Classification of Unexploded Ordnance Using Geophysics – A Practical Reality

For classification problems where the test data are labeled sequentially, the point at which all true positives are first identified is often of critical importance. This article develops hypothesis tests to assess whether all true positives have been labeled in the test data. The tests use a partial receiver operating characteristic (ROC) that is generated from a labeled subset of the test data. These methods are developed in the context of unexploded ordnance (UXO) classification, but are applicable to any binary classification problem. First, the likelihood of the observed ROC given binormal model parameters is derived using order statistics, leading to a nonlinear parameter estimation problem. I then derive the approximate distribution of the point on the ROC at which all true instances are found. Using estimated binormal parameters, this distribution can be integrated up to a desired confidence level to define a critical false alarm rate (FAR). If the selected operating point is before this critical point, then additional labels out to the critical point are required. A second test uses the uncertainty in binormal parameters to determine the critical FAR. These tests are demonstrated with UXO classification examples and both approaches are recommended for testing operating points.

The Very Early Time Electromagnetic (VETEM) system, the High-Frequency Sounder (HFS), and the Tensor Magnetic Gradiometer System (TMGS) were developed by the U.S. Geological Survey (USGS) for applications other than unexploded ordnance (UXO). However, features of these systems may make them adaptable to UXO applications. VETEM is a time-domain electromagnetic (EM) induction system that utilizes full waveform recording. The HFS is a flexible frequency-domain EM system. The TMGS uses four sensitive 3-axis fluxgate magnetometers so that gradients of field components are recorded. Tests were conducted in air using a selection of inert UXO objects ranging from 20 mm to 105 mm in diameter. A nulled overlapped antenna arrangement was used for both VETEM and the HFS. It was found that this antenna configuration resulted in distinctly different responses to horizontal UXO objects aligned along the direction of traverse versus horizontal and perpendicular to the traverse. Vertical orientation of UXO objects resulted in smaller amplitude signals and an interesting slightly asymmetric “crater” shape response in plan view. The HFS response was a strong function of frequency, which is good for discrimination. The TMGS data are the most sensitive to ferromagnetic objects and the multi-component data hold the promise of accurate location, including depth, to UXO objects using a dipole mapping algorithm. Plans include increased sensitivity and multicomponent capability for the EM systems, advanced algorithms to enhance UXO discrimination for EM and TMGS systems and field evaluations at standard UXO test sites of the Strategic Environmental Research and Development Program (SERDP). Introduction UXO is one of the most pressing problems facing the Department of Defense and other government agencies that have lands that were once used for military training and are now closed or closing and being returned to civilian use. Cleanup of all UXO-contaminated lands using existing methods would be prohibitively expensive, so a great deal of effort is being directed to finding better ways to detect, locate and identify buried UXO. It is not sufficient to merely detect buried metal objects, because many of these objects are not UXO and pose no hazards. In many cases of range cleanup 70% or more of the cost consists of locating and removing harmless metal including soda cans, broken plow blades and fragments of ordnance that exploded as designed. Among the primary geophysical methods for detecting and classification are various time and frequency domain EM systems and magnetometer systems. Since the USGS developed some prototype EM and magnetic systems for applications other than UXO, and because some of these prototype systems have some unusual characteristics, we proposed to test, evaluate, and modify three of these systems, the VETEM system, the HFS, and the TMGS. We are now in the process of carrying out this work and this paper describes some of the results from laboratory tests.

The detection and classification of unexploded ordnance (UXO) is considered a multidimensional pattern recognition problem. Standard techniques in solving multidimensional detection and classification problems involve using large sets of templates or libraries. This paper shows that by using wavelet transformation a single library will allow a particular class of ordnance to be classified over a range of depths.

: Buried Unexploded Ordnance (UXO) is a serious and prevalent environmental problem facing Department of Defense (DoD) facility managers. Not limited to active sites and test ranges, these problems also occur at DoD sites that are currently inactive and in areas adjacent to military ranges that belong to the civilian sector or are under control of other government agencies. Over the past decade, SERDP and ESTCP have invested heavily in developing survey data analysis and processing techniques for use with commercial and custom sensor suites that improve the detection, characterization, and classification of UXO and anthropic clutter. These techniques include procedures for quantitatively characterizing and classifying the anomalies. In this report, we compare and contrast the performances of feature-based discrimination decisions on magnetic and EMI data acquired at the Aberdeen Proving Ground (APG) Standardize Test Site. The APG test site was selected because it was designed to imitate conditions commonly found at munitions response sites, and includes emplaced ordnance ranging from 20mm projectiles to general practice bombs.

