Unexploded Ordnance Discrimination Research Articles

Synthetic case study: discrimination of unexploded ordnance (UXO) and non-UXO sources with varying remanent magnetization strength using magnetic data

SUMMARY We investigate if it is theoretically possible to discriminate between unexploded ordnance (UXO) and non-UXO sources by modelling the magnetic dipole moment for ferrous objects of different shapes and sizes. This is carried out by approximating the volumetric demagnetization factors of rectangular prisms, representing shapes similar to a long rod or flat steel plate. By modelling different UXO as prolate spheroids the demagnetization factors can be determined which can be compared with the magnetic response of a prism. The inversion is carried out in a probabilistic framework, where the UXO forward model and the non-UXO forward model are assigned individual prior models in terms of shape, size, orientation and remanent magnetization of the object. 95 independent realizations of the prism prior model are generated to make 95 synthetic anomalies exemplifying non-UXO objects, which are inverted for using the UXO model. It is investigated if an identical magnetic moment can be produced between the two models and how well resolved the magnetic moment is in terms of the measured anomaly. The case study is carried out in two steps where we first have little prior information of expected UXO properties and another where a UXO prior is introduced with expected values of aspect ratio and size of 24 different UXO, that are often encountered in the North Sea. With no prior information of expected UXO, discrimination is at many times implausible, unless elongated rod prism objects are considered, where the magnetic moment often can not be reproduced by a spheroid. Introducing the UXO prior we achieve a much better discrimination rate when using the list of expected UXO properties. By using the UXO prior we can account for a much higher remanent magnetization allowed in the prior, and still achieve high discrimination capabilities in comparison to a case with no UXO prior.

Learning Machine Identification of Ferromagnetic UXO Using Magnetometry

The fundamental problem in applying geophysical mapping to locate unexploded ordnance (UXO) is distinguishing true UXO from non-UXO. Enhancing the accuracy of UXO detection has multiple benefits, especially in the areas of cost savings and safety. We investigated discrimination approaches using both magnetic field data and numerically modeled data. Libraries of total field magnetic (TFM) responses were calculated using finite element modeling for three UXO types found at a Montana National Guard training site. UXO model parameters were varied over ranges of azimuth, declination, and depth resulting in approximately 600 models per UXO type. The modeled responses of finite-element model (FEM) and actual TFM field data were then used as training data in discrimination and classification approaches comparing neural networks (NN), random forests (RF), and support vector machines (SVMs). The prediction targets in the training process comprised three classes: 1) binary [UXO or noninteresting object (NIO)]; 2) multiclass (UXO round type and NIO); and 3) classes derived from multiclass self-organizing feature map (SOFM) analysis. The multiclass SOFM targets generated from site-specific field data were found to be optimal for UXO discrimination. The best performing combination of class selection types using recentered data for UXO detection rates of 100% resulted in a false alarm rate (FAR) of 28%.

The Orthonormalized Volume Magnetic Source Model for Discrimination of Unexploded Ordnance

We introduce a fast and accurate numerical technique for the solution of electromagnetic induction sensing problems called the orthonormalized volume magnetic source model. The model assumes that the secondary magnetic field measured by a sensor originates from a set of magnetic dipole sources distributed over a volume that coincides with the interrogated area. The Green functions associated with the responding sources are turned into an orthonormal basis using a generalization of the Gram-Schmidt method, enabling one to determine the sources' strengths directly from measured data without having to invert large and potentially ill-conditioned matrices. The method treats multitarget cases naturally and robustly. Several examples are presented to illustrate the applicability of the method in the discrimination of unexploded ordnance (UXO). In particular, we analyze data taken by the Time-Domain Electromagnetic Multisensor Towed Array Detection System sensor array at a test stand and during a blind test administered at a UXO live site. The method is highly successful in distinguishing UXO from among other UXO and from accompanying clutter.

