This PDF file contains the front matter associated with SPIE Proceedings Volume 8709, including the Title Page, Copyright information, Table of Contents, and the Conference Committee listing.

The Portable Decoupled Electromagnetic Induction Sensor (Pedemis) is a new instrument designed to provide diverse, high-quality data for detection and discrimination of unexploded ordnance in rocky, treed, or otherwise forbidding terrain. It consists of a square array of nine transmitters and a similar arrangement of receivers that measure all three vector components of the time-dependent magnetic field at nine different locations. The receiver assembly can be fixed to the transmitters or detached from them for enhanced flexibility and convenience. The latter mode requires a positioning system that finds the location of the receivers with respect to the transmitters at any time without hampering portability or requiring communication with outside agents (which may be precluded by field conditions). The current system examines the primary field during the transmitters’ on-time phase and optimizes to find the location at which it is most likely to obtain the combination of measured values. We have developed an algorithm that computes mutual inductances analytically and exploits their geometric information to predict location. The method does full justice to Faraday’s Law from the start and incorporates the fine structure of both transmitters and receivers; it is exact and involves only elementary functions, making it unnecessary to set up and monitor approximations and guaranteeing robustness and stability everywhere; it uses a fraction of the memory and is orders-of-magnitude faster than methods based on numerical quadrature. We have tested the algorithm on the current Pedemis prototype and have obtained encouraging results which we summarize in this paper.

Pedemis (PortablE Decoupled Electromagnetic Induction Sensor) is a time-domain man-portable electromagnetic induction (EMI) instrument with the intended purpose of improving the detection and classification of UneXploded Ordnance (UXO). Pedemis sports nine coplanar transmitters (the Tx assembly) and nine triaxial receivers held in a fixed geometry with respect to each other (the Rx assembly) but with that Rx assembly physically decoupled from the Tx assembly allowing flexible data acquisition modes and deployment options. Such flexibility is expected to be instrumental in non-trivial terrains exhibiting either an abundant vegetation or being highly contaminated by large or dense clutter. Before validating the sensor in such challenging configurations, however, Pedemis was taken to Aberdeen Proving Ground, MD, for its first test site validation. We describe Pedemis, including its operation and data acquisition modes along with our Aberdeen Proving Ground results.

This paper details methods for automatic classification of Unexploded Ordnance (UXO) as applied to sensor data from the Spencer Range live site. The Spencer Range is a former military weapons range in Spencer, Tennessee. Electromagnetic Induction (EMI) sensing is carried out using the 5x5 Time-domain Electromagnetic Multi-sensor Towed Array Detection System (5x5 TEMTADS), which has 25 receivers and 25 co-located transmitters. Every transmitter is activated sequentially, each followed by measuring the magnetic field in all 25 receivers, from 100 microseconds to 25 milliseconds. From these data target extrinsic and intrinsic parameters are extracted using the Differential Evolution (DE) algorithm and the Ortho-Normalized Volume Magnetic Source (ONVMS) algorithms, respectively. Namely, the inversion provides x, y, and z locations and a time series of the total ONVMS principal eigenvalues, which are intrinsic properties of the objects. The eigenvalues are fit to a power-decay empirical model, the Pasion-Oldenburg model, providing 3 coefficients (k, b, and g) for each object. The objects are grouped geometrically into variably-sized clusters, in the k-b-g space, using clustering algorithms. Clusters matching a priori characteristics are identified as Targets of Interest (TOI), and larger clusters are automatically subclustered. Ground Truths (GT) at the center of each class are requested, and probability density functions are created for clusters that have centroid TOI using a Gaussian Mixture Model (GMM). The probability functions are applied to all remaining anomalies. All objects of UXO probability higher than a chosen threshold are placed in a ranked dig list. This prioritized list is scored and the results are demonstrated and analyzed.

ESTCP live-site UXO classification results are presented for cued data collected by the Man Portable Vector (MPV) handheld sensor, at the Former Spencer Artillery Range in Tennessee. The site was contaminated with assorted munitions, ranging in caliber from 37 mm to 155 mm. The MPV data were collected in two areas: dynamic and wooded. The data sets are inverted using an advanced forward EMI model, the ortho-normalized volume magnetic source (ONVMS) model, combined with a direct-search optimization algorithm known as differential evolution. All data are inverted assuming one, two, and three sources. For each inversion, the targets’ intrinsic parameters are extracted and used in a library matching technique. Anomalies are classified as targets of interest (TOI) or clutter. Prioritized dig lists were constructed and submitted to the Institute for Defense Analysis for independent scoring. The result revealed an excellent classification performance by the advanced EMI models when applied to the Spencer Range MPV data. This paper describes the MPV sensor and the advanced models and demonstrates the Receiver Operating Characteristic curves for the cued MPV data collected at the Spencer Range.

Advanced electromagnetic induction (EMI) sensors currently feature multi-axis illumination of targets and tri-axial vector sensing (e.g., MetalMapper), or exploit multi-static array data acquisition (e.g., TEMTADS). They produce data of high density, quality, and diversity, and have been combined with advanced EMI models to provide superb classification performance relative to the previous generation of single-axis, monostatic sensors. However, these advances yet have to improve significantly our ability to classify small, deep, and otherwise challenging targets. Particularly, recent live-site discrimination studies at Camp Butner, NC and Camp Beale, CA have revealed that it is more challenging to detect and discriminate small munitions (with calibers ranging from 20 mm to 60 mm) than larger ones. In addition, a live-site test at the Massachusetts Military Reservation, MA highlighted the difficulties for current sensors to classify large, deep, and overlapping targets with high confidence. There are two main approaches to overcome these problems: 1) adapt advanced EMI models to the existing systems and 2) improve the detection limits of current sensors by modifying their hardware. In this paper we demonstrate a combined software/hardware approach that will provide extended detection range and spatial resolution to next-generation EMI systems; we analyze and invert EMI data to extract classification features for small and deep targets; and we propose a new system that features a large transmitter coil.

