parent material. The rate of weathering is controlled by rainfall and temperature; thus knowing the climate zone, the amount of iron in the lithology and the age of the surface will help predict the amount and forms of iron oxide. We have compiled analyses of the types, amounts, and magnetic properties of iron oxides from soils over a wide climate range, from semi arid grasslands, to temperate regions, and tropical forests. We find there is a predictable range of iron oxide type and magnetic susceptibility according to the climate zone, the age of the soil and the amount of iron in the unweathered regolith.

Magnetic soils are a major source of false positives when searching for unexploded ordnance with electromagnetic induction sensors. In adverse areas up to 30% of identified electromagnetic induction anomalies have been attributed to geology. In the presence of magnetic soil, sensor movement and surface topography can cause anomalies in the data that have similar size and shape to those from compact metallic targets. In areas where the background geological response is small relative to the response of metallic targets, electromagnetic induction data can be inverted for the dipole polarization tensor. However, spatially correlated noise from the presence of a geologic background greatly reduces the accuracy of dipole polarization estimates. In this presentation we examine the effects of sensor movement on the measured EM response of a magnetic background signal. We demonstrate how sensor position and orientation information can be used to model the background soil response and improve estimates of a target's dipole polarization tensor.

Since 2002, our research group at Tohoku University has developed a new hand-held land mine detection dual-sensor ALIS. ALIS is equipped with a metal detector and a GPR, and it has a sensor tracking system, which can record the GPR and Metal detector signal with its location. It makes possible to process the data afterwards, including migration. The migration processing drastically increases the quality of the image of the buried objects. ALIS evaluation test was conducted in Croatia in October 2007. Then after, we stared a half-year evaluation test of ALIS in QC test in Croatia in December 2007. This test will be conducted in various soil and environmental conditions in Croatia.

Object depth is a simple characteristic that can indicate an object's type. Popular instruments like radar, metal detectors, and magnetometers are often used to detect the presence of a subsurface object. The next question is often, "How deep is it?" Determining the answer, however, is not as straight forward as might be expected. This paper explores the determination of depth using metal detectors. More specifically, it looks at a popular metal detector (the Geonics EM61) and makes use of its vertically separated coils to generate a depth estimate. Estimated depths are shown for UXO and small surface clutter from flush buried down to 48". Ultimately a statistical depth resolution is determined. An alternative approach is then considered that casts the depth determination problem as one of classification. Only two classes are considered important "deep" and "shallow". Results are shown that illustrate the utility of the classifier approach. The traditional estimator can provide a depth estimate of the object, but the classifier approach can distinguish between small shallow, large deep, and large shallow object classes.

This paper proposes a technique for using infrared (IR) imagery to eliminate false forward-looking ground penetrating radar (FLGPR) detections by examining areas in IR images corresponding to FLGPR alarm locations. The FLGPR and IR co-location is based on the assumption of a flat earth and the pinhole camera model. The parameters of the camera and its location on the vehicle are not assumed to be known. The parameters of the model are estimated using a set of correspondences gathered from the data utilizing the covariance matrix adaptation evolution strategy (CMA-ES) optimization algorithm. Detection of false alarms is accomplished by generating a descriptor, consisting of various statistics calculated from the IR images along with the FLGPR confidence value, for each alarm location. The alarms are then classified based on the Mahalanobis distance between their descriptor and a multivariate normal distribution used to model false alarms. The false alarm distribution is computed from training data where the validity of each alarm location is already known. Using this technique, generally fifteen to twenty percent or more of the FLGPR false alarms can be eliminated without losing any true alarms.

