This PDF file contains the front matter associated with SPIE Proceedings Volume 9823, including the Title Page, Copyright information, Table of Contents, and Conference Committee listing.

Intermediate electrical conductivity (IEC) materials (10¹S/m < σ < 10⁴S/m), such as carbon fiber (CF), have recently been used to make smart bombs. In addition, homemade improvised explosive devices (IED) can be produced with low conducting materials (10^-4S/m < σ < 1S/m), such as Ammonium Nitrate (AN). To collect unexploded ordnance (UXO) from military training ranges and thwart deadly IEDs, the US military has urgent need for technology capable of detection and identification of subsurface IEC objects. Recent analytical and numerical studies have showed that these targets exhibit characteristic quadrature response peaks at high induction frequencies (100kHz − 15MHz, the High Frequency Electromagnetic Induction (HFEMI) band), and they are not detectable with traditional ultra wideband (UWB) electromagnetic induction (EMI) metal detectors operating between 100Hz − 100kHz. Using the HFEMI band for induction sensing is not so simple as driving existing instruments at higher frequencies, though. At low frequency, EMI systems use more wire turns in transmit and receive coils to boost signal-to-noise ratios (SNR), but at higher frequencies, the transmitter current has non-uniform distribution along the coil length. These non-uniform currents change the spatial distribution of the primary magnetic field and disturb axial symmetry and thwart established approaches for inferring subsurface metallic object properties. This paper discusses engineering tradeoffs for sensing with a broader band of frequencies ever used for EMI sensing, with particular focus on coil geometries.

A method using feedback is presented that reduces several measurement errors inherent in electromagnetic induction sensors. Errors associated with coupling between receive coils and errors associated with operating near magnetic soils will both be reduced. The method uses feedback that is directly injected into the receive coils and does not require secondary coils. A simple circuit is introduced to perform the feedback and is optimized to reduce the errors and make the circuit stable. Experimental results are presented to show the effectiveness of the feedback.

Continuous-wave (CW) electromagnetic induction (EMI) systems used for subsurface sensing typically employ separate transmit and receive coils placed in close proximity. The closeness of the coils is desirable for both packaging and object pinpointing; however, the coils must have as little mutual coupling as possible. Otherwise, the signal from the transmit coil will couple into the receive coil, making target detection difficult or impossible. Additionally, mineralized soil can be a significant problem when attempting to detect small amounts of metal because the soil effectively couples the transmit and receive coils. Optimization of wire coils to improve their performance is difficult but can be made possible through a stream-function representation and the use of partially convex forms. Examples of such methods have been presented previously, but these methods did not account for certain practical issues with coil implementation. In this paper, the power constraint introduced into the optimization routine is modified so that it does not penalize areas of high current. It does this by representing the coils as plates carrying surface currents and adjusting the sheet resistance to be inversely proportional to the current, which is a good approximation for a wire-wound coil. Example coils are then optimized for minimum mutual coupling, maximum sensitivity, and minimum soil response at a given height with both the earlier, constant sheet resistance and the new representation. The two sets of coils are compared both to each other and other common coil types to show the method’s viability.

The implementation of new advanced electromagnetic induction (EMI) sensor surveys at sites containing unexploded ordnance (UXO) and explosive remnants of war (ERW) is an effective method for accurate mapping and for discriminating clutter from targets of interest. We present development and integration of a next generation advanced EMI sensor onto a cart-based sensing platform to combine the mapping capability of previous digital geophysical survey instruments with the high-resolution discrimination capability of advanced characterization arrays. The EMI sensor employs a multi-axis receiver configuration to produce data sufficient for anomaly discrimination. We discuss platform design and development, data acquisition and post-processing software development, and results from field tests demonstrating the detection and discrimination capability of the cart-based system. Platform development and design focused on navigation and EMI sensor integration onto a custom, low-noise, metal-free platform. Data acquisition is via an Android application with emphasis on ease-of-use and real-time quality control (QC) of collected data. Post-processing methods emphasize QC, inversion-based anomaly location estimation, and automated or supervised polarizability-based discrimination methods to produce a prioritized dig list. Integration of the detection, clutter rejection and QC methods into the post-processing software module reduces the time required between sensor data collection and generation of a prioritized dig list. System concept of operations (CONOPs), data collection, QC, data processing procedures, and performance against various clutter objects and targets of interest will also be discussed.

Recently developed feature extraction methods proposed in the explosive hazard detection community have yielded many features that potentially provide complementary information for explosive detection. Finding the right combination of features that is most effective in distinguishing targets from clutter, on the other hand, is extremely challenging due to a large number of potential features to explore. Furthermore, sensors employed for mine and buried explosive hazard detection are typically sensitive to environmental conditions such as soil properties and weather as well as other operating parameters. In this work, we applied Bayesian cross-categorization (CrossCat) to a heterogeneous set of features derived from electromagnetic induction (EMI) sensor time-series for purposes of buried explosive hazard detection. The set of features used here includes simple, point-wise measurements such as the overall magnitude of the EMI response, contextual information such as soil type, and a new feature consisting of spatially aggregated Discrete Spectra of Relaxation Frequencies (DSRFs). Previous work showed that the DSRF characterizes target properties with some invariance to orientation and position. We have developed a novel approach to aggregate point-wise DSRF estimates. The spatial aggregation is based on the Bag-of-Words (BoW) model found in the machine learning and computer vision literatures and aims to enhance the invariance properties of point-wise DSRF estimates. We considered various refinements to the BoW model for purpose of buried explosive hazard detection and tested their usefulness as part of a Bayesian cross-categorization framework on data collected from two different sites. The results show improved performance over classifiers using only point-wise features.

Detecting and classifying small (i.e., with calibers ranging from 20 to 60 mm) and deep targets (burial depth more than 11 times targets diameter) is still a challenging problem using current advanced EMI sensors and signal processing approaches. In order to overcome this problem, the standard time-domain NRL TEMTADS 2x2 electromagnetic induction (EMI) instrument is updated. Namely, the NRL TEMTADS 2x2 system’s transmitter electronics is modified to increase transmitter (Tx) currents from 6 Amperes to 14 Amperes. The instrument has a Tx array with four coplanar square coils, together with four tri-axial receivers (Rx) placed at the center of each Tx. Each Rx cube contains three orthogonal coils and thus registers all three vector components of the impinging signals. The Tx coils, with transmitter currents of ~14 A, illuminate a buried target, and the target responses are collected with a 500 kHz sample rate after turn off of the excitation pulse. The system operates in both static (cued) and dynamic modes. For cued mode, the raw decay measurements are grouped into 121 logarithmically-spaced “gates” whose center times range from 25 μs to 24.35 ms with 5% widths. The sensor is placed on a cart which provides a sensor-to-ground offset of 20 cm or less. In this paper, studies for APG Calibration, Blind, and Small Munitions Grids are presented and analyzed. The areas are arranged in grids of test cells and the cell center positions are known. Each target position is flagged with a non-metallic pin flag using cm-level GPS. The sensor is positioned over each target in turn. With the system positioned over the target, each Tx is activated sequentially and during off the Tx current, all four Rx record data. The capabilities of this sensor platform is rigorously investigated for UXO classification at APG blind and small munitions grids.

The adaptive coherence estimator (ACE) estimates the squared cosine of the angle between a known target vector and a sample vector in a transformed coordinate space. The space is transformed according to an estimation of the background statistics, which directly effects the performance of the statistic as a target detector. In this paper, the ACE detection statistic is used to detect buried explosive hazards with data from a Wideband Electromagnetic Induction (WEMI) sensor. Target signatures are based on a dictionary defined using a Discrete Spectrum of Relaxation Frequencies (DSRF) model. Results are summarized as a receiver operator curve (ROC) and compared to other leading methods.