This study is designed to illustrate the discrimination performance at two UXO active sites (Oklahoma’s Fort Sill and the Massachusetts Military Reservation) of a set of advanced electromagnetic induction (EMI) inversion/discrimination models which include the orthonormalized volume magnetic source (ONVMS), joint diagonalization (JD), and differential evolution (DE) approaches and whose power and flexibility greatly exceed those of the simple dipole model. The Fort Sill site is highly contaminated by a mix of the following types of munitions: 37-mm target practice tracers, 60-mm illumination mortars, 75-mm and 4.5′′ projectiles, 3.5′′, 2.36′′, and LAAW rockets, antitank mine fuzes with and without hex nuts, practice MK2 and M67 grenades, 2.5′′ ballistic windshields, M2A1-mines with/without bases, M19-14 time fuzes, and 40-mm practice grenades with/without cartridges. The site at the MMR site contains targets of yet different sizes. In this work we apply our models to EMI data collected using the MetalMapper (MM) and 2 × 2 TEMTADS sensors. The data for each anomaly are inverted to extract estimates of the extrinsic and intrinsic parameters associated with each buried target. (The latter include the total volume magnetic source or NVMS, which relates to size, shape, and material properties; the former includes location, depth, and orientation). The estimated intrinsic parameters are then used for classification performed via library matching and the use of statistical classification algorithms; this process yielded prioritized dig-lists that were submitted to the Institute for Defense Analyses (IDA) for independent scoring. The models’ classification performance is illustrated and assessed based on these independent evaluations.

Broadband, continuous-wave, electromagnetic induction sensors have been shown to perform well, but they usually use power inefficiently because of the power wasted in the linear amplifier used to drive the coils. Generally, much more power is dissipated in the power amplifier than in the coil in these systems. Methods for reducing this power are investigated using both linear and switched-mode amplifiers. Significant reductions in power dissipated by the linear amplifier are achieved by optimizing the signal used to drive the coil. Much greater reductions are achieved with the switched-mode amplifier using an optimized signal that has most of its energy in the desired frequencies for the sensor. Both of these techniques are shown to perform well in a prototype system.

Advanced electromagnetic induction arrays that feature high sensitivity wideband magnetic field and electromagnetic induction receivers provide significant capability enhancement to landmine, unexploded ordnance, and buried explosives detection applications. Specifically, arrays that are easily and quickly configured for integration with a variety of ground vehicles and mobile platforms offer improved safety and efficiency to personnel conducting detection operations including route clearance, explosive ordnance disposal, and humanitarian demining missions. We present experimental results for explosives detection sensor concepts that incorporate both magnetic and electromagnetic modalities. Key technology components include a multi-frequency continuous wave EMI transmitter, multi-axis induction coil receivers, and a high sensitivity chip scale atomic magnetometer. The use of multi-frequency transmitters provides excitation of metal encased threats as well as low conductivity non-metallic explosive constituents. The integration of a radio frequency tunable atomic magnetometer receiver adds increased sensitivity to lower frequency components of the electromagnetic response. This added sensitivity provides greater capability for detecting deeply buried targets. We evaluate the requirements for incorporating these sensor modalities in forward mounted ground vehicle operations. Specifically, the ability to detect target features in near real-time is critical to non-overpass modes. We consider the requirements for incorporating these sensor technologies in a system that enables detection of a broad range of explosive threats that include both metallic and non-metallic components.

The prevalence of unexploded ordnance (UXO), discarded military munitions (DMM), and munitions constituents (MC) at both active and formerly used defense sites (FUDS) has created a necessity for remediation efforts to mitigate the potential environmental and public health hazards posed by these munitions and explosives of concern (MEC). UXO remediation operations typically incorporate electromagnetic induction (EMI) or magnetometer surveys to identify potential MEC hazards located throughout cleanup sites. Often, significant costs are allocated for the intrusive investigation by dig teams of magnetic field anomalies associated with harmless objects such as fragmentation, scrap, or geological clutter at these sites. Recent developments in advanced EMI sensor technologies, i.e., those that employ multi-axis transmitter and receiver configurations, have enabled classification of a vast majority of these non-hazardous objects prior to excavation. One of the key requirements for successfully implementing MEC classification is the acquisition of high quality EMI data prior to analysis. Factors such as improper sensor positioning, low signal-to-noise ratio, or insufficient data sampling can lead to poor performance of classification algorithms. We present results from recent field evaluations of an approach for incorporating an in-field analysis of data quality metrics as part of the EMI survey process. Specifically, this approach applies a dipole inversion routine to the EMI data immediately after acquisition is complete. Data and model parameters are subsequently used to extract quality metrics, which are supplied to the operator in the form of a quality decision. This process provides the operator with high confidence that the data will yield effective classification results.

Electromagnetic induction (EMI) systems often use separate transmit and receive coils. In these systems, it is desirable for the transmit and receive coils to have minimal mutual coupling and a maximum eld product, thus maximizing the detection depth. We demonstrate that a pair of spiral coils can be optimized to achieve these desired properties. A mathematical representation is chosen for the coils that allows the coil pair to be optimized using an iterative convex method, which, due to its convexity, is very fast. We then present results showing both a pair of nonuniformly-wound, single-sided spiral coils and a pair of nonuniformly-wound, double-sided, spiral coils created with this optimization.

Dictionary matching techniques have been an e ective way to detect the location and orientation of buried targets using electromagnetic induction (EMI) sensors. Two problems with dictionary detection are that they require a large amount of computer storage to enumerate nine dimensions, and fine discretization of the parameter space must be used to reduce modeling error. The proposed method shrinks the dictionary size by five orders of magnitude, and reduces modeling error by directly solving for the 3×3 tensor model of the target. A robust lowrank matrix approximation algorithm has been implemented which can also account for directional insensitivities in the measurements.