A Comparison of the Finite-Element Method and Analytical Method for Modeling Unexploded Ordnance Using Magnetometry

Unexploded ordnance (UXO) is military ordnance that was fired, dropped, or emplaced but failed to function as intended and thus constitutes an explosive hazard. UXO is a worldwide problem that kills or maims thousands of civilians each year. Magnetic surveys are an efficient means of locating UXO containing ferrous metal when geologic conditions are sufficiently free of magnetic soil and rock. However, discrimination of UXO from non-UXO is complicated by the fact that UXO is often associated with high levels of clutter from ordnance fragmentation. To date, magnetic modeling of UXO has been based on calculations for a simple body of revolution geometry (prolate spheroids). We conducted an investigation to show how numerical modeling, in particular, finite-element modeling of more realistic geometries, compares to prolate spheroid results. Our results show that the calculated dipole moment response for complex models resembling actual UXO is up to 50% higher than the dipole moments for the prolate spheroid model. We also found that altering the shape of a model from a prolate spheroid to a complex shape has a greater effect on dipole moment than maintaining the same shape and altering the volume. Finally, in comparing the surface response from our models to real total field magnetic data, we find that complex models more closely match actual field data than prolate spheroid models. We suggest that modeling and, ultimately, discrimination using more realistic UXO shapes could result in significant improvements in distinguishing UXO from magnetic clutter and geology.

Joint diagonalization applied to the detection and discrimination of unexploded ordnance

Efforts to discriminate buried unexploded ordnance from harmless surrounding clutter are often hampered by the uncertainty in the number of buried targets that produce a given detected signal. We present a technique that helps determine that number with no need for data inversion. The procedure is based on the joint diagonalization of a set of multistatic response (MSR) matrices measured at different time gates by a time-domain electromagnetic induction sensor. In particular, we consider the Naval Research Laboratory’s Time-Domain Electromagnetic Multisensor Towed Array Detection System (TEMTADS), which consists of a [Formula: see text] square grid of concentric transmitter/receiver pairs. The diagonalization process itself generalizes one of the standard procedures for extracting the eigenvalues of a single matrix; in terms of execution time, it is comparable to diagonalizing the matrices one by one. We present the method, discuss and illustrate its mathematical basis and physical meaning, and apply it to several actual measurements carried out with TEMTADS at a test stand and in the field at the former Camp Butner in North Carolina. We find that each target in a measurement is associated with a set of nonzero time-dependent MSR eigenvalues (usually three), which enables estimation of the number of targets interrogated. These eigenvalues have a characteristic shape as a function of time that does not change with the location and orientation of the target relative to the sensor. We justify analytically and empirically that symmetric targets have pairs of eigenvalues with constant ratios between them.

The quantitative advantages of using B-field sensors in time-domain EM measurement for mineral exploration and unexploded ordnance search

The use of simple models for decay of conductive targets under conductive overburden and for the decay of magnetically permeable conductive steel objects allows quantitative consideration of the advantages of the use of magnetic-field detectors in time-domain electromagnetic (TEM) measurements, or more generally, the advantage of step response over impulse response TEM systems. We identified eight advantages of the step response versus impulse-response systems. The first two advantages relate to the inductive limit (early time) decay behavior, in which a target response amplitude is largely dependent on geometrical rather than conductivity parameters. Five further advantages occur when measuring response of a target in a conductive host or under conductive overburden; the maximum target-to-overburden response occurs 25%–30% earlier in time, the earliest target detection time occurs a factor 2–4 earlier, and the amplitude advantage of target-to-overburden response is a factor in the range of 1–10 for the step versus impulse-response systems, respectively. These advantages agree quantitatively with field observations on a chalcopyrite orebody under conductive cover. We used a model response for a conductive permeable sphere to derive mathematically consistent approximations for the power-law and exponential decay behaviors for step and impulse responses of metal objects, from which the onset of late-time exponential decay of EM responses of unexploded ordnance occurs about a factor of two earlier in time for the step response. This earlier-time transition together with the higher signal-to-noise ratio available from the step-response measurement makes measurement of the fundamental time-constant of unexploded ordnance (UXO) possible for medium and large UXO where the time constant is in the range of tens of milliseconds. This time-constant thus becomes accessible as an additional parameter for UXO characterization and discrimination.