Previous research has developed an information-theoretic sensor management framework for improving static target detection performance. This framework has been successfully applied to a large dataset of real landmine data; performance using the sensor manager on this dataset was demonstrated to be superior to performance using a direct search technique in which sensors blindly sweep through the gridded region of interest. In previous work, the sensor manager has modeled the observations made in each grid cell as being independent from the other observations made in that cell by the same sensor and also as being independent from observations made in that cell by other sensors. Such a modeling approach fails to account for the correlations that will result between observations made both by the same and different sensors. This paper alters the modeling framework that has been used previously to incorporate observation correlation, which will more realistically model the interrelationships between sensor observations. After introducing the new modeling approach, results are then presented that compare the performance of the sensor manager to the performance of an unmanaged direct search procedure. The sensor manager is again demonstrated to outperform direct search. Furthermore, the performance effects of modeling and failing to model correlation are examined through simulation. Failing to model correlation that is present in the data is demonstrated to substantially degrade performance and cause direct search to outperform the sensor manager. However, when correlated modeling is used to model correlated data, the sensor manager is again demonstrated to outperform direct search.

In this paper, a fast approximate version of the Kernel RX-algorithm, termed FastKRX is presented. The original Kernel RX-algorithm is reformulated using a spatial weighting function. In the proposed framework, a single kernel Gram matrix is defined over the entire image domain, and the detector statistics for the whole image can be obtained directly from the centered kernel Gram matrix. A methodology based on spatial-spectral clusters is presented for the fast computation of the centered kernel Gram matrix using a multivariate Taylor series approximation. Comparative detection performance on representative airborne multispectral data for both the proposed FastKRX algorithm and the RX anomaly detector is presented. Comparative computational complexity and results on speed of execution are also presented.

In a typical minefield detection problem, the minefield decision is based on the number of detected targets in a given field segment. The detected target locations are obtained by an anomaly detector, such as the RX, using constant target rate (CTR) or constant false alarm rate (CFAR) thresholding. Specific shape and spectral features at the detection locations are used to assign "mineness" or "non-mineness" measures to the detections, which are further used for false alarm mitigation (FM). The remaining detections after FM are used to assign a minefield metric based on a spatial point process (SPP) formulation. This paper investigates how this "mineness" attribute of the detected targets can be exploited to improve the performance of scatterable minefield detection over and above that which is possible by FM. The distribution of the detections in the segment is formulated as a marked point process (MPP), and the minefield decision is based on the log-likelihood ratio test of a binary hypothesis problem. An elegant, linear complexity algorithm is developed to maximize this log-likelihood ratio. An iterative expectation maximization algorithm is used to estimate the unknown probability of the detection of mines. The minefield detection performance, based on SPP with false alarm mitigation and MPP formulation under both CTR and CFAR thresholding methods, is compared using thousands of simulated minefields and background segments.

We present an algorithm suite, Hybrid Airborne Mine Detector (HAMD), developed for the detection of small scatterable surface and buried mines, using multispectral airborne images. This algorithm suite is composed of a number of components designed for specific tasks such as image segmentation based on unsupervised clustering, localized image enhancement, generalized signature extraction and construction, and mine classification and fusion. Since both surface and buried mines in low contrast images are difficult to detect, a new algorithm has been developed to enhance images locally. The signature extraction component extracts different signatures based on surface or buried mines. To extract small surface mine signatures, moment invariance (MI) is used. However, to extract buried mine signatures, thermal variations and spatial distributions are employed. To make the system suitable for different operational environments, a small number of general signatures are constructed and stored in the signature library. Test results based on airborne images have shown that signatures collected can be used to detect mines placed in different environments such as vegetation and sandy areas. For mine classification and false alarm mitigation, statistical hypothesis tests, such as Fisher's Discriminant Ratio (FDR) test and the Kolmogorov-Smirnov (KS) test, are used.

In remotely sensed hyperspectral imagery, many samples are collected on a given flight and many variable factors contribute to the distribution of samples. Measurements made at different flight times over the same swath may result in different spectral responses due to various environmental conditions and sensor calibration. Many classification methods attempt to classify a sample using labeled datasets or a priori information about the samples. We present a possibilistic context-based approach for class estimation within a random set model. This approach includes novel formulations for model parameters with an intuitive base in probability and measure theory. This approach implicitly retains contextually correlated information in the data and uses it to estimate class labels in the presence of unknown factors-hidden contexts. This new method is applied to AHI (hyperspectral) imagery for the purposes of landmine detection. The results are compared to conventional methods and analyzed.