Many effective supervised discriminative dictionary learning methods have been developed in the literature. However, when training these algorithms, precise ground-truth of the training data is required to provide very accurate point-wise labels. Yet, in many applications, accurate labels are not always feasible. This is especially true in the case of buried object detection in which the size of the objects are not consistent. In this paper, a new multiple instance dictionary learning algorithm for detecting buried objects using a handheld WEMI sensor is detailed. The new algorithm, Task Driven Extended Functions of Multiple Instances, can overcome data that does not have very precise point-wise labels and still learn a highly discriminative dictionary. Results are presented and discussed on measured WEMI data.

Low-frequency electromagnetic induction (EMI) sensors are commonly used in subsurface detection applications because of their efficacy at detecting even small fragments of metal when they are buried near the surface. This efficacy can become a shortcoming when the detector is expected to locate specific classes of targets that are buried among metallic clutter. For these applications, broadband EMI sensors have shown considerable promise at being able to detect, classify and locate targets such as land mines, and discriminate between them and the clutter with low false-alarm rates. In such cases, where differentiating targets from clutter is a significant obstacle, detection strategies based on the discrete spectrum of relaxation frequencies (DSRF) have been shown to be highly effective. For such purposes, a dictionary of DSRF of targets of interest must be computed a priori. Several classes of targets such as sphere and rings have DSRF that can be derived analytically, however, in general, the DSRF must be computed numerically. Previously, numerical strategies have been presented for thin conducting shells and rotationaly symmetric targets. In this paper, we will present a strategy to compute the DSRF of arbitrary conducting targets using a null space free Jacobi Davidson iteration (NFJD).

Ultrawide band electromagnetic induction (EMI) instruments have been traditionally used to detect high electric conductivity discrete targets such as metal unexploded ordnance. The frequencies used for this EMI regime have typically been less than 100 kHz. To detect intermediate conductivity objects like carbon fiber, even less conductive saturated salts, and even voids embedded in conducting soils, higher frequencies up to the low megahertz range are required in order to capture characteristic responses. To predict EMI phenomena at frequencies up to 15 MHz, we first modeled the response of intermediate conductivity targets using a rigorous, first-principles approach, the Method of Auxiliary Sources. A newly fabricated benchtop high-frequency electromagnetic induction instrument produced EMI data at frequencies up to that same high limit. Modeled and measured characteristic relaxation signatures compare favorably and indicate new sensing possibilities in a variety of scenarios.

This work introduces two advances in wide-band electromagnetic induction (EMI) processing: a novel adaptive matched filter (AMF) and matched subspace detection methods. Both advances make use of recent work with a subspace SVD approach to separating the signal, soil, and noise subspaces of the frequency measurements The proposed AMF provides a direct approach to removing the EMI self-response while improving the signal to noise ratio of the data. Unlike previous EMI adaptive downtrack filters, this new filter will not erroneously optimize the EMI soil response instead of the EMI target response because these two responses are projected into separate frequency subspaces. The EMI detection methods in this work elaborate on how the signal and noise subspaces in the frequency measurements are ideal for creating the matched subspace detection (MSD) and constant false alarm rate matched subspace detection (CFAR) metrics developed by Scharf The CFAR detection metric has been shown to be the uniformly most powerful invariant detector.

Using both computer simulation and physical experiment, we demonstrate principal limitations of standard Time Domain Spectroscopy based on a broadband THz pulse for the detection and identification of substance placed inside the disordered structure. The interaction of a THz pulse with a disordered layered structure was simulated in order to show the influence of the disordered layers on the spectral characteristics of the transmitted and reflected signals. Spectral characteristics of these signals were analyzed in a direct comparison with the incident pulse spectrum. We showed that a disordered structure disturbs the reflected pulse spectrum dramatically. To avoid this, we used the integral correlation criteria in real experiment. Computer simulation results were confirmed by physical experiment. We provided the experiments with paper bag, and ordinary sheets of paper, and napkins.

This paper considers the efficacy of 3D scanning and printing technologies to produce duplicate keys. Duplication of keys, based on remote-sensed data represents a significant security threat, as it removes pathways to determining who illicitly gained access to a secured premises. Key to understanding the threat posed is the characterization of the easiness of gaining the required data for key production and an understanding of how well keys produced under this method work. The results of an experiment to characterize this are discussed and generalized to different key types. The effect of alternate sources of data on imaging requirements is considered.

Explosive hazards, above and below ground, are a serious threat to civilians and soldiers. In an attempt to mitigate these threats, different forms of explosive hazard detection (EHD) exist; e.g., multi-sensor hand-held platforms, downward looking and forward looking vehicle mounted platforms, etc. Robust detection of these threats resides in the processing and fusion of different data from multiple sensing modalities, e.g., radar, infrared, electromagnetic induction (EMI), etc. Herein, we focus on a new energy-based prescreener in hand-held ground penetrating radar (GPR). First, we Curvelet filter B-scan signal data using either Reverse-Reconstruction followed by Enhancement (RRE) or selectivity with respect to wedge information in the Curvelet transform. Next, we aggregate the result of a bank of matched filters and run a size contrast filter with Bhattacharyya distance. Alarms are then combined using weighted mean shift clustering. Results are demonstrated in the context of receiver operating characteristics (ROC) curve performance on data from a U. S. Army test site that contains multiple target and clutter types, burial depths and times of the day.

The challenge in detecting explosive hazards is that there are multiple types of targets buried at different depths in a highlycluttered environment. A wide array of target and clutter signatures exist, which makes detection algorithm design difficult. Such explosive hazards are typically deployed in past and present war zones and they pose a grave threat to the safety of civilians and soldiers alike. This paper focuses on a new image enhancement technique for hand-held ground penetrating radar (GPR). Advantages of the proposed technique is it runs in real-time and it does not require the radar to remain at a constant distance from the ground. Herein, we evaluate the performance of the proposed technique using data collected from a U.S. Army test site, which includes targets with varying amounts of metal content, placement depths, clutter and times of day. Receiver operating characteristic (ROC) curve-based results are presented for the detection of shallow, medium and deeply buried targets. Preliminary results are very encouraging and they demonstrate the usefulness of the proposed filtering technique.

regions with significant amount of metal debris. The challenge for the handheld GPR is to reduce the false alarm rate and limit the undesirable human operator effect. This paper proposes the use of log-Gabor features to improve the detection performance. In particular, we apply 36 log-Gabor filters to the B-scan of the GPR data in the time domain for the purpose to extract the edge behaviors of a prescreener alarm. The 36 log-Gabor filters cover the entire frequency plane with different bandwidths and orientations. The energy of each filter output forms an element of the feature vector and an SVM is trained to perform target vs non-target classification. Experimental results using the experimental hand held demonstrator data collected at a government site supports the increase in detection performance by using the log-Gabor features.

Ground Penetrating Radar (GPR) is one of the most popular subsurface sensing devices. It has a wide range of applications such as landmine detection, archeological investigations, road condition survey and so on. Hardware and software requirements of the GPR system are strongly dependent on type of applications. Principally, lower frequencies provide deeper penetration and low resolution, but higher frequencies are able to detect shallow objects with high resolution. As a fundamental design criterion, there is a trade-off between penetration depth and vertical resolution. In impulse radar, pulse duration (frequency related) is a key parameter because it affects the system detection performance. Specially, optimization of the pulse duration is a challenging problem for landmine detection because the GPR performance has been strongly affected from mine types, varying terrain and environmental conditions. In this work, two GPR systems with pulse durations of 650 ps and 870 ps are compared for evaluation of their detection performance. The pulses are tested with extensive data sets collected from different soil types by using surrogate mines and other objects. Receiver Operating Characteristic (ROC) curves of the system is also calculated. It seems that the 650 ps pulse duration gives better performance than the 870 ps pulse duration for the shallow landmine detection.