Cart-Mounted Geolocation System for Unexploded Ordnance With Adaptive ZUPT Assistance

Reliable and precise navigation technology is essential for robust detection and discrimination of unexploded ordnance (UXO) in a wide range of field conditions, including GPS-challenged environment. Based on the fact that the UXO detection involves frequent stops, this paper proposes an adaptive ZUPT (AZUPT) algorithm in a sliding time window to guarantee the correct identification of the zero velocity condition and improve the position accuracy of UXO geolocation. Furthermore, a known zero velocity-based prefiltering method is used to remove INS sensor errors. A pushcart-based prototype in which the tactical-grade inertial measurement unit and geodetic-grade GPS receiver are mounted is designed for the geolocation system. Field test was undertaken, and simulated GPS outages were introduced to the data. The result shows that AZUPT precisely identified the zero velocity and greatly improved the positioning accuracy of the cart-mounted UXO geolocation system.

A Man-Portable Vector Sensor for Identification of Unexploded Ordnance

The identification and discrimination of unexploded ordnance using low-frequency electromagnetic induction is an expensive and difficult process, typically beset by low data diversity and high positioning uncertainty. In this paper, we present the Man-Portable Vector (MPV) sensor, a new time-domain instrument designed to remedy these shortcomings by measuring all three vector components of the secondary magnetic field at five distinct points around each transmitter location. The MPV also has a laser positioning system that can give its location with millimeter precision. After describing the instrument in detail, we study its performance in various sets of measurements, using the tensor dipole model to analyze the data. We find that the sensor can detect deeply buried targets and identify some standard ordnance items. It can also resolve separate targets in cases where two objects share the field of view and produce overlapping signals. A new incarnation of the MPV, the MPV-II, is in an advanced stage of development.

Incorporating Uncertainty in Unexploded Ordnance Discrimination

We examine representations of feature vector uncertainty in the context of unexploded ordnance (UXO) discrimination with electromagnetic data. We compare a local uncertainty estimate derived from the curvature of the misfit function with global estimates of the model posterior probability density (PPD) obtained with Markov chain sampling. For well-posed experiments (i.e., with high SNR and adequate spatial coverage), the two methods of uncertainty appraisal agree. However, when the inverse problem is ill posed, we find out that the PPD can be multimodal. To incorporate these uncertainties in discrimination, we first develop an extension of discriminant analysis which integrates over the posterior distribution of the model. When dealing with multimodal PPDs, we show that an effective solution is to input all modes of the PPD-corresponding to all models at local minima of the misfit-into discrimination and, then, to classify on the basis of the model which is most likely a UXO.

Dual-mode, Fluxgate-Induction Sensor for UXO Detection and Discrimination

We describe an innovative, dual-mode, fluxgate-induction sensor (FIS) that combines a DC magnetometer and an electromagnetic (EM) sensor. This sensor will enable one-pass unexploded ordnance (UXO) detection and discrimination using the joint measurements. The FIS is based on a [Formula: see text] long, high-permeability magnetic core and a sensing coil that are shared by both EM and magnetic (MAG) modes. This integration makes the sensor very compact and removes the potential crosstalk problem of the core material of one sensor dominating the response of the other. A prototype receiver was developed and true serial, dual-mode operation demonstrated. The FIS achieved sensitivities of [Formula: see text] rms in MAG mode and [Formula: see text] at [Formula: see text] in EM mode, which compare favorably to existing commercial instruments. Triaxial dipole modeling confirms that three-component EM data are better for shape characterization than one (vertical) component. Assuming axisymmetry, inversions of three-component EM measurements of 22 cylindrical and disk-shaped targets yielded 100% correct classification of UXO-like objects (cylinders) and 38% misclassification of disks as cylinders. Degraded classification performance with increasing vertical distance may be caused by the small cross-sectional area of the FIS; an array will provide better field sampling. Triaxial dipole modeling of the MAG data confirmed its utility for detecting deeper targets using only the vertical component. Based on these results, the FIS is ready for field demonstration and validation.