This paper presents an application of a simple surface magnetic charge model to discriminate<br>between objects of interest such as unexploded ordnance (UXO) and innocuous items, in cases when<br>signals from buried objects are a mixture of responses from two or more items. In the low frequency<br>(ten’s of Hertz up to several hundred’s of kHz) electromagnetic induction (EMI) sensing considered<br>here, both conduction and displacement currents may be neglected within the medium surrounding a<br>metallic object. Therefore, the scattered magnetic field outside the object is represented in terms of<br>scalar potential fields, from which one can obtain all scattered magnetic fields. While these are time<br>dependent by virtue of forcing functions or boundary conditions, they correspond in structure to static<br>fields. Such fields are appropriately, and readily, produced mathematically by equivalent elementary<br>magnetic charges placed on a convenient fictitious closed surface [1, 2, 3]. This forward model is<br>physically complete in the sense that all heterogeneity, near and far field, and internal interaction effects<br>within the object are included. It is very fast; in particular it can be implemented in inversion<br>calculations on a PC. According to the Gauss’s law,∇⋅B=ρeq , the net flux of magnetic field through any<br>closed surface equals the total (equivalent) magnetic charge ρeq inside the surface. The frequency<br>spectrum of this total induced equivalent charge is used here as a discriminant. Based on measured data<br>from two scatterers together, an iterative two step procedure is used in conjunction with the differential<br>evolution (DE) algorithm [4, 5]. One step determines of each object’s location and orientation and the<br>other determines the amplitudes of the responding fictitious magnetic charges. Once the objects are<br>isolated, the total magnetic charge for each is calculated as a function of frequency and compared to<br>cataloged/library data. Finally, blind classification analyses are performed for a single object as well as<br>for multiple subsurface scatterers, when two objects appear simultaneously within the field of view of<br>the sensor.

Efficient three-dimensional (3-d) convolutional neural networks (CNNs) are designed and trained for an underwater target classification task with volumetric synthetic aperture sonar (SAS) imagery. (The third dimension of the data represents depth into the sediment, thereby enabling the consideration of buried underwater objects.) The use of tiny networks containing relatively few parameters makes training with enormous input data volumes feasible even with modest computational power and limited computer memory. The promise of the approach is demonstrated for both buried and proud man-made objects present in real, measured SAS data cubes collected at aquatic sites by an experimental volumetric sonar system, called the Sediment Volume Search Sonar (SVSS). The classification performance of each 3-d CNN exhibits marked improvement over a prescreening detection algorithm alone, and the utility of an ensemble approach is also quantified. An analysis of the effective functionality of the learned networks is provided, with this also accompanied by figures showing example trained filters as well as intermediate representations of a data volume containing unexploded ordnance (UXO). The predictions of the 3-d CNN classifiers can provide valuable guidance for the efficient allocation of resources during real-world UXO remediation operations.

Mapping of the seabed is important for numerous applications, including the detection of buried objects such as cables, pipelines, unexploded ordnance (UXO), off-shore engineering, and soil remediation. This paper introduces multi-band interferometric sonar as a technique to map a volume of the seabed. It uses a high- frequency interferometric side scan sonar to map the sea surface, whereas lower frequencies are used to map the interior of the seabed. The combination of these two techniques enables the detection and localization of objects or features of interest, including accurate estimates of the depth of burial. This has been demonstrated on data acquired in the Haringvliet, the Netherlands. It is shown that the multi-band interferometric sonar system has the capability to detect buried objects and to map the thickness of a soft sediment layer (mud) in an efficient fashion.