Detection of buried landmines and other explosive objects using ground penetrating radar (GPR) has been investigated for almost two decades and several classifiers have been developed. Most of these methods are based on the supervised learning paradigm where labeled target and clutter signatures are needed to train a classifier to discriminate between the two classes. Typically, large and diverse labeled training samples are needed to improve the performance of the classifier by overcoming noise and adding robustness and generalization to unseen examples. Unfortunately, even though unlabeled GPR data may be abundant, labeled data are often available in small quantities as the labeling process is tedious and can be ambiguous for most of the data. In this paper, we propose an algorithm for detecting landmines and buried objects that uses unlabeled data to help labeled data in the classification process. Our algorithm is graph-based and propagates the nodes labels to neighboring nodes according to their proximity in the feature space. For labeled data, we use a set of prototypes that are extracted from a small set of labeled training samples. For unlabeled data, we use a collection of signatures that are extracted from the vicinity of the alarm being tested. This choice is based on the assumption that many spatially close signatures are expected to have similar features and thus, unlabeled samples can create dense regions that link different regions of the labeled samples and propagate their labels to test samples. In other words, unlabeled samples are explored to create a context for each test alarm. To validate the proposed label propagation based classifier, we use it to detect buried explosive objects in GPR data collected by an experimental hand held demonstrator. We show that our approach is robust and computationally efficient to be used for both target discrimination and prescreening.

Buried explosive hazards (BEHs), including traditional landmines and homemade improvised explosives, have proven difficult to detect and defeat during and after conflicts around the world. Despite their various sizes, shapes and construction material, ground penetrating radar (GPR) is an excellent phenomenology for detecting BEHs due to its ability to sense localized differences in electromagnetic properties. Handheld GPR detectors are common equipment for detecting BEHs because of their flexibility (in part due to the human operator) and effectiveness in cluttered environments. With modern digital electronics and positioning systems, handheld GPR sensors can sense and map variation in electromagnetic properties while searching for BEHs. Additionally, large-scale computers have demonstrated an insatiable appetite for ingesting massive datasets and extracting meaningful relationships. This is no more evident than the maturation of deep learning artificial neural networks (ANNs) for image and speech recognition now commonplace in industry and academia. This confluence of sensing, computing and pattern recognition technologies offers great potential to develop automatic target recognition techniques to assist GPR operators searching for BEHs. In this work deep learning ANNs are used to detect BEHs and discriminate them from harmless clutter. We apply these techniques to a multi-antennae, handheld GPR with centimeter-accurate positioning system that was used to collect data over prepared lanes containing a wide range of BEHs. This work demonstrates that deep learning ANNs can automatically extract meaningful information from complex GPR signatures, complementing existing GPR anomaly detection and classification techniques.

This paper describes procedures and approaches our team took to demonstrate the capability of advanced electromagnetic induction (EMI) forward and inverse models to perform subsurface metallic objects picking and classification at live-UXO sites from dynamic data sets. Over the past seven years, blind classification tests at live-UXO sites have revealed two main challenges: 1) consistent selection of targets for cued interrogation, (e.g., for the recent SWPG2 study, two independent performers that processed the same MetalMapper dynamic data picked different targets for cued interrogation); and 2) positioning of the cued sensor close enough to the actual cued target to accurately perform classification (particularly when multiple targets or magnetic soils are present). To overcome these problems, in this paper we introduced an innovative and robust approach for subsurface metallic targets picking and classification from dynamic data sets. This approach first inverts for target locations and polarizabilities from each dynamic data point, and then clusters the inverted locations and defines each cluster as a target/source. Finally, the method uses the extracted polarizabilities for classifying UXO from non-UXO items. The studies are done for the 2x2 TEMTADS dynamic data set collected at Camp Hale, CO. The targets picking and classification results are illustrated and validated against ground truth.

An electromagnetic induction system, suitable for 2D imaging of metallic samples of different electrical conductivities, has been developed. The system is based on a parallel LCR circuit comprising a ferrite-cored coil (7.8 mm x 9.5 mm, L=680 μH at 1 KHz), a variable resistor and capacitor. The working principle of the system is based on eddy current induction inside a metallic sample when this is introduced into the AC magnetic field created by the coil. The inductance of the LCR circuit is modified due to the presence of the sample, to an extent that depends on its conductivity. Such modification is known to increase when the system is operated at its resonant frequency. Characterizing different metals based on their values of conductivity is therefore possible by utilizing a suitable system operated at resonance. Both imaging and material characterization were demonstrated by means of the proposed electromagnetic induction technique. Furthermore, the choice of using a system with an adjustable resonant frequency made it possible to select resonances that allow magnetic-field penetration through conductive screens. Investigations on the possibility of imaging concealed metals by penetrating such shields have been carried out. A penetration depth of δ~3 mm through aluminium (Al) was achieved. This allowed concealed metallic samples- having conductivities ranging from 0.54 to 59.77 MSm^-1 and hidden behind 1.5-mm-thick Al shields- to be imaged. Our results demonstrate that the presence of the concealed metallic objects can be revealed. The technique was thus shown to be a promising detection tool for security applications.

We describe our research programme on the use of atomic magnetometers to detect conductive objects via electromagnetic induction. The extreme sensitivity of atomic magnetometers at low frequencies, up to seven orders of magnitude higher than a coil-based system, permits deep penetration through different media and barriers, and in various operative environments. This eliminates the limitations usually associated with electromagnetic detection.

This paper investigates the enhancements to detection of buried unexploded ordinances achieved by combining ground penetrating radar (GPR) data with electromagnetic induction (EMI) data. Novel features from both the GPR and the EMI sensors are concatenated as a long feature vector, on which a non-parametric classifier is then trained. The classifier is a boosting classifier based on tree classifiers, which allows for disparate feature values. The fusion algorithm was applied to a government-provided dataset from an outdoor testing site, and significant performance enhancements were obtained relative to classifiers trained solely on the GPR or EMI data. It is shown that the performance enhancements come from a combination of improvements in detection and in clutter rejection.

Within the analysis of cases relating to the use of explosives for crimes, we have experienced a shift from using industrial explosives towards substances made in amateur and illegal way. Availability of industrial explosives is increasingly limited to a narrow sphere of subjects with a relevant permission. Thus, on the part of perpetrators, terrorists, ever greater attention is paid to illegal production of explosives that are easily made from readily available raw materials. Another alarming fact is the availability of information found on the internet. Procedures of preparation are often very simple and do not require even a deeper professional knowledge. Explosive characteristics are not actually accessible for many of these substances (detonation velocity, sensitivity, working capacity, brisance, physical and chemical stability, etc.). Therefore, a project is being implemented, which on grounds of assessment of individual information available in literature and on the internet, aiming at choosing individual areas of potentially abusable substances (e.g. mixtures of nitric acid (98%) with organic substances, mixtures nitromethane and tetranitromethane with organic substances, mixtures of chlorates and perchlorates of alkali metals with organic substances, chemically individual compounds of organic base type of perchloric acid, azides, fulminates, acetylides, picrates, styphnates of heavy metals, etc.). It is directed towards preparation of these explosives also in non-stoichiometric mixtures, conducting test explosives, determination of explosive characteristics (if they are unknown) and analysis of both primary phases and post-blast residues through available analytical techniques, such as gas and liquid chromatography with mass detection, FTIR, micro-Raman spectrometry, electron microscopy with microanalysis and Raman microspectrometry directly in SEM chamber for analysis at the level of individual microparticles. The received characteristics will be used to extend knowledge database for security forces.