Realistic Subsurface Anomaly Discrimination Using Electromagnetic Induction and an SVM Classifier

The environmental research program of the United States military has set up blind tests for detection and discrimination of unexploded ordnance. One such test consists of measurements taken with the EM-63 sensor at Camp Sibert, AL. We review the performance on the test of a procedure that combines a field-potential (HAP) method to locate targets, the normalized surface magnetic source (NSMS) model to characterize them, and a support vector machine (SVM) to classify them. The HAP method infers location from the scattered magnetic field and its associated scalar potential, the latter reconstructed using equivalent sources. NSMS replaces the target with an enclosing spheroid of equivalent radial magnetization whose integral it uses as a discriminator. SVM generalizes from empirical evidence and can be adapted for multiclass discrimination using a voting system. Our method identifies all potentially dangerous targets correctly and has a false-alarm rate of about 5%.

Unexploded ordnance discrimination using magnetic and electromagnetic sensors: Case study from a former military site

In a study at a military range with the objective to discriminate potentially hazardous [Formula: see text] mortars from nonhazardous shrapnel, range, and cultural debris, six different discrimination techniques were tested using data from an array of magnetometers, a time-domain electromagnetic induction (EMI) cart, an array of time-domain sensors, and a time-domain EMI cart with a wider measurement bandwidth. Discrimination was achieved using rule-based or statistical classification of feature vectors extracted from dipole or polarization tensor models fit to detected anomalies. For magnetics, the ranking by moment yielded better discrimination results than that of apparent remanence from relatively large remanent magnetizations of several of the seeded items. The magnetometer results produced very accurate depths and fewer failed fits attributable to noisy data or model insuffi-ciency. The EMI-based methods were more effective than the magnetometer for intrinsic discrimination ability. The higher signal-to-noise ratio, denser coverage, and more precise positioning of the EM-array data resulted in fewer false positives than the EMI cart. When depth constraints from the magnetometer data were used to constrain the EMI fits through cooperative inversion, discrimination performance improved considerably. The wide-band EMI sensor was deployed in a cued-interrogation mode over a subset of anomalies. This produced the highest-quality data because of collecting the densest data around each target and the additional late time-decay information available with the wide-band sensor. When the depth from the magnetometer was used as a constraint in the cooperative inversion process, all [Formula: see text] mortars were recovered before any false positives were encountered.

Numerical Modeling of Higher Order Magnetic Moments in UXO Discrimination

The surface magnetic anomaly observed in unexploded ordnance (UXO) clearance is mainly dipolar, and consequently, the dipole is the only magnetic moment regularly recovered in UXO discrimination. The dipole moment contains information about the intensity of magnetization but lacks information about the shape of the target. In contrast, higher order moments, such as quadrupole and octupole, encode asymmetry properties of the magnetization distribution within the buried targets. In order to improve our understanding of magnetization distribution within UXO and non-UXO objects and to show its potential utility in UXO clearance, we present a numerical modeling study of UXO and related metallic objects. The tool for the modeling is a nonlinear integral equation describing magnetization within isolated compact objects of high susceptibility. A solution for magnetization distribution then allows us to compute the magnetic multipole moments of the object, analyze their relationships, and provide a depiction of the anomaly produced by different moments within the object. Our modeling results show the presence of significant higher order moments for more asymmetric objects, and the fields of these higher order moments are well above the noise level of magnetic gradient data. The contribution from higher order moments may provide a practical tool for improved UXO discrimination.