Detection and discrimination of unexploded ordnance (UXO) in areas of prior conflict is of high importance to the international community and the United States government. For humanitarian applications, sensors and processing methods need to be robust, reliable, and easy to train and implement using indigenous UXO removal personnel. This paper describes system characterization, system testing, and a continental United States (CONUS) Operational Field Evaluations (OFE) of the PAC-MAG man-portable UXO detection system. System testing occurred at a government test facility in June, 2010 and December, 2011 and the OFE occurred at the same location in June, 2012. NVESD and White River Technologies personnel were present for all testing and evaluation. The PAC-MAG system is a manportable magnetometer array for the detection and characterization of ferrous UXO. System hardware includes four Cesium vapor magnetometers for detection, a Real-time Kinematic Global Position System (RTK-GPS) for sensor positioning, an electronics module for merging array data and WiFi communications and a tablet computer for transmitting and logging data. An odometer, or “hipchain” encoder, provides position information in GPS-denied areas. System software elements include data logging software and post-processing software for detection and characterization of ferrous anomalies. The output of the post-processing software is a dig list containing locations of potential UXO(s), formatted for import into the system GPS equipment for reacquisition of anomalies. Results from system characterization and the OFE will be described.

Currents on remote metallic objects such as landmines can be induced by projecting strong magnetic fields. These currents result in electromagnetic fields that can be subsequently detected. The magnetic field varies slowly as it passes from air into the ground and is sufficient to excite currents in buried metallic objects. Traditionally strong magnetic fields are produced using short-range transformer like inductive coupling, or as a component of powerful propagating electromagnetic fields. The strength of the magnetic component of the propagating electromagnetic field is restricted by regulatory limits on the total radiated radio frequency power. There is a need for a means to produce forward projected strong magnetic field at medium ranges with low-level propagation. This paper reports on a non-radiating loop antenna which maintains a constant amplitude and phase current around the loop and projects a strong magnetic field. The radiated field is small and results from the relativistic time-of-flight effect from one side of the loop to the other. The result is that a very strong magnetic field is produced in the near- to mid-field region, up to one wavelength away from the loop. Experiments with a prototype antenna and modeling show that the H-field is very high, radiated electromagnetic fields are negligible, and the drop off in field strength is inversely proportional to the distance squared. This agreement between experiments and modeling allows for a design based on computer simulations.

Many technologies are being developed to improve the detection of buried threats (e.g., landmines) and to discriminate these threats from clutter in an operational environment. Current systems have implemented ground-penetrating radar (GPR) and electromagnetic induction (EMI) sensors. The detection performance of these systems is assessed in field testing, where algorithms are used to determine when a buried threat has been encountered [1]. Similar to work done by Rosen and Ayers [2], this paper focuses on developing a method to study EMI sensor performance independent of any aided target recognition (ATR) algorithm used. Rosen and Ayers developed a method and a simple metric for assessing the mine-detection capabilities of down-looking GPR systems before an ATR algorithm is applied. This paper reports the development of two metrics for a wide-band EMI sensor based on the method used by Rosen and Ayers. In this initial effort, the values of the metrics developed are presented over different targets, and observations are made regarding potential use of this metric.

Identifying the important discriminating information demonstrated by objects in SAS imagery is important for automatic target recognition. We present a method for determining which information is important using a generative model for documents, introduced by Blei, Ng, and Jordan³ in which each document is generated by choosing a distribution over topics and then choosing each word in the document from a topic selected according to this distribution. We use this algorithm to analyze synthetic aperture sonar data by using Bayesian model selection to establish the number of topics. We show that the extracted topics capture meaningful structure in the SAS data, consistent with the class designations provided, and demonstrate the transfer of this knowledge across sensor domains.

In this paper we describe an unsupervised approach to seabed co-segmentation over the multiple sonar images collected in sonar surveys. We adapt a traditional single image segmentation texton-based approach to the sonar survey task by modifying the texture extraction filter bank to better model possible sonar image textures. Two different algorithms for building a universal texton library are presented that produce common pixel labels across multiple images. Following pixel labeling with the universal texton library, images are quantized into superpixels and co-segmented using a DP clustering algorithm. The segmentation results for both texton library selection criteria are contrasted and compared for a labeled set of SAS images with various discernable textures.

Land reclamation efforts in post-conflict regions are often hampered by the presence of Unexploded Ordnance (UXO) or other Explosive Remnants of War (ERW). Surface geophysical methods, such as Electromagnetic Induction (EMI) and magnetometry, are typically applied to screen rehabilitation areas for UXO prior to excavation; however, the prevalence of innocuous magnetic clutter related to indigenous scrap, fragmentation, or geology can severely impede the progress and efficiency of these remediation efforts. Additionally, the variability in surface conditions and local topography necessitates the development of sensor technologies that can be applied to a range of sites including those that prohibit the use of vehicle-mounted or large array systems. We present a man-portable EMI sensor known as the Electromagnetic Packable Technology (EMPACT) system that features a multi-axis sensor configuration in a compact form factor. The system is designed for operation in challenging site conditions and can be used in low ground-standoff modes to detect small and low-metal content objects. The EMPACT acquires high spatial density, multi-axis data that enable high resolution of small objects. This high density data can also be used to provide characterization of target physical features, such as size, material content, and shape. We summarize the development of this system for humanitarian demining operations and present results from preliminary system evaluations against a range of target types. Specifically, we assess the general detection capabilities of the EMPACT system and we evaluate the potential for target classification based on analysis of data and target model features.

We are in the process of developing a high-resolution landmine scanning system “ALIS” which produces horizontal slices of the shallow subsurface for visualization of buried explosives and inert clutter. As many AP mines contain minimum amounts of metal, metal detectors need to be combined with a complimentary subsurface imaging sensor. Ground Penetrating Radar (GPR) is widely accepted for subsurface sensing in the fields of geology, archaeology and utility detection. The demining application requires real-time imaging results with centimetre resolution in a highly portable package. The key requirement for sharp images of the subsurface is the precise tracking of the geophysical sensor(s) during data collection. We should also notice that GPR system is a very wide band radar system, and equivalent to UWB radar, which has recently been developed for short-range high-accuracy radar. We are testing simplified but effective signal processing for imaging mines. We are currently testing a dual sensor ALIS which is a realtime sensor tracking system based on a CCD camera and image processing. In this paper we introduce the GPR systems which we have developed for detection of buried antipersonnel mines and small size explosives. ALIS has been deployed in Cambodia since 2009 and detected more than 70 mines in mine fields, and returned more than 13ha cleaned fields to local farmers. We also report the current status of ALIS in Cambodia.