Explosive hazards are a deadly threat in modern conflicts; hence, detecting them before they cause injury or death is of paramount importance. One method of buried explosive hazard discovery relies on data collected from ground penetrating radar (GPR) sensors. Threat detection with downward looking GPR is challenging due to large returns from non-target objects and clutter. This leads to a large number of false alarms (FAs), and since the responses of clutter and targets can form very similar signatures, classifier design is not trivial. One approach to combat these issues uses robust principal component analysis (RPCA) to enhance target signatures while suppressing clutter and background responses, though there are many versions of RPCA. This work applies some of these RPCA techniques to GPR sensor data and evaluates their merit using the peak signal-to-clutter ratio (SCR) of the RPCA-processed B-scans. Experimental results on government furnished data show that while some of the RPCA methods yield similar results, there are indeed some methods that outperform others. Furthermore, we show that the computation time required by the different RPCA methods varies widely, and the selection of tuning parameters in the RPCA algorithms has a major effect on the peak SCR.

Detection of concealed hazardous materials is a pressing need for the global defense community. To address this need, the development of reliable and readily-deployable sensing devices is a key area of research. A multitude of infrared sensing techniques are being studied which allow for reliable sensing of concealed threats. Continued development in this field is working to increase the selectivity of such infrared sensors, while at the same time reducing their complexity, size and cost. We have recently developed a biomimetic optical filter based approach, based on human color vision, that utilizes multiple, broadband, overlapping infrared (IR) filters to clearly discriminate between hazardous target chemicals and interferents with very similar mid-IR spectral signatures. This technique was extensively studied in order to select filters which provide optimum selectivity for specific chemical sets. Using this knowledge, we designed and assembled a gas-phase sensor which uses three broadband mid-IR filters to detect and discriminate between a target chemical, fuel oil, and various interferents with strongly overlapping IR absorption bands in the carbon – hydrogen stretch region of the IR absorption spectrum 2700 cm^-1 - 3300 cm^-1 (3.0 μm - 3.7 μm). We present an overview of the design and performance of this filter-based system and explore the ability of this system to detect and discriminate between strongly overlapping target and interferent chemicals. The detection results using the filter-based system are compared to numerical methods to demonstrate the operation of this methodology. We present the results of experiments with both target and interferent chemicals present with chemicals both in and out of the detection set, and discuss future field development and application of this approach.

The technique of high-power laser-induced plasma acceleration can be used to generate a variety of diverse effects including the emission of X-rays, electrons, neutrons, protons and radio-frequency radiation. A compact variable source of this nature could support a wide range of potential applications including single-sided through-barrier imaging, cargo and vehicle screening, infrastructure inspection, oncology and structural failure analysis. This paper presents a verified particle physics simulation which replicates recent results from experiments conducted at the Central Laser Facility at Rutherford Appleton Laboratory (RAL), Didcot, UK. The RAL experiment demonstrated the generation of backscattered X-rays from test objects via the bremsstrahlung of an incident electron beam, the electron beam itself being produced by Laser Wakefield Acceleration. A key initial objective of the computer simulation was to inform the experimental planning phase on the predicted magnitude of the backscattered X-rays likely from the test objects. This objective was achieved and the computer simulation was used to show the viability of the proposed concept (Laser-induced X-ray ‘RADAR’). At the more advanced stages of the experimental planning phase, the simulation was used to gain critical knowledge of where it would be technically feasible to locate key diagnostic equipment within the experiment. The experiment successfully demonstrated the concept of X-ray ‘RADAR’ imaging, achieved by using the accurate timing information of the backscattered X-rays relative to the ultra-short laser pulse used to generate the electron beam. By using fast response X-ray detectors it was possible to derive range information for the test objects being scanned. An X-ray radar ‘image’ (equivalent to a RADAR B-scan slice) was produced by combining individual X-ray temporal profiles collected at different points along a horizontal distance line scan. The same image formation process was used to generate images from the modelled data. The simulated images show good agreement with the experimental images both in terms of the temporal and spatial response of the backscattered X-rays. The computer model has also been used to simulate scanning over an area to generate a 3D image of the test objects scanned. Range gating was applied to the simulated 3D data to show how significant signal-to-noise ratio enhancements could be achieved to resulting 2D images when compared to conventional backscatter X-ray images. Further predictions have been made using the computer simulation including the energy distribution of the backscatter X-rays, as well as multi-path and scatter effects not measured in the experiment. Multi-path effects were shown to be the primary contributor to undesirable image artefacts observed in the simulated images. The computer simulation allowed the sources of these artefacts to be identified and highlighted the importance of mitigating these effects in the experiment. These predicted effects could be explored and verified through future experiments. Additionally the model has provided insight into potential performance limitations of the X-ray RADAR concept and informed on possible solutions. Further model developments will include simulating a more realistic electron beam energy distribution and incorporating representative detector characteristics.

Nuclear Quadrupole Resonance (NQR) is a highly selective spectroscopic method that can be used to detect and identify a number of chemicals of interest to the defense, national security, and law enforcement community. In the past, there have been several documented attempts to utilize NQR to detect nitrogen bearing explosives using induction sensors to detect the NQR RF signatures. We present here our work on the NQR detection of explosive simulants using optically pumped RF atomic magnetometers. RF atomic magnetometers can provide an order of magnitude (or more) improvement in sensitivity versus induction sensors and can enable mitigation of RF interference, which has classically has been a problem for conventional NQR using induction sensors. We present the theory of operation of optically pumped RF atomic magnetometers along with the result of laboratory work on the detection of explosive simulant material. An outline of ongoing work will also be presented along with a path for a fieldable detection system.

Nuclear Quadrupole Resonance (NQR) has been demonstrated for the detection of 14-N in explosive compounds. Application of a material specific radio-frequency (RF) pulse excites a response typically detected with a wire- wound antenna. NQR is non-contact and material specific, however fields produced by NQR are typically very weak, making demonstration of practical utility challenging. For certain materials, the NQR signal can be increased by transferring polarization from hydrogen nuclei to nitrogen nuclei using external magnetic fields. This polarization enhancement (PE) can enhance the NQR signal by an order of magnitude or more. Atomic magnetometers (AM) have been shown to improve detection sensitivity beyond a conventional antenna by a similar amount. AM sensors are immune to piezo-electric effects that hamper conventional NQR, and can be combined to form a gradiometer for effective RF noise cancellation. In principle, combining polarization enhancement with atomic magnetometer detection should yield improvement in signal-to-noise ratio that is the product of the two methods, 100-fold or more over conventional NQR. However both methods are even more exotic than traditional NQR, and have never been combined due to challenges in operating a large magnetic field and ultra-sensitive magnetic field sensor in proximity. Here we present NQR with and without PE with an atomic magnetometer, demonstrating signal enhancement greater than 20-fold for ammonium nitrate. We also demonstrate PE for PETN using a traditional coil for detection with an enhancement factor of 10. Experimental methods and future applications are discussed.

A mass-spectrometric study of photo processes initiated by ultraviolet (UV) laser radiation in explosives adsorbed on metal and dielectric substrates has been performed. A calibrated quadrupole mass spectrometer was used to determine a value of activation energy of desorption and a quantity of explosives desorbed by laser radiation. A special vacuumoptical module was elaborated and integrated into a vacuum mass-spectrometric system to focus the laser beam on a sample. It has been shown that the action of nanosecond laser radiation set at q= 10⁷ - 10⁸ W/cm², λ=266 nm on adsorbed layers of molecules of trinitrotoluene (TNT ) and pentaerytritoltetranitrate (PETN) leads not only to an effective desorption, but also to the non-equilibrium dissociation of molecules with the formation of nitrogen oxide NO. The cyclotrimethylenetrinitramine (RDX) dissociation products are observed only at high laser intensities (q> 109 W/cm²) thus indicating the thermal nature of dissociation, whereas desorption of RDX is observed even at q> 10⁷ W/cm² from all substrates. Desorption is not observed for cyclotetramethylenetetranitramine (HMX) under single pulse action: the dissociation products NO and NO₂ are registered only, whereas irradiation at 10Hz is quite effective for HMX desorption. The results clearly demonstrate a high efficiency of nanosecond laser radiation with λ = 266 nm, q ~ 10⁷ - 10⁸ W/cm², E_pulse= 1mJ for desorption of molecules of explosives from various surfaces.