Assessing The Quality of Electromagnetic Data for The Discrimination of UXO Using Figures of Merit

The need for assessing data quality in unexploded ordnance (UXO) remediation problems arises from two sources. In the planning stage it is essential that the data are acquired in sufficient numbers and with sufficient accuracy to answer the detection or discrimination problem of relevance. At the interpretation stage it is critical to objectively assess whether the data are of sufficient quality to warrant subsequent processing, inversion, and classification. Faced with this practical challenge of defining data quality we propose a Figure of Merit (FOM). FOM is a reliability indicator derived from quantities that affect the quality of data, such as anomaly coverage, line spacing, station spacing, instrument noise, survey location errors, etc. The FOM can also include informative features of the inversion, such as the variance of key model parameters, and thus it depends on the inverse model to be applied. Anomalies associated with higher values of FOM should have increased reliability in classification. A...

Cooperative Inversion of Time Domain Electromagnetic and Magnetometer Data for The Discrimination of Unexploded Ordnance

Magnetic and electromagnetic induction (EMI) surveys are the primary techniques used for unexploded ordnance (UXO) remediation projects. Magnetometry is a valuable geophysical tool for UXO detection because of the ease of data acquisition and its ability to detect relatively deep targets. However, magnetic data can have large false alarm rates caused by geological noise, and there is an inherent non-uniqueness when trying to determine the orientation, size and shape of a target. EMI surveys, on the other hand, are relatively immune to geologic noise and are more diagnostic for target shape and size but have a reduced depth of investigation. We aim to improve discrimination ability by developing an interpretation method that takes advantage of the strengths of both techniques. We consider cooperative inversion, where information from the inversion of one type of data is used as a constraint for inverting another. We compare the confidence with which we can discriminate UXO from non-UXO targets when inverting the data sets cooperatively, to results from individual inversions. Examples are given of the application of the methodology to time domain electromagnetic induction (TEM) and magnetic data sets collected at the Yuma Proving Ground UXO Standardized Test Site calibration grid and the Former Camp Sibert.

Selecting a Discrimination Algorithm for Unexploded Ordnance Remediation

We review the algorithms that have been used to discriminate between hazardous unexploded ordnance (UXO) and harmless clutter. Statistical classifiers use model parameters estimated from geophysical data to formulate a decision rule. This rule tries to discriminate between UXO and clutter using the available information. In contrast, library-based discrimination algorithms make decisions using a predefined library of signatures for expected UXO types. Given the variety of algorithms that are available for UXO discrimination, we describe two metrics for evaluating discrimination performance - the area under the receiver operating characteristic and the false-alarm rate. We propose a bootstrapping algorithm for estimating these metrics when limited data are available. Last, we demonstrate this approach on real electromagnetic and magnetic data sets.

Unexploded ordnance discrimination using time-domain electromagnetic induction and self-organizing maps

Self-organizing maps (SOM) are implemented for discrimination of geologic noise, buried metal objects and unexploded ordnance using the geophysical method of time-domain electromagnetic induction. The learning and misfit measures are based on a Euclidean metric. The U*-matrix method is shown to be a reliable tool for determining data clusters and cluster boundaries. The performance of SOM for data-type discrimination was tested using three synthetic, idealized geophysical datasets consisting of exponential, multi-exponential and stretched-exponential decaying transients. In addition, experimental data were acquired using a modified Geonics EM63 instrument. Results from the synthetic examples show that SOM clusters the data based on their functional origin, when represented using U*-matrices. The percentage of correct classification is 100%. Unsupervised learning using the field dataset obtained with the Geonics EM63 succeeded in producing a multi-clustered map in which the background transients cluster themselves and are separated from clusters associated with metal clutter objects and UXO. Even though in some cases the SOM did not produce a single cluster for each type of causative body, it was able to separate clutter data from target data by producing several small clusters. The results are encouraging in view of the heterogeneity and sparsity of the training dataset.