This paper describes a system for stand-off vapor detection based on Resonant Raman spectroscopy, RRS. The system is a step towards a RRS LIDAR (Light Detection And Ranging) system, capable of detecting vapors from explosives and explosives precursors at long distances. The current system was used to detect the vapor of nitromethane and mononitrotoluene outdoors in the open air, at a stand-off distance of 11–13 meters. Also, the signal dependence upon irradiation wavelength and sample concentration was studied in controlled laboratory conditions. A tunable Optical Parametric Oscillator pumped by an Nd:YAG laser, with a pulse length of 6 ns, was operated in the UV range of interest, 210–400 nm, illuminating the sample vapor. The backscattered Raman signal was collected by a telescope and a roundto- slit optical fiber was used to transmit collected light to the spectrometer with minimum losses. A gated intensified charge-coupled device (ICCD) registered the spectra. The nitromethane cross section was resonance enhanced more than a factor 30 700, when measured at 220 nm, compared to the 532 nm value. The results show that a decrease in concentration can have a positive effect on the sensitivity of the system, due to a decrease in absorption and selfabsorption in the sample.

We are developing a technique for the stand-off detection of trace explosives on relevant substrate surfaces using photo-thermal infrared (IR) imaging spectroscopy (PT-IRIS). This approach leverages one or more compact IR quantum cascade lasers, tuned to strong absorption bands in the analytes and directed to illuminate an area on a surface of interest. An IR focal plane array is used to image the surface and detect small increases in thermal emission upon laser illumination. The PT-IRIS signal is processed as a hyperspectral image cube comprised of spatial, spectral and temporal dimensions as vectors within a detection algorithm. The ability to detect trace analytes on relevant substrates is critical for stand-off applications, but is complicated by the optical and thermal analyte/substrate interactions. This manuscript describes recent PT-IRIS experimental results and analysis for traces of RDX, TNT, ammonium nitrate (AN) and sucrose on relevant substrates (steel, polyethylene, glass and painted steel panels). We demonstrate that these analytes can be detected on these substrates at relevant surface mass loadings (10 μg/cm² to 100 μg/cm²) even at the single pixel level.

The application of CW radar techniques to Nuclear Quadrupole Resonance (NQR) detection of nitrogen based explosives and chlorine based narcotics enables the use of low power levels, in the range of 10’s of watts, to yield high signal strengths. By utilizing Rabi transitions the nucleus oscillates between states one and two under the time dependent incident electromagnetic field and alternately absorbs energy from the incident field while emitting coherent energy via stimulated emission. Through the application of a cancellation algorithm the incident field is eliminated from the NQR response, allowing the receive signal to be measured while transmitting. The response signal is processed using matched filters of the NQR response which enables the direct detection of explosives. This technology has applicability to the direct detection of explosives and narcotics for security screening, all at safe low power levels, opposed to the current XRay and Millimeter wave screening systems that detect objects that may contain explosives and utilize high power. The quantum mechanics theoretical basis for the approach and an application for a system for security screening are described with empirical results presented to show the effects observed.

Recently, the development of methods for the identification of explosive materials that are faster, more sensitive, easier to use, and more cost-effective has become a very important issue for homeland security and counter-terrorism applications. However, limited applicability of several analytical methods such as, the incapability of detecting explosives in a sealed container, the limited portability of instruments, and false alarms due to the inherent lack of selectivity, have motivated the increased interest in the application of Raman spectroscopy for the rapid detection and identification of explosive materials. Raman spectroscopy has received a growing interest due to its stand-off capacity, which allows samples to be analyzed at distance from the instrument. In addition, Raman spectroscopy has the capability to detect explosives in sealed containers such as glass or plastic bottles. We report a rapid and sensitive recognition technique for explosive compounds using Raman spectroscopy and principal component analysis (PCA). Seven hundreds of Raman spectra (50 measurements per sample) for 14 selected explosives were collected, and were pretreated with noise suppression and baseline elimination methods. PCA, a well-known multivariate statistical method, was applied for the proper evaluation, feature extraction, and identification of measured spectra. Here, a broad wavenumber range (200- 3500 cm^-1) on the collected spectra set was used for the classification of the explosive samples into separate classes. It was found that three principal components achieved 99.3 % classification rates in the sample set. The results show that Raman spectroscopy in combination with PCA is well suited for the identification and differentiation of explosives in the field.

Recently, improvised explosive device (IED) has been a serious threat for many countries. One of the approaches to alleviate this threat is standoff detection of explosive molecules used in IEDs. Raman spectroscopy is a prospective method among many technologies under research to achieve this goal. It provides unique information of the target materials, through which the ingredients used in IEDs can be analyzed and identified. The main problem of standoff Raman spectroscopic detection is the large background noise hindering weak Raman signals from the target samples. Typical background noise comes from both ambient fluorescent lights indoor and sunlight outdoor whose intensities are usually much larger than that of Raman scattering from the sample. Under the proper condition using pulse laser and ICCD camera with nanosecond pulse width and gating technology, we succeed to separate and remove these background noises from Raman signals. For this experiment, we build an optical system for standoff detection of explosive molecules. We use 532 nm, 10 Hz, Q-switching Nd:YAG laser as light source, and ICCD camera triggered by laser Qswitching time with proper gate delay regarding the flight time of Raman from target materials. Our detection system is successfully applied to detect and identify more than 20 ingredients of IEDs including TNT, RDX, and HMX which are located 10 to 54 meters away from the system.

A novel cold cathode field emission array (FEA) X-ray source based on ultra-nanocrystalline diamond (UNCD) field emitters is being constructed as an alternative for detection of obscured objects and material. Depending on the geometry of the given situation the flat-panel X-ray source could be used in tomography, radiography, or tomosynthesis. Furthermore, the unit could be used as a portable X-ray scanner or an integral part of an existing detection system. UNCD field emitters show great field emission output and can be deposited over large areas as the case with carbon nanotube “forest” (CNT) cathodes. Furthermore, UNCDs have better mechanical and thermal properties as compared to CNT tips which further extend the lifetime of UNCD based FEA. This work includes the first generation of the UNCD based FEA prototype which is being manufactured at the Center for Nanoscale Materials within Argonne National Laboratory with standard microfabrication techniques. The prototype is a 3x3 pixel FEA, with a pixel pitch of 500 μm, where each pixel is individually controllable.