Imaging Raman spectroscopy based on tunable filters is an established technique for detecting single explosives particles at stand-off distances. However, large light losses are inherent in the design due to sequential imaging at different wavelengths, leading to effective transmission often well below 1 %. The use of digital micromirror devices (DMD) and compressive sensing (CS) in imaging Raman explosives trace detection can improve light throughput and add significant flexibility compared to existing systems. DMDs are based on mature microelectronics technology, and are compact, scalable, and can be customized for specific tasks, including new functions not available with current technologies. This paper has been focusing on investigating how a DMD can be used when applying CS-based imaging Raman spectroscopy on stand-off explosives trace detection, and evaluating the performance in terms of light throughput, image reconstruction ability and potential detection limits. This type of setup also gives the possibility to combine imaging Raman with non-spatially resolved fluorescence suppression techniques, such as Kerr gating. The system used consists of a 2nd harmonics Nd:YAG laser for sample excitation, collection optics, DMD, CMOScamera and a spectrometer with ICCD camera for signal gating and detection. Initial results for compressive sensing imaging Raman shows a stable reconstruction procedure even at low signals and in presence of interfering background signal. It is also shown to give increased effective light transmission without sacrificing molecular specificity or area coverage compared to filter based imaging Raman. At the same time it adds flexibility so the setup can be customized for new functionality.

Three-dimensional point clouds generated by LIDAR offer the potential to build a more complete understanding of the environment in front of a moving vehicle. In particular, LIDAR data facilitates the development of a non-parametric ground plane model that can filter target predictions from other sensors into above-ground and below-ground sets. This allows for improved detection performance when, for example, a system designed to locate above-ground targets considers only the set of above-ground predictions. In this paper, we apply LIDAR-based ground plane filtering to a forward looking ground penetrating radar (FLGPR) sensor system and a side looking synthetic aperture acoustic (SAA) sensor system designed to detect explosive hazards along the side of a road. Additionally, we consider the value of the visual magnitude of the LIDAR return as a feature for identifying anomalies. The predictions from these sensors are evaluated independently with and without ground plane filtering and then fused to produce a combined prediction confidence. Sensor fusion is accomplished by interpolating the confidence scores of each sensor along the ground plane model to create a combined confidence vector at specified points in the environment. The methods are tested along an unpaved desert road at an arid U.S. Army test site.

The remarkable popularity of smartphones over the past decade has led to a technological race for dominance in market share. This has resulted in a flood of new processors and sensors that are inexpensive, low power and high performance. These sensors include accelerometers, gyroscope, barometers and most importantly cameras. This sensor suite, coupled with multicore processors, allows a new community of researchers to build small, high performance platforms for low cost. This paper describes a system using off-the-shelf components to perform position tracking as well as environment modeling. The system relies on tracking using stereo vision and inertial navigation to determine movement of the system as well as create a model of the environment sensed by the system.

The capability of detecting an unpaved road in arid environments can greatly enhance an explosive hazard detection system. One approach is to segment out the off-road area and the area above the horizon, which is considered to be irrelevant for the task in hand. Segmenting out irrelevant areas, such as the region above the horizon, allows the explosive hazard detection system to process a smaller region in a scene, enabling a more computationally complex approach. In this paper, we propose a novel approach for speeding up the detection algorithms based on random projection and random selection. Both methods have a low computational cost and reduce the dimensionality of the data while approximately preserving, with a certain probability, the pair-wise point distances. Dimensionality reduction allows any classifier employed in our proposed algorithm to consume fewer computational resources. Furthermore, by applying the random projections directly to image intensity patches, there is no feature extraction needed. The data used in our proposed algorithms are obtained from sensors on board a U.S. Army countermine vehicle. We tested our proposed algorithms on data obtained from several runs on an arid climate road. In our experiments we compare our algorithms based on random projection and random selection to Principal Component Analysis (PCA), a popular dimensionality reduction method.

Underground structures can affect their surrounding landscapes in different ways such as soil moisture content, soil composition and vegetation vigor. Vegetation vigor is often observed on the ground as a crop mark; a phenomenon which can be used as a proxy to denote the presence of underground and not visible structures. This paper presents the results obtained from field spectroradiometric campaigns at ‘buried’ underground structures in Cyprus. A SVC-1024 field spectroradiometer was used and in-band reflectances were determined for medium and high resolution satellite sensors, including Landsat. A number of vegetation indices such as NDVI were obtained while a ‘smart index’ was developed. The aim of the 'smart index' is to detect underground military structures by using existing vegetation indices or other in-band algorithms. In this study, test areas were identified, analyzed and modeled. The areas were analyzed and tested in different scenarios, including: (a) the ‘natural state’ of the underground structure (b) the different type of crop over the underground structure and imported soil (c) the different types of non-natural material over the underground structure. A reference target in the nearby area was selected as a baseline. Controllable meteorological and environmental parameters were acquired and monitored.

Modern improvised explosive device (IED) and mine detection sensors using microwave technology are based on ground penetrating radar operated by a ground vehicle. Vehicle size, road conditions, and obstacles along the troop marching direction limit operation of such sensors. This paper presents a new conceptual design using a rotary unmanned aerial vehicle (UAV) to carry subsurface imaging radar for roadside IED detection. We have built a UAV flight simulator with the subsurface imaging radar running in a laboratory environment and tested it with non-metallic and metallic IED-like targets. From the initial lab results, we can detect the IED-like target 10-cm below road surface while carried by a UAV platform. One of the challenges is to design the radar and antenna system for a very small payload (less than 3 lb). The motion compensation algorithm is also critical to the imaging quality. In this paper, we also demonstrated the algorithm simulation and experimental imaging results with different IED target materials, sizes, and clutters.

A new generation of ultra-high sensitivity magnetic sensors based on innovative micro-electromechanical systems (MEMS) are being developed and incorporated into military systems. Specifically, we are currently working to fully integrate the latest generation of MicroFabricated Atomic Magnetometers (MFAMs) developed by Geometrics on defense mobility systems such as unmanned systems, military vehicles and handheld units. Recent reductions in size, weight, and power of these sensors has enabled new deployment opportunities for improved sensitivity to targets of interest, but has also introduced new challenges associated with noise mitigation, mission configuration planning, and data processing. Our work is focused on overcoming the practical aspects of integrating these sensors with various military platforms. Implications associated with utilizing these combined sensor systems in working environments are addressed in order to optimize signal-to-noise ratios, detection probabilities, and false alarm mitigation. Specifically, we present collaborative work that bridges the gap between commercial specialists and operation platform integration organizations including magnetic signature characterization and mitigation as well as the development of simulation tools that consider a wide array of sensor, environmental, platform, and mission-level parameters. We discuss unique deployment concepts for explosive hazard target geolocation, and data processing. Applications include configurations for undersea and underground threat detection - particularly those associated with stationary or mobile explosives and compact metallic targets such as munitions, subsea threats, and other hazardous objects. We show the potential of current and future features of miniaturized magnetic sensors including very high magnetic field sensitivities, bandwidth selectivity, and array processing.