Recent research sponsored by ESTCP and SERDP has emphasized advanced, multiple transmit/receive coil, electromagnetic induction (EMI) sensors to be used in a stationary mode to identify buried metallic objects as either unexploded ordnance (UXO) or clutter. Sponsored field demonstrations have used commercial EMI sensors to locate and flag anomalies. Advanced sensors then measure over each flag and indicate whether the buried item is dangerous UXO or harmless clutter. The collection of stationary data allows for multiple coil combinations and significant noise reduction. In turn, this high fidelity data allows for reliable model-based inversion and accurate identification of buried items. Work is underway to deploy some of these advanced EMI platforms in a survey mode. There are some advantages to this approach: (1) Only one sensor has to be deployed, (2) More channels of data are collected, (3) similar data between survey and stationary, and (4) the possibility of identifying some items with survey data only. Some of the down sides are: positioning errors, higher noise levels, and difficulty in deploying a sensor with many coils. We will look at data collected with the TEMTADS 2x2 cart in survey mode. Noise levels and sources will be presented and compared to the stationary sensor data. Much of the noise is due to sensor motion and it may be possible to reduce the noise by changing the platform's deployment, i.e. four wheels versus two or wheels versus skids. Current survey detection is carried out by looking for regions of peak signal. With the added channels of data from advanced sensors (more coils and time gates), there may be better ways to carry out detection. We will present preliminary results from a model-based inversion technique.

The problem of detecting buried unexploded ordnance (UXO) is addressed with a sensor deployed from a shallow-draft surface vessel. This sonar system produces three dimensional synthetic aperture sonar (SAS) imagery of both surficial and buried UXO across a range of environments. The sensor's hardware design was based in part upon data created using a hybrid modeling approach that combined results from separate environmental scattering and target scattering models. This hybrid model produced synthetic sensor data where the sensor/environment/target space could be modified to explore the expected operating conditions. The simulated data were also used to adapt a set of existing signal processing algorithms for formation of three-dimensional acoustic imagery. Recently, the sonar system has been integrated to a test platform, and experiments have been conducted at a trial site in the Foster Joseph Sayers Reservoir near Howard, PA. This test site has been prepared with several buried man-made objects. Initial results show that fully buried targets can be detected.

Bringing the State Back In? US and UK Government Regulation of Corporate Social Responsibility (CSR) in International Business

Advancements are being made to a system that uses an inertial measurement unit (IMU) to locally position an electromagnetic induction (EMI) sensor. The Small Area Inertial Navigation Tracking system (SAINT) has been interfaced to an EM61-HH with the EMI data logged on a handheld computer. To improve timing accuracy of the IMU and EMI data, a single board computer (SBC) is being implemented to collect measurements from both. This SBC system will include digital I/O options to interface to other advanced EMI systems under development. Operationally, the system can be swept about over a localized EMI anomaly for thirty seconds. In post-processing, the EMI sensor data is mapped out in both position and orientation. The positions and EMI data can be used in a dipole-based inversion. The inversion finds the magnetic polarization terms along an object’s axes which can be related to an object’s size and shape. Accurate inversion of these terms is possible because of the centimeter level positioning provided by the IMU system. With these terms well determined, the system can be used to classify objects as either unexploded ordnance or as clutter. Issues of sweep pattern and sensor time response on the inversion are being explored. Results from fitting repeated data over a four inch steel sphere are presented. The data was collected using three different sweep patterns over the sphere. No significant difference in the fits was observed between the sweep patterns.

Exploration of natural and artificial underwater structures is an actual field in development and research projects. Acoustic detection methods are well known and can be used to create seabed and subbottom profiles. The systems have limitations for detection of small objects, especially in presence of large acoustic reflectors. The detection of flat buried objects like unexploded ordnance (UXO) is very difficult with acoustic systems. This work presents a simple setup that utilizes three unfocused transceivers to detect small acoustic disturbances in presence of large acoustic reflectors. We suggest a differential ultrasonic receiver to eliminate equal signal components of two overlapping receivers to detect small objects. Furthermore, we present a differential single-receiver system and an evaluation of different algorithms for data visualization and compare with the frequently applied envelope in single receiver systems. It is possible to detect small objects with the diameter of 6 mm located in the water-sediment interface with the differential receiver setup.

From 2007 to 2017, the Environmental Security Technology Certification Program (ESTCP) sponsored a series of live-site demonstrations to assess the capabilities of newly developed advanced geophysical classification (AGC) technologies to detect and classify buried targets of interest (TOI) (unexploded ordnance (UXO), inert and surrogate munitions, seeds, and any other objects for which the site team agrees that removal is required). These technologies have been shown to detect TOI and reject clutter with a high probability of detection and low probability of false alarm, saving the DoD in remediation costs. Previous retrospective analyses summarized the capabilities of AGC to differentiate TOI from clutter without taking the size of the detected objects into account. The capability to predict the size of a buried object could allow the DoD to more appropriately set the minimum separation distance (MSD) in UXO remediation projects, thus saving more in remediation costs. We have therefore conducted a retrospective analysis of the