The goal of the proposed research is to automate the vehicle threat detection with X-ray images when a vehicle crosses the country border or the gateway of a secured facility (military base). The proposed detection system requires two inputs: probe images (from X-ray machine) and gallery images (from database). For each vehicle, the gallery images include the X-ray images of fully-loaded (with typical cargo) and unloaded (empty) vehicle. The proposed system produces two types of outputs for threat detection: the detected anomalies and the synthesized images (e.g., grayscale fusion, color fusion, and differential images). The anomalies are automatically detected with the block-wise correlation analysis between two temporally aligned images (probe versus gallery). The locations of detected anomalies can be marked with small rectangles on the probe X-ray images. The several side-view images can be combined into one fused image in gray scale and in colors (color fusion) that provides more comprehensive information to the operator. The fused images are suitable for human analysis and decision. We analyzed a set of vehicle X-ray images, which consists of 4 images generated from AS and E OmniView Gantry™. The preliminary results of detected anomalies and synthesized images are very promising; meanwhile the processing speed is very fast.

In this paper we present a high resolution magnetic field sensor that is based on the perturbation of the optical modes (whispering gallery mode, WGM) of a spherical dielectric resonator. The optical resonator is side coupled to a tapered single mode optical fiber. One side of the optical fiber is coupled to a distribute feedback diode laser, while the other end is connected to a photodiode. The optical modes of the dielectric cavity are perturbed using a metglas sheet that is in contact with the resonator. When the metglas sheet is exposed to an external magnetic field it elongates perturbing the optical modes of the dielectric cavity. This in turn leads to a shift in the optical resonances. By measuring the induced WGM shift the magnetic field can be measured. Preliminary results show sensor resolution of a few nanoteslas.

Cross-border smuggling tunnels enable unmonitored movement of people and goods, and pose a severe threat to homeland security. In recent years, we have been working on the development of a system based on fiber- optic Brillouin time domain reflectometry (BOTDR) for detecting tunnel excavation. In two previous SPIE publications we have reported the initial development of the system as well as its validation using small-scale experiments. This paper reports, for the first time, results of full-scale experiments and discusses the system performance. The results confirm that distributed measurement of strain profiles in fiber cables buried at shallow depth enable detection of tunnel excavation, and by proper data processing, these measurements enable precise localization of the tunnel, as well as reasonable estimation of its depth.

This study describes a cost-effective check-in baggage screening system, based on ‘on-belt tomosynthesis’ (ObT) and close-range photogrammetry, which is designed to address the limitations of the most common method of baggage screening, conventional projection radiography. The latter’s limitations can lead to loss of information and an increase in baggage handling time, as baggage is manually searched or screened with more advanced systems. This project proposes a system that overcomes such limitations creating a cost-effective automated pseudo-3D imaging system, by combining x-ray and optical imaging to form digital tomograms. Tomosynthesis is the creation of pseudo-3D images from a number of 2D projections which are acquired at a range of orientations around a static object. In the ObT system, instead of moving the source and detectors around the object, as in conventional CT, the movement of bags around bends in the baggage transport system provides the required relative motion between source, object and a fan configuration of stripdetectors. For image reconstruction it is necessary to accurately establish the sequential position and orientation of each bag as it is imaged. For this, a low-cost photogrammetric solution is used, based on geometrically calibrated web– cameras positioned around the bends where the bags are imaged. This paper describes a study demonstrating the feasibility of implementing close-range photogrammetry to a potential ObT system, for accurate determination of the object location. After this, an optimum ObT system is designed and built, the process of which is presented in this paper.

This manuscript discusses investigations into currents induced in linear conductors. The induced current is a useful indicator of the amount of scattering an electromagnetic field encounters in the presence of a linear conductor, and hence, the ease with which such a linear conductor could be sensed using electromagnetic radiation. The variation of induced current with several parameters is assessed using Method of Moment calculations. The final portion of the presentation will involve a comparison of the modeling results with acquired experimental data. Such comparisons are important for benchmarking theoretical models and will undoubtedly stimulate ongoing research.

A broadband antenna system has been developed to use polarization diversity for ground penetrating radar (GPR) applications. The antenna system uses four, crossed, resistive-vee dipole (RVD) antennas operating bistaticly to measure the transmission and reception of multiple polarizations. The RVD was selected because it has low self clutter, low radar cross section and wideband performance. The RVD is linearly polarized, but other polarizations can be synthesized through the use of two orthogonal RVDs to transmit or receive orthogonal field components. The antenna system is able to distinguish rotationally symmetric and linear targets with its ability to transmit right-hand circularly polarized (RHCP) fields and receive both left-hand circularly polarized (LHCP) and RHCP scattered fields. A target’s type can be identified by comparing the relative amplitudes of the received LHCP fields and RHCP fields. For example, the polarimetric antenna will be able to identify linear targets such as wires or pipes because linear targets scatter LHCP and RHCP fields in equal amounts. The configuration of the RVDs was optimized through simulation to achieve good circular polarization at close range and low coupling between the RVDs. Further simulations were performed which demonstrate that the polarimetric antenna provides sufficient information to identify linear targets from nonlinear ones, even at different target orientations. The polarimetric antenna was constructed and has been shown to also correctly detect and identify linear targets. Additional experiments revealed that the polarimetric antenna is effective for ground penetrating radar applications.

Millimeter wave absorption relative to background soil can be used for detection landmines with little or no metal content. At these frequencies, soil and landmine absorb electromagnetic energy differently. Stepped frequency measurements from 20 GHz to 60 GHz were used to detect buried surrogate landmines in the soil. The targets were 3 cm and 5 cm beneath the soil surface and coherent transmission and reflection was used in the experimental setup. The measurement set-up was mounted on a handheld portable device, and this device was on a rail for accurate displacement such that the rail could move freely along the scan axis. Measurements were performed with network analyzer and scattering data in frequency domain were recorded for processing, namely for inverse Fourier Transform and background subtraction. Background subtraction was performed through a numerical filter to achieve higher contrast ratio. Although the numerical filter used was a simple routine with minimal computational burden, a specific detection method was applied to the background subtracted GPR data, which was based on correlation summation of consecutive A-scan signals in a predefined window length.