Ground-penetrating radar (GPR) technology is an effective method of detecting buried explosive threats. The system uses a binary classifier to distinguish “targets”, or buried threats, from “nontargets” arising from system prescreener false alarms; this classifier is trained on a dataset of previously-observed buried threat types. However, the threat environment is not static, and new threat types that appear must be effectively detected even if they are not highly similar to every previously-observed type. Gathering a new dataset that includes a new threat type is expensive and time-consuming; minimizing the amount of new data required to effectively detect the new type is therefore valuable. This research aims to reduce the number of training examples needed to effectively detect new types using transfer learning, which leverages previous learning tasks to accelerate and improve new ones. Further, new types have attribute data, such as composition, components, construction, and size, which can be observed without GPR and typically are not explicitly included in the learning process. Since attribute tags for buried threats determine many aspects of their GPR representation, a new threat type’s attributes can be highly relevant to the transfer-learning process. In this work, attribute data is used to drive transfer learning, both by using attributes to select relevant dataset examples for classifier fusion, and by extending a relevance vector machine (RVM) model to perform intelligent attribute clustering and selection. Classification performance results for both the attribute-only case and the low-data case are presented, using a dataset containing a variety of threat types.

Ground penetrating radar (GPR) is a popular remote sensing modality for buried threat detection. Many algorithms have been developed to detect buried threats using GPR data. One on-going challenge with GPR is the detection of very deeply buried targets. In this work a detection approach is proposed that improves the detection of very deeply buried targets, and interestingly, shallow targets as well. First, it is shown that the signal of a target (the target “signature”) is well localized in time, and well correlated with the target’s burial depth. This motivates the proposed approach, where GPR data is split into two disjoint subsets: an early and late portion corresponding to the time at which shallow and deep target signatures appear, respectively. Experiments are conducted on real GPR data using the previously published histogram of oriented gradients (HOG) prescreener: a fast supervised processing method operated on HOG features. The results show substantial improvements in detection of very deeply buried targets (4.1% to 17.2%) and in overall detection performance (81.1% to 83.9%). Further, it is shown that the performance of the proposed approach is relatively insensitive to the time at which the data is split. These results suggest that other detection methods may benefit from depth-based processing as well.

This paper investigates the use of the apex-shifted hyperbolic Radon transform to improve detection of buried unexploded ordinances with ground penetrating radar (GPR). The forward transform, motivated by physical signatures generated by targets, is defined and implemented. The adjoint of the transform is derived and implemented as well. The transform and its adjoint are used to filter out responses that do not exhibit the hyperbolic structure characteristic of GPR target responses. The effectiveness of filtering off clutter via this hyperbolic Radon transform filtering procedure is demonstrated qualitatively on several examples of GPR B-scan imagery from a government-provided dataset collected at an outdoor testing site. Furthermore, a quantitative assessment of the utility within a detection algorithm is given in terms of improved ROC curve performance on the same dataset.

We present a new method, based on the Fisher Vector (FV), for detecting buried explosive objects using ground- penetrating radar (GPR) data. First, low-level dense SIFT features are extracted from a grid covering a region of interest (ROIs). ROIs are identified as regions with high-energy along the (down-track, depth) dimensions of the 3-D GPR cube, or with high-energy along the (cross-track, depth) dimensions. Next, we model the training data (in the SIFT feature space) by a mixture of Gaussian components. Then, we construct FV descriptors based on the Fisher Kernel. The Fisher Kernel characterizes low-level features from an ROI by their deviation from a generative model. The deviation is the gradient of the ROI log-likelihood with respect to the generative model parameters. The vectorial representation of all the deviations is called the Fisher Vector. FV is a generalization of the standard Bag of Words (BoW) method, which provides a framework to map a set of local descriptors to a global feature vector. It is more efficient to compute than the BoW since it relies on a significantly smaller codebook. In addition, mapping a GPR signature into one global feature vector using this technique makes it more efficient to classify using simple and fast linear classifiers such as Support Vector Machines. The proposed approach is applied to detect buried explosive objects using GPR data. The selected data were accumulated across multiple dates and multiple test sites by a vehicle mounted mine detector (VMMD) using GPR sensor. This data consist of a diverse set of conventional landmines and other buried explosive objects consisting of varying shapes, metal content, and burial depths. The performance of the proposed approach is analyzed using receiver operating characteristics (ROC) and is compared to other state-of-the-art feature representation methods.

In this paper, a decision level fusion using multiple pre-screener algorithms is proposed for the detection of buried landmines from Ground Penetrating Radar (GPR) data. The Kernel Least Mean Square (KLMS) and the Blob Filter pre-screeners are fused together to work in real time with less false alarms and higher true detection rates. The effect of the kernel variance is investigated for the KLMS algorithm. Also, the results of the KLMS and KLMS+Blob filter algorithms are compared to the LMS method in terms of processing time and false alarm rates. Proposed algorithm is tested on both simulated data and real data collected at the field of IPA Defence at METU, Ankara, Turkey.

There is an increasing demand for noninvasive real-time detection and classification of buried objects in various civil and military applications. The problem of detection and annihilation of landmines is particularly important due to strong safety concerns. The requirement for a fast real-time decision process is as important as the requirements for high detection rates and low false alarm rates. In this paper, we introduce and demonstrate a computationally simple, timeefficient, energy-based preprocessing approach that can be used in ground penetrating radar (GPR) applications to eliminate reflections from the air-ground boundary and to locate the buried objects, simultaneously, at one easy step. The instantaneous power signals, the total energy values and the cumulative energy curves are extracted from the A-scan GPR data. The cumulative energy curves, in particular, are shown to be useful to detect the presence and location of buried objects in a fast and simple way while preserving the spectral content of the original A-scan data for further steps of physics-based target classification. The proposed method is demonstrated using the GPR data collected at the facilities of IPA Defense, Ankara at outdoor test lanes. Cylindrically shaped plastic containers were buried in fine-medium sand to simulate buried landmines. These plastic containers were half-filled by ammonium nitrate including metal pins. Results of this pilot study are demonstrated to be highly promising to motivate further research for the use of energy-based preprocessing features in landmine detection problem.

Automatic landmine detection system using ground penetrating radar has been widely researched. For the automatic mine detection system, system speed is an important factor. Many techniques for mine detection have been developed based on statistical background. Among them, a detection technique employing the Principal Component Analysis(PCA) has been used for clutter reduction and anomaly detection. However, the PCA technique can retard the entire process, because of large basis dimension and a numerous number of inner product operations. In order to overcome this problem, we propose a fast anomaly detection system using 2D DCT and PCA. Our experiments use a set of data obtained from a test site where the anti-tank and anti- personnel mines are buried. We evaluate the proposed system in terms of the ROC curve. The result shows that the proposed system performs much better than the conventional PCA systems from the viewpoint of speed and false alarm rate.

A serious threat to civilians and soldiers is buried and above ground explosive hazards. The automatic detection of such threats is highly desired. Many methods exist for explosive hazard detection, e.g., hand-held based sensors, downward and forward looking vehicle mounted platforms, etc. In addition, multiple sensors are used to tackle this extreme problem, such as radar and infrared (IR) imagery. In this article, we explore the utility of feature and decision level fusion of learned features for forward looking explosive hazard detection in IR imagery. Specifically, we investigate different ways to fuse learned iECO features pre and post multiple kernel (MK) support vector machine (SVM) based classification. Three MK strategies are explored; fixed rule, heuristics and optimization-based. Performance is assessed in the context of receiver operating characteristic (ROC) curves on data from a U.S. Army test site that contains multiple target and clutter types, burial depths and times of day. Specifically, the results reveal two interesting things. First, the different MK strategies appear to indicate that the different iECO individuals are all more-or-less important and there is not a dominant feature. This is reinforcing as our hypothesis was that iECO provides different ways to approach target detection. Last, we observe that while optimization-based MK is mathematically appealing, i.e., it connects the learning of the fusion to the underlying classification problem we are trying to solve, it appears to be highly susceptible to over fitting and simpler, e.g., fixed rule and heuristics approaches help us realize more generalizable iECO solutions.