Optimizing decision threshold and weighting parameter for UXO discrimination

Establishing a decision threshold to separate unexploded ordnance (UXO) from clutter is a critical task in UXO discrimination procedures. It is, however, a difficult task because the decision threshold depends on many parameters such as local geology, distribution of UXO and clutter, and cultural and environmental noise at a given UXO cleanup site. Some discrimination algorithms also use site-dependent weighting factors. In practice, one determines these parameters empirically from calibration experiments. To facilitate this task, we introduce a simple function called the degree of discrimination, constructed from calibration plots at a UXO cleanup site. First, we compute the degree of discrimination for a representative set of thresholds and weighting factors using ground truth data from the calibration site. Then, we find the optimum decision threshold and weighting factors associated with the best degree of discrimination. Finally, we use these optimal parameters for the UXO site cleanup. We have conducted two experiments using the broadband electromagnetic (EM) sensor with a concentric coil configuration. Our experiments show that 92% of UXO is declared and 74% of clutter is rejected using our optimized threshold and weights. A blind test shows that optimizing the threshold and weights increases the degree of discrimination from 0 to 0.15. Our tests also show that the detection scheme with a collective threshold for all UXO types is easier to use and yields better discrimination than the scheme based on individual thresholds for each UXO type.

Non-linear inversion of controlled source multi-receiver electromagnetic induction data for unexploded ordnance using a continuation method

This work adopts a continuation approach, based on path tracking in model space, to solve the non-linear least-squares problem for discrimination of unexploded ordnance (UXO) using multi-receiver electromagnetic induction (EMI) data. The forward model corresponds to a stretched-exponential decay of eddy currents induced in a magnetic spheroid. We formulate an over-determined, or under-parameterized, inverse problem. An example using synthetic multi-receiver EMI responses illustrates the efficiency of the method. The fast inversion of actual field multi-receiver EMI responses of inert, buried ordnances is also shown. Software based on the continuation method could be installed within a multi-receiver EMI sensor and used for near-real-time UXO decision-making purposes without the need for a highly-trained operator.

Experiences with unexploded ordnance discrimination using magnetometry at a live-site in Montana

Advanced discrimination methods and careful optimization of operational procedures are critical for efficient remediation of unexploded ordnance (UXO) contaminated sites. In this paper, we report on our experiences with a 200 acre magnetic survey that was collected and processed under production survey conditions at Chevallier Ranch, Montana. All anomalies with fitted moments above 0.05 Am 2 were excavated. During the survey the magnetic remanence metric was predicted but not used to guide the discrimination. The retrospective analysis presented here reveals that discrimination using remanence would have significantly reduced the total number of anomalies (with good dipolar fits) that needed to be excavated, from 524 to 290 while still recovering all 69 UXO. The false alarm rate (FAR = number of non-UXOs excavated divided / number of UXO found) was reduced from 6.3 to 2.9. At a cut-off of 75% remanence, 77% of anomalies due to shrapnel and metallic debris and 64% of geological anomalies were rejected. Geological anomalies due to variations in magnetite concentration introduced a significant human-element into the interpretation process. Three different interpreters added a total of 305 additional anomalies that were not fit with a dipole model and which were later found to be non-UXO. Between 40 and 50% of anomalies picked by the two relatively inexperienced interpreters who analyzed the data turned out to be geology, as compared to 14% for an experienced interpreter. Critical analysis of results, operator training and feedback from the UXO technicians validating the anomaly are essential components towards improving the quality and consistency of the anomaly interpretations. This is consistent with the tenants of Total Quality Management (TQM). We compare the actual FAR that resulted during the survey when there was little feedback between UXO technician validation results, to a hypothetical result that could have been achieved had there been a constant feedback system in place at the onset of operations. Feedback would have significantly reduced the number of geological anomalies and decreased the FAR from 10.7 to 4.0. The hypothetical results presented here demonstrate the value of using TQM principles to guide the UXO remediation process. They further show that improvements in the efficiency and costs of UXO remediation require both technological advances and operational optimization of the technology when implemented in a production setting. Furthermore, by treating geophysical modeling and UXO validation as separate entities, both with respect to contracting and operational reporting, there is little incentive for the geophysicist to leave an anomaly off the dig-sheet. Only potential negative consequences will result if that anomaly is later found to be a UXO. An incentive based mechanism that rewards the geophysicist for reductions in follow-on costs would have a strong potential to reduce the number of unnecessary excavations, and hence reduce the total cost of the UXO remediation effort.