Recent advances in ground penetrating radar (GPR) design and fabrication have resulted in improved fidelity responses from relatively small, shallow-buried objects like landmines and improvised explosive devices. As the responses measured with GPR improve, more and more advanced processing techniques can be brought to bear on the problem of target identification in GPR data. From an electromagnetic point of view, the problem of target detection in GPR signal processing is reducible to inferring the presence or absence of changes in the electromagnetic properties of soils and thus the presence or absence of buried targets. Problems arise because the algorithms required for the full electromagnetic inversion of GPR signals are extremely computationally expensive, and usually rely on assumptions of electromagnetically constant transmission media; these problems typically make the real-time implementation of purely electromagnetic-inspired algorithms infeasible. On the other hand, purely statistical or signal-processing inspired approaches to target identification in GPR often lack a solid theoretical basis in the underlying physics, which is fundamental to understanding responses in GPR. In this work, we propose a model for responses in time-domain ground penetrating radar that attempts to incorporate the underlying physics of the problem, but avoids several of the issues inherent in assuming constant media with known electrical parameters by imposing a statistical model over the observed parameters of interest in A-scans - namely the signal gains, times of arrival, etc. The spatial requirements of the proposed statistical model suggests the application of Markov random field (MRF) distributions which provide expressive, but computationally simple models of spatial interactions. In this work we will explore the application of physics-based MRF's as generative models for time-domain GPR data, the pre-screening algorithms that this model motivates, and discuss how the model can be extended to other applications in GPR processing. Preliminary results showing how the MRF approach to understanding the underlying physics can improve performance are also shown.

In this paper, we propose an efficient Discrete Hidden Markov Models (DHMM) for landmine detection that rely on training data to learn the relevant features that characterize different signatures (mines and non-mines), and can adapt to different environments and different radar characteristics. Our work is motivated by the fact that mines and clutter objects have different characteristics depending on the mine type, soil and weather conditions, and burial depth. Thus, ideally different sets of specialized features may be needed to achieve high detection and low false alarm rates. The proposed approach includes three main components: feature extraction, clustering, and DHMM. First, since we do not assume that the relevant features for the different signatures are known a priori, we proceed by extracting several sets of features for each signature. Then, we apply a clustering and feature discrimination algorithm to the training data to quantize it into a set of symbols and learn feature relevance weights for each symbol. These symbols and their weights are then used in a DHMM framework to learn the parameters of the mine and the background models. Preliminary results on large and diverse ground penetrating radar data show that the proposed method outperforms the basic DHMM where all the features are treated equally important.

This paper proposes the use of subspace approach to model the energy density spectra (EDS) of landmine targets, for the purpose to improve the detection of weak scattering landmines and their discrimination with clutter objects. The effectiveness of subspace technique to model the landmine EDS depends on the subspace selection. A slight modification of the eigenspace separation transform was applied to generate the subspace basis. In addition to the experimental results of performance improvement through subspace modeling, fusion performance with the Hidden Markov Model (HMM), the Edge Histogram Descriptor (EHD) and the NUKEv6 confidence values will be provided.

The paper discusses the fusing of results from classifiers and discriminant functions. It explores the relationship between the Bayesian opinion pooling of classifier results and the linear pooling of ranks generated by normalizing discriminant values. This work is closely related to current research rank preference aggregation. I discuss a method of using gradient ascent to choose appropriate aggregation parameters from a surface comprising the Wilcoxon-Mann-Whitney statistic and discuss the importance of including negative weights in that search. Finally I recap the results of applying this method to a large-scale collection of landmine data from a vehicular mounted mine detection system.