In this work we present a comprehensive analysis of the scattered signals from buried landmine-like targets via accurate numerical modeling of Ground Penetrating Radar (GPR) responses considering various antenna-soil-target scenarios. Different characteristics in time and frequency domain are extracted and interpreted for each configuration. The acquired knowledge is useful to better understand the scattering mechanisms of subsurface objects and can be incorporated to target recognition procedures. A brief explanation of the results is also provided together with an overview of the most relevant temporal and spectral features encountered.

Visual detection of soil disturbances is a surprisingly effective, but far from perfect way of detecting buried explosive threats such as landmines and improvised explosive devices (IEDs). This effort builds upon the few systematic studies of optical detection in this area. It investigates observer sensitivity to optical information produced by the burial of anti-tank and small anti-personnel landmines asking “How detectable are disturbed soil signatures captured in visible (VIS), shortwave infrared (SWIR), and thermal infrared (TIR), bands?” “Which band or bands are most effective for detection?” and “How well does each band support detection in the natural environment over time?” Using signal detection procedures this study presented young adults photographs showing soil disturbed by landmine burial or adjacent undisturbed surfaces with instructions to make decisions about the presence or absence of a disturbance. Stimuli spanned a six-week time period over which VIS, SWIR, and TIR imagery was collected. Results show that (a) signal strength persists surprisingly well over the observation period, (b) generally, SWIR and VIS show consistently strong performance for large anti-tank mines and SWIR shows the soil signature for the small, anti-personnel mine remarkably well. TIR lags the other two bands when using d’ to measure performance, but shows promising hit rates for anti-tank mine signatures under appropriate conditions. Generally, results show that the SWIR and VIS bands show most promise as a practical means of explosive hazards detection, although TIR can work effectively for large anti-tank mines under certain conditions. Limitations and implications for further research are discussed.

Previous research by many groups has shown that broad-band thermal infrared (TIR) imagers can detect buried explosive threat devices, such as unexploded ordnance (UXO), landmines and improvised explosive devices (IEDs). Broad-band detection measures the apparent temperature - an average over the wave band of the product of the true soil surface temperature and the emissivity. Broad-band detection suffers from inconsistent performance (low signal, high clutter rates), due in part to diurnal variations, environmental and meteorological conditions, and soil surface effects. It has been suggested that hyperspectral TIR imaging might have improved performance since it can, in principle, allow extraction of the wavelength-dependent emissivity and the true soil surface temperature. This would allow the surface disturbance effects to be separated from the soil column (bulk) effects. A significant, and as yet unanswered, question is whether hyperspectral TIR images provide better detection capability (higher probability of detection and/or lower false alarm rate) than do broad-band thermal images. TIR hyperspectral image data of threat objects, buried and surface-laid in bare soil, were obtained in arid, desert-like conditions over full diurnal cycles for several days. Regions of interest containing threat objects and backgrounds were extracted throughout the time period. Simulated broad-band images were derived from the hyperspectral images. The diurnal variation of the images was studied. Hyperspectral was found to provide some advantage over broad-band imaging in detection of buried threat objects for the limited data set studied.

The developments of optical methods to characterize soils and various surface contaminants require complete and reliable databases of spectral signatures of various objects, including chemical and representative background surfaces. Ideally, the databases should be acquired in the field to properly consider the chemical mixing and heterogeneity of the surfaces. Spectral characterization instruments are common in the visible and the shortwave infrared but there are few solutions in the midwave and thermal infrared regions.

ABB recently developed a broad band reflectometer based on a small FTIR spectrometer. It is capable of measuring diffuse spectral reflectance from various surfaces in the infrared from 0.7 to 13.5 microns. This sensor has been developed to be operated in the field by one person. It is lightweight (about 12 kg); it is battery powered and ruggedized for operation in harsh environments. Its operation does not require sophisticated training; it has been designed to be operated by a non-specialist. The sensor can be used to generate spectral libraries or to perform material identification if a spectral library already exists.

Examples of measurements in the field will be presented.

Thermal inertia (effusivity) contrast marks the borders of naturally heated below ground object and void sites. The Dual Infrared Effusivity Computed Tomography (DIRECT) method, patent pending, detects and locates the presence of enhanced heat flows from below ground object and void sites at a given area. DIRECT maps view contrasting surface temperature differences between sites with normal soil and sites with soil disturbed by subsurface, hollow or semi-empty object voids (or air gaps) at varying depths. DIRECT utilizes an empirical database created to optimize the scheduling of daily airborne thermal surveys to view and characterize unseen object and void types, depths and volumes in “blind” areas.

In this paper we investigate a new approach for representing objects in FLIR images based on shearlets. Similar to wavelets, shearlets represent an affine system for image representation obtained by scaling and translation of a generating function called mother shearlet. Unlike wavelets, the mother shearlet has an extra parameter called shear that allows the shearlet transform to be anisotropic. Anisotropic property of the shearlet transform could allow for a better representation of objects with irregular shape. We test our representation methodology on Froward looking long wave infrared (LWIR) images obtained from an IR camera installed on a moving vehicle. Objects of interest (spots) are detected in each frame using a prescreener presented in our previous work. Each spot is then represented using its shearlet features and assigned a confidence coming from a support vector machine classifier. We compare shearlets to various traditional features such as local binary patterns (LPB) and histogram of gradients (HOG). The comparison is performed on a large dataset that consists of 16 runs at a US Army test site.

In this paper we describe an approach for optimizing the parameters of a Support Vector Machine (SVM) as part of an algorithm used to detect buried objects in forward looking infrared (FLIR) imagery captured by a camera installed on a moving vehicle. The overall algorithm consists of a spot-finding procedure (to look for potential targets) followed by the extraction of several features from the neighborhood of each spot. The features include local binary pattern (LBP) and histogram of oriented gradients (HOG) as these are good at detecting texture classes. Finally, we project and sum each hit into UTM space along with its confidence value (obtained from the SVM), producing a confidence map for ROC analysis. In this work, we use an Evolutionary Computation Algorithm (ECA) to optimize various parameters involved in the system, such as the combination of features used, parameters on the Canny edge detector, the SVM kernel, and various HOG and LBP parameters. To validate our approach, we compare results obtained from an SVM using parameters obtained through our ECA technique with those previously selected by hand through several iterations of “guess and check”.