The forward-looking ground penetrating radar (FLGPR) is a remote sensing modality that has recently been investigated for buried threat detection. The FLGPR considered in this work uses stepped frequency sensing followed by filtered backprojection to create images of the ground, where each image pixel corresponds to the radar energy reflected from the subsurface at that location. Typical target detection processing begins with a prescreening operation where a small subset of spatial locations are chosen to consider for further processing. Image statistics, or features, are then extracted around each selected location and used for training a machine learning classification algorithm. A variety of features have been proposed in the literature for use in classification. Thus far, however, predominantly hand-crafted or manually designed features from the computer vision literature have been employed (e.g., HOG, Gabor filtering, etc.). Recently, it has been shown that image features learned directly from data can obtain state-of-the-art performance on a variety of problems. In this work we employ a feature learning scheme using k-means and a bag-of-visual-words model to learn effective features for target and non-target discrimination in FLGPR data. Experiments are conducted using several lanes of FLGPR data and learned features are compared with several previously proposed static features. The results suggest that learned features perform comparably, or better, than existing static features. Similar to other feature learning results, the features consist of edges or texture primitives, revealing which structures in the data are most useful for discrimination.

Forward looking ground penetrating radar (FLGPR) is an alternative buried threat sensing technology designed to offer additional standoff compared to downward looking GPR systems. Due to additional flexibility in antenna configurations, FLGPR systems can accommodate multiple sensor modalities on the same platform that can provide complimentary information. The different sensor modalities present challenges in both developing informative feature extraction methods, and fusing sensor information in order to obtain the best discrimination performance. This work uses convolutional neural networks in order to jointly learn features across two sensor modalities and fuse the information in order to distinguish between target and non-target regions. This joint optimization is possible by modifying the traditional image-based convolutional neural network configuration to extract data from multiple sources. The filters generated by this process create a learned feature extraction method that is optimized to provide the best discrimination performance when fused. This paper presents the results of applying convolutional neural networks and compares these results to the use of fusion performed with a linear classifier. This paper also compares performance between convolutional neural networks architectures to show the benefit of fusing the sensor information in different ways.

Many remote sensing modalities have been developed for buried target detection (BTD), each one offering relative advantages over the others. There has been interest in combining several modalities into a single BTD system that benefits from the advantages of each constituent sensor. Recently an approach was developed, called multi-state management (MSM), that aims to achieve this goal by separating BTD system operation into discrete states, each with different sensor activity and system velocity. Additionally, a modeling approach, called Q-MSM, was developed to quickly analyze multi-modality BTD systems operating with MSM. This work extends previous work by demonstrating how Q-MSM modeling can be used to design BTD systems operating with MSM, and to guide research to yield the most performance benefits. In this work an MSM system is considered that combines a forward-looking infrared (FLIR) camera and a ground penetrating radar (GPR). Experiments are conducted using a dataset of real, field-collected, data which demonstrates how the Q-MSM model can be used to evaluate performance benefits of altering, or improving via research investment, various characteristics of the GPR and FLIR systems. Q-MSM permits fast analysis that can determine where system improvements will have the greatest impact, and can therefore help guide BTD research.

Forward looking ground penetrating radar (FLGPR) has the benefit of detecting objects at a significant standoff distance. The FLGPR signal is radiated over a large surface area and the radar signal return is often weak. Improving detection, especially for buried in road targets, while maintaining an acceptable false alarm rate remains to be a challenging task. Various kinds of features have been developed over the years to increase the FLGPR detection performance. This paper focuses on investigating the use of as many features as possible for detecting buried targets and uses the sequential feature selection technique to automatically choose the features that contribute most for improving performance. Experimental results using data collected at a government test site are presented.

The operational constraints associated with a forward-looking ground-penetrating radar (GPR) limit the ability of the radar to resolve targets in the dimension orthogonal to the ground. As such, detection performance of buried targets is greatly inhibited by the relatively large response due to surface clutter. The response of buried targets differs from surface targets due to the interaction at the boundary and propagation through the ground media. The electromagnetic properties of the media, interrogation frequency, depth of buried target, and location of the target with respect to the the sensing platform all contribute to the shape, position, and magnitude of the point spread function (PSF). The standard FLGPR scenario produces a wide-band data set collected over a fixed set of observation points. By observing the shape, position, and amplitude behavior of the PSF as a function of frequency and sensor position (time), energy resulting from surface clutter can be separated from energy resulting from buried targets. There are many possible ways beyond conventional image resolution that might be exploited to improve distinction between buried targets and surface clutter. This investigation exploits the frequency dependence of buried targets compared to surface targets using a set of sub-banded images.

Buried explosives hazards are one of the many deadly threats facing our Soldiers, thus the U.S. Army is interested in the detection and neutralization of these hazards. One method of buried target detection uses forward-looking ground-penetrating radar (FLGPR), and it has grown in popularity due to its ability to detect buried targets at a standoff distance. FLGPR approaches often use machine learning techniques to improve the accuracy of detection. We investigate an approach to explosive hazard detection that exploits multi-instance features to discriminate between hazardous and non-hazardous returns in FLGPR data. One challenge this problem presents is a high number of clutter and non-target objects relative to the number of targets present. Our approach learns a bag of words model of the multi-instance signatures of potential targets and confuser objects in order to classify alarms as either targets or false alarms. We demonstrate our method on test data collected at a U.S. Army test site.

The U.S. Army RDECOM CERDEC Night Vision and Electronic Sensors Directorate (NVESD) is executing a program to assess the performance of a variety of sensor modalities for standoff detection of roadside explosive hazards. The program objective is to identify an optimal sensor or combination of fused sensors to incorporate with autonomous detection algorithms into a system of systems for use in future route clearance operations. This paper provides an overview of the program, including a description of the sensors under consideration, sensor test events, and ongoing data analysis.

Commercial sensor technology has the potential to bring cost-effective sensors to a number of U.S. Army applications. By using sensors built for a widespread of commercial application, such as the automotive market, the Army can decrease costs of future systems while increasing overall capabilities. Additional sensors operating in alternate and orthogonal modalities can also be leveraged to gain a broader spectrum measurement of the environment. Leveraging multiple phenomenologies can reduce false alarms and make detection algorithms more robust to varied concealment materials. In this paper, this approach is applied to the detection of roadside hazards partially concealed by light-to-medium vegetation. This paper will present advances in detection algorithms using a ground vehicle-based commercial LADAR system. The benefits of augmenting a LADAR with millimeter-wave automotive radar and results from relevant data sets are also discussed.

In this paper, we develop an approach to detect explosive hazards designed to attack vehicles from the side of a road, using a side looking synthetic aperture acoustic (SAA) sensor. This is done by first processing the raw data using a back-projection algorithm to form images. Next, an RX prescreener creates a list of possible targets, each with a designated confidence. Initial experiments are performed on libraries of the highest confidence hits for both target and false alarm classes generated by the prescreener. Image chips are extracted using pixel locations derived from the target’s easting and northing. Several feature types are calculated from each image chip, including: histogram of oriented gradients (HOG), and generalized column projection features where the column aggregator takes the form of the minimum, maximum, mean, median, mode, standard deviation, variance, and the one-dimensional fast Fourier transform (FFT). A support vector machine (SVM) classifier is then utilized to evaluate feature type performance during training and testing in order to determine whether the two classes are separable. This will be used to build an online detection system for road-side explosive hazards.