We present a novel method for fusing the results of multiple landmine detection algorithms that use different types of features and different classification methods. Our approach, called Context Extraction for Local Fusion (CELF), is motivated by the fact that the relative performance of different detectors can vary significantly depending on the mine type, geographical site, soil and weather conditions, and burial depth. CELF 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. This objective function is defined and optimized to produce contexts via unsupervised clustering while simultaneously providing optimal fusion parameters for each context. Results on large and diverse Ground Penetrating Radar data collections show that the proposed method can identify meaningful and coherent contexts and that different expert algorithms can be identified for the different contexts. Typically, the contexts correspond to groups of alarm signatures that share common attributes such as mine type, geographical site, soil and weather conditions. Our initial experiments have also indicated that the proposed context-dependent fusion outperforms all individual detectors and other standard fusion methods.

The prohibitive costs of excavating all geophysical anomalies are well known and are one of the greatest impediments to efficient clean-up of unexploded ordnance (UXO)-contaminated lands at Department of Defense (DoD) and Department of Energy (DOE) sites. Innovative discrimination techniques that can reliably distinguish between hazardous UXO and non-hazardous metallic items are required. The key element to overcoming these difficulties lies in the development of advanced processing techniques that can treat complex data sets to maximize the probability of accurate classification and minimize the false alarm rate. To address these issues, this paper uses a new approach that combines a physically complete EMI forward model called the Generalized Standardized Excitation Approach (GSEA) with a statistical signal processing approach named Mixed Modeling (MM). UXO discrimination requires the inversion of digital geophysical data, which could be divided into two pars: 1) linear - estimating model parameters such as the amplitudes of the responding GSEA sources and 2) non-linear - inverting an object's location and orientation. Usually the data inversion is an ill-posed problem that requires regularization. Determining the regularization parameter is not straightforward, and in many cases depends on personal experience. To overcome this issue, in this paper we employ the statistical approach to estimate regularization parameters from actual data using the un-surprised mixed model approach. In addition, once the non-linear inverse scattering parameters are estimated then for UXO discrimination a covariance matrix and confidence interval are derived. The theoretical basis and practical realization of the combined GSEA-Mixed Model algorithm are demonstrated. Discrimination studies are done for ATC-UXO sets of time-domain EMI data collected at the ERDC UXO test stand site in Vicksburg, Mississippi.

The Man Portable Vector (MPV) sensor is a new mono/multistatic time-domain EMI detector that provides a detailed electromagnetic picture of a target by measuring all three magnetic field components at five distinct receiver positions in over 100 time channels. We have adapted the data-derived Standardized Excitation Approach (SEA) to this sensor. The SEA has been found in the past to make sound predictions in near-field situations, where schemes like the dipole model fail, and in cases where the target under interrogation is heterogeneous and the interactions between its different sections affect the detectable signal. The method replaces a given target with a set of sources placed on a surrounding spheroid and decomposes the sensor primary field into a set of standardized modes. Each of these modes elicits a response from the sources that is intrinsic to the object; it is only the relative weights of the modes that vary with the position and orientation of the target relative to the sensor. The strengths of the sources can be determined by fitting experimental data. Here we review some of the results we obtain when we apply the technique to problems relevant to the identification of unexploded ordnance (UXO). We extract the source parameters using high-quality measurements collected at a UXO test stand and invert unused data sets for location and to discriminate between different objects. We carry out similar experiments with buried objects in order to assess the performance of the method in realistic situations.

Magnetic soils are a major source of false positives when searching for landmines or unexploded ordnance (UXO) with electromagnetic induction sensors. The viscosity effects of magnetic soil can be accurately modeled by assuming a ferrite relaxation with a log-uniform distribution of time constants. The frequency domain response of ferrite soils has a characteristic negative log-linear in-phase and constant quadrature component. After testing and validating that assumption, we process frequency domain electromagnetic data collected over UXO buried in a viscous remanent magnetic host. The first step is to estimate a spatially smooth background magnetic susceptibility model from the sensor. The response of the magnetically susceptibility background is then subtracted from the sensor data. The background removed data are then inverted to obtain estimates of the dipole polarization tensor. This technique is demonstrated for the discrimination of UXO with hand-held Geophex GEM3 data collected at a contaminated site near Denver, Colorado.