There is a strong need to develop an automatic buried explosive hazards detection (EHD) system for purposes such as route clearance. In this article, we put forth a new automatic detection system, which consists of keypoint identification, feature extraction, classification and clustering. In particular, we focus on a new soft feature extraction process from forwardlooking long-wave infrared (FL-LWIR) imagery based on the use of an importance map derived from a bank of Gabor energy filters. Experiments are conducted using a variety of target types buried at varying depths at a U.S. Army test site. An uncooled LWIR camera is used and the collected data spans multiple lanes and times of day (due to diurnal temperature variation that occurs in IR). The preliminary receiver operating characteristic (ROC) curve-based performance presented herein is extremely encouraging for FL-EHD.

In previous work an automatic detection system for locating buried explosive hazards in forward-looking longwave infrared (FL-LWIR) and forward-looking ground penetrating radar (FL-GPR) data was presented. This system consists of an ensemble of trainable size-contrast filters prescreener coupled with a secondary classification step which extracts cell-structured image space features, such as local binary patterns (LBP), histogram of oriented gradients (HOG), and edge histogram descriptors (EHD), from multiple looks and classifies the resulting feature vectors using a support vector machine. Previously, this system performed image space to UTM coordinate mapping under a flat earth assumption. This limited its applicability to flat terrain and short standoff distances. This paper demonstrates a technique for dense 3D scene reconstruction from a single vehicle mounted FL-LWIR camera. This technique utilizes multiple views and standard stereo vision algorithms such as polar rectification and optimal correction. Results for the detection algorithm using this 3D scene reconstruction approach on data from recent collections at an arid US Army test site are presented. These results are compared to those obtained under the flat earth assumption, with special focus on rougher terrain and longer standoff distance than in previous experiments. The most recent collection also allowed comparison between uncooled and cooled FL-LWIR cameras for buried explosive hazard detection.

Forward Looking Infrared (FLIR) cameras have recently been studied as a sensing modality for use in buried threat detection systems. FLIR-based detection systems benefit from larger standoff distances and faster rates of advance than other sensing modalities, but they also present significant signal processing challenges. FLIR imagery typically yields multiple looks at each surface area, each of which is obtained from a different relative camera pose and position. This multi-look imagery can be exploited for improved performance, however open questions remain as to the best ways to process and fuse such data. Further, the utility of each look in the multi-look imagery is also unclear: How many looks are needed, from what poses, etc? In this work we propose a general framework for processing FLIR imagery wherein FLIR imagery is partitioned according to the particular relative camera pose from which it was collected. Each partition is then projected into a common spatial coordinate system resulting in several distinct images of the surface area. Buried threat detection algorithms can then be applied to each of these resulting images independently, or in aggregate. The proposed framework is evaluated using several detection algorithms on an FLIR dataset collected at a Western US test site and the results indicate that the framework offers significant improvement over detection in the original FLIR imagery. Further experiments using this framework suggest that multiple looks by the FLIR camera can be used to improve detection performance.

Frequency-domain electromagnetic induction (EMI) sensors have been shown to provide target signatures which enable discrimination of landmines from harmless clutter. In particular, frequency-domain EMI sensors are well-suited for target characterization by inverting a physics-based signal model. In many model-based signal processing paradigms, the target signatures can be decomposed into a weighted sum of parameterized basis functions, where the basis functions are intrinsic to the target under consideration and the associated weights are a function of the target sensor orientation. When sensor array data is available, the spatial diversity of the measured signals may provide more information for estimating the basis function parameters. After model inversion, the basis function parameters can form the foundation of model-based classification of the target as landmine or clutter. In this work, sparse model inversion of spatial frequency-domain EMI sensor array data followed by target classification using a statistical model is investigated. Results for data measured with a prototype frequency-domain EMI sensor at a standardized test site are presented. Preliminary results indicate that extracting physics-based features from spatial frequency-domain EMI sensor array data followed by statistical classification provides an effective approach for classifying targets as landmine or clutter.

A possibilistic K-Nearest Neighbors classifier is presented to classify mine and non-mine objects using data collected from a wideband electromagnetic induction (WEMI) sensor. The proposed classifier is motivated by the observation that buried objects often have consistent signatures depending on their metal content, size, shape, and depth. Given a joint orthogonal matching pursuits (JOMP) sparse representation, particular target types consistently selected the same dictionary elements. The proposed classifier distinguishes between target types using the frequency of dictionary elements selected by potential landmine alarms. Results are shown on data containing sixteen landmine types and several non-mine examples.

Handheld ground penetrating radar (GPR) devices, such as the AN/PSS-14, produce image data for a detection sequence. Sequences contain sweeps of left to right and right to left swings of the device. By smoothing the image scan and examining local minima, we can determine the sweep ranges and turn around points contained within the data. Different filters are used to determine the interval between sweeps and approximate the exact turn around point for each sweep. Images are then annotated with the start and end of sweep locations. Results presented are both qualitative, based on comparison to labeling by humans, and quantitative, based on robot- collected data. Dynamic Time Warping (DTW) helps us align overlapping regions of a left to right sweep with its corresponding right to left sweep.

The detection of shallow buried low-metal content objects using ground penetrating radar (GPR) is a challenging task. This is because these targets are right underneath the ground and the ground bounce reflection interferes with their detections. They do not create distinctive hyperbolic signatures as required by most existing GPR detection algorithms due to their special geometric shapes and low metal content. This paper proposes the use of the Autoregressive (AR) modeling method for the detection of these targets. We fit an A-scan of the GPR data to an AR model. It is found that the fitting error will be small when such a target is present and large when it is absent. The ratio of the energy in an Ascan before and after AR model fitting is used as the confidence value for detection. We also apply AR model fitting over scans and utilize the fitting residual energies over several scans to form a feature vector for improving the detections. Using the data collected from a government test site, the proposed method can improve the detection of this kind of targets by 30% compared to the pre-screener, at a false alarm rate of 0.002/m².