Explosive hazards in current and former conflict zones are a threat to both military and civilian personnel. As a result, much effort has been dedicated to identifying automated algorithms and systems to detect these threats. However, robust detection is complicated due to factors like the varied composition and anatomy of such hazards. In order to solve this challenge, a number of platforms (vehicle-based, handheld, etc.) and sensors (infrared, ground penetrating radar, acoustics, etc.) are being explored. In this article, we investigate the detection of side attack explosive ballistics via a vehicle-mounted acoustic sensor. In particular, we explore three acoustic features, one in the time domain and two on synthetic aperture acoustic (SAA) beamformed imagery. The idea is to exploit the varying acoustic frequency profile of a target due to its unique geometry and material composition with respect to different viewing angles. The first two features build their angle specific frequency information using a highly constrained subset of the signal data and the last feature builds its frequency profile using all available signal data for a given region of interest (centered on the candidate target location). Performance is assessed in the context of receiver operating characteristic (ROC) curves on cross-validation experiments for data collected at a U.S. Army test site on different days with multiple target types and clutter. Our preliminary results are encouraging and indicate that the top performing feature is the unrolled two dimensional discrete Fourier transform (DFT) of SAA beamformed imagery.

A main problem of effective landmine and UXO decontamination is efficient and reliable detection and localization of suspicious objects in reasonable time. This requirement demands for fast sensors investigating large areas with sufficient spatial resolution and sensitivity. Ground penetrating radar (GPR) is a suitable tool and is considered as a complementing sensor since nearly two decades. However, most GPRs operate in very close distance to ground in a rather punctual method of operation. In contrast, synthetic aperture radar (SAR) is a technique allowing fast and laminar stand-off investigation of an area. TIRAMI-SAR is imaging radar at lower microwaves for fast close-in detection of buried and unburied objects on a larger area. This allows efficient confirmation of a threat by investigating such regions of detection by other sensors. For proper object detection sufficient spatial resolution is required. Hence the SAR principle is applied. SAR for landmine/UXO detection can be applied by side-looking radar moved on safe ground along the area of interest, being typically the un-safe ground. Additionally, reliable detection of buried and unburied objects requires sufficient suppression of background clutter. For that purpose TIRAMI-SAR is using several antennas in multi-static configuration and wave polarization together with advanced SAR processing. The advantages and necessity of a multi-static antenna configuration for this kind of GPR approach is illustrated in the paper.

Coherent Change Detection (CCD) is a process of highlighting an area of activity in scenes (seafloor) under survey and generated from pairs of synthetic aperture sonar (SAS) images of approximately the same location observed at two different time instances. The problem of CCD and subsequent anomaly feature extraction/detection is complicated due to several factors such as the presence of random speckle pattern in the images, changing environmental conditions, and platform instabilities. These complications make the detection of weak target activities even more difficult. Typically, the degree of similarity between two images measured at each pixel locations is the coherence between the complex pixel values in the two images. Higher coherence indicates little change in the scene represented by the pixel and lower coherence indicates change activity in the scene. Such coherence estimation scheme based on the pixel intensity correlation is an ad-hoc procedure where the effectiveness of the change detection is determined by the choice of threshold which can lead to high false alarm rates. In this paper, we propose a novel approach for anomalous change pattern detection using the statistical normalized coherence and multi-pass coherent processing. This method may be used to mitigate shadows by reducing the false alarms resulting in the coherent map due to speckles and shadows. Test results of the proposed methods on a data set of SAS images will be presented, illustrating the effectiveness of the normalized coherence in terms statistics from multi-pass survey of the same scene.

The Art Gallery Problem (AGP) is the name given to a constrained optimization problem meant to determine the maximum amount of sensor coverage while utilizing the minimum number of resources. The AGP is significant because a common issue among surveillance and interdiction systems is obtaining an understanding of the optimal position of sensors and weapons in advance of enemy combatant maneuvers. The implication that an optimal position for a sensor to observe an event or for a weapon to engage a target autonomously is usually very clear after the target has passed, but for autonomous systems the solution must at least be conjectured in advance for deployment purposes. This abstract applies the AGP as a means to solve where best to place underwater sensor nodes such that the amount of information acquired about a covered area is maximized while the number of resources used to gain that information is minimized. By phrasing the ISR/interdiction problem this way, the issue is addressed as an instance of the AGP. The AGP is a member of a set of computational problems designated as nondeterministic polynomial-time (NP)-hard. As a member of this set, the AGP shares its members' defining feature, namely that no one has proven that there exists a deterministic algorithm providing a computationally-tractable solution to the AGP within a finite amount of time. At best an algorithm meant to solve the AGP can asymptotically approach perfect coverage with minimal resource usage but providing perfect coverage would either break the minimal resource usage constraint or require an exponentially-growing amount of time. No perfectly-optimal solution yet exists to the AGP, however, approximately optimal solutions to the AGP can approach complete area or barrier coverage while simultaneously minimizing the number of sensors and weapons utilized. A minimal number of underwater sensor nodes deployed can greatly increase the Mean Time Between Operational Failure (MTBOF) and logistical footprint. The resulting coverage optimizes the likelihood of encounter given an arbitrary sensor profile and threat from a free field statistical model approach. The free field statistical model is particularly applicable to worst case scenario modeling in open ocean operational profiles where targets to do not follow a particular pattern in any of the modeled dimensions. We present an algorithmic testbed which shows how to achieve approximately optimal solutions to the AGP for a network of underwater sensor nodes with or without effector systems for engagement while operating under changing environmental circumstances. The means by which we accomplish this goal are three-fold: 1) Develop a 3D model for the sonar signal propagating through the underwater environment 2) Add rigorous physics-based modeling of environmental events which can affect sensor information acquisition 3) Provide innovative solutions to the AGP which account for the environmental circumstances affecting sensor performance.

This paper addresses selected computational aspects of collaborative search planning when multiple search agents seek to find hidden objects (i.e. mines) in operating environments where the detection process is prone to false alarms. A Receiver Operator Characteristic (ROC) analysis is applied to construct a Bayesian cost objective function that weighs and combines missed detection and false alarm probabilities. It is shown that for fixed ROC operating points and a validation criterion consisting of a prerequisite number of detection outcomes, an interval exists in the number of conducted search passes over which the risk objective function is supermodular. We show that this property is not retained beyond validation criterion boundaries. We investigate the use of greedy algorithms for distributing search effort and, in particular, examine the double greedy algorithm for its applicability under conditions of varying criteria. Numerical results are provided to demonstrate the effectiveness of the approach.

Littoral regions typically present to passive sensors as a high noise acoustic environment, particularly with respect to port and harbor regions where tidal variation, often characterized as pink, mixes with reverberation from on-shore business and commercial shipping, often characterized as white. Some fish in these regions, in particular epiphenalius Guttatus or more commonly the red hind grouper, emit relatively narrowband tones in low frequencies to communicate with other fish in such regions. The impact of anthropogenic noise sources on the red Hind and other fish is a topical area of interest for wildlife fisheries, private sportsmen and military offices that is not considered here; the fact that fish species continue to populate and communicate in these regions in the presence of high noise content lends some study to the signal content and modeling of a potential biologically inspired receiver structure.

The important part to any multi-input and multi-output (MIMO) system is the set of signals. These signals must be designed so that they can be simultaneously transmitted and received with minimal signal-to-signal interference and low spatial ambiguity. In addition, the signals must adhere to strict time and bandwidth constraints. To achieve this, we use the prolate spheroidal wave functions (PSWFs) as the basis functions for the approximate signals. Then, we define a cost function on the oblique manifold and use known Riemannian steepest descent to find the coefficients of the signals so that the output signals satisfy the MIMO design criteria.