This paper evaluates and compares the performance of two algorithms that have previously demonstrated their potential in underground target detection. Field data was obtained on specially prepared test fields, where various mine simulants, reference objects, and mine-like clutter where placed at precise locations in different soil types. The efficiency of both algorithms in terms of detection accuracies (ROC curves) and computational burden is compared, as well as the impact of preprocessing strategies. Based on the results, we discuss the convenience of both methods to be integrated in a real - time signal processing system considering their benefits and drawbacks.

The Run Packing (RP) fusion method is a novel algorithm that addresses the con dence level fusion problem when M different sensors (or alarm sources) produce alarms independently. The goal of such a fusion method is to map the output confidence range of each alarm source to a global range shared by all of the alarm sources. The shared global confidence range allows a single receiver operating characteristics (ROC) curve to be created, and this ROC then shows the global system performance trade-o s across all alarm sources. We explain the run packing algorithm, show its application to a multi-sensor buried explosive object detection system, and compare its performance to other fusion techniques.

Ground Penetrating Radar (GPR) is a widely used technology for the detection of subsurface buried threats. Although GPR data contains a representation of 3D space, during training, target and false alarm locations are usually only provided in 2D space along the surface of the earth. To overcome uncertainty in target depth location, many algorithms simply extract features from multiple depth regions that are then independently used to make mine/non-mine decisions. A similar technique is employed in hidden Markov models (HMM) based landmine detection. In this approach, sequences of downtrack GPR responses over multiple depth regions are utilized to train an HMM, which learns the probability of a particular sequence of GPR responses being generated by a buried target. However, the uncertainty in object depth complicates learning for discriminating targets/non-targets since features at the (unknown) target depth can be significantly different from features at other depths but in the same volume. To mitigate the negative impact of the uncertainty in object depth, mixture models based on Multiple Instance Learning (MIL) have previously been developed. MIL is also applicable in the landmine detection problem using HMMs because features that are extracted independently from sequences of GPR signals over several depth bins can be viewed as a set of unlabeled time series, where the entire set either corresponds to a buried threat or a false alarm. In this work, a novel framework termed as multiple instance hidden Markov model (MIHMM) is developed. We show that the performance of the proposed approach for discriminating targets from non-targets in GPR data is promising.

Multiple instance learning is a recently researched learning paradigm in machine intelligence which operates under conditions of uncertainty. A Multiple Instance Hidden Markov Model (MI-HMM) is investigated with applications to landmine detection using ground penetrating radar data. Without introducing any additional parameters, the MI-HMM provides an elegant and simple way to learn the parameters of an HMM in a multiple instance framework. The efficacy of the model is shown on a real landmine dataset. Experiments on the landmine dataset show that MI-HMM learning is effective.

Identifying the ground within a ground penetrating radar (GPR) image is a critical component of automatic and assisted target detection systems. As these systems are deployed to more challenging environments they encounter rougher terrain and less-ideal data, both of which can cause standard ground detection methods to fail. This paper presents a means of improving the robustness of ground detection by adapting a technique from image processing in which images are segmented by local entropy. This segmentation provides the rough location of the air-ground interface, which can then act as a “guide” for more precise but fragile techniques. The effectiveness of this two-step “coarse/fine” entropyguided detection strategy is demonstrated on GPR data from very rough terrain, and its application beyond the realm of GPR data processing is discussed.

Prior to the calculation of target detection features, ground penetrating radar (GPR) data typically requires extensive preprocessing to suppress clutter artifacts and enhance signals corresponding to weaker targets. Optimization of this GPR signal preprocessing pipeline is necessary to provide the best opportunity at visual detection and automatic target recognition. Manual, independent adjustment of the many configuration parameters in the data preprocessing pipeline is inefficient and not guaranteed to find an optimal result. In this paper, the authors present a new metric for GPR processed data quality and demonstrate its utility in an automated parameter sweep optimization of a large set of algorithm configuration parameters. The observed costs and benefits of using automated preprocessing optimization are presented and discussed.

For preprocessing optimization and evaluation, a cost function was desired that is independent of the target detection features – to enable independent evaluation of the various components of the GPR target detection software. The proposed cost function, JSUM, is a signal-to-clutter ratio (SCR) metric, derived from the known KSUM metric. JSUM was developed to be sensitive to a particular type of noise in GPR data not captured by KSUM. The response of JSUM and KSUM to different common types of noise was explored to qualify the usefulness of the metric.

JSUM was used as the cost function for a parameter sweep optimization across a set of preprocessing parameters. The outcomes of this optimization are presented for discussion.

Utilizing methods from the image processing and computer vision fields has led to advances in high resolution Ground Penetrating Radar (GPR) based threat detection. By analyzing 2-D slices of GPR data and applying various image processing algorithms, it is possible to discriminate between threat and non-threat objects. In initial attempts to utilize such approaches, object instance-matching algorithms were applied to GPR images, but only limited success was obtained when utilizing feature point methods to identify patches of data that displayed landmine-like characteristics. While the approach worked well under some conditions, the instance-matching method of classification was not designed to identify a type of class, only reproductions of a specific instance. In contrast, our current approach is focused on identifying methods that can account for within-class variations that result from changing target types and varying operating conditions that a GPR system regularly encounters. Image category recognition is an area of research that attempts to account for within class variation of objects within visual images. Instead of finding a reproduction of a particular known object within an image, algorithms for image categorization are designed to learn the qualities of images that contain an instance belonging to a known class. The results illustrate how image category recognition algorithms can be successfully applied to threat identification in GPR data.

Multiple Instance Learning is a recently researched learning paradigm in machine intelligence which operates under conditions of uncertainty with the cost of increased computational burden. This increase in computational burden can be avoided by embedding these so-called multiple instances using a kernel function or other embedding function. In the following, a family of fast multiple instance relevance vector machines are used to learn and classify landmine signatures in ground penetrating radar data. Results indicate a significant reduction in computational complexity without a loss in classification accuracy in operating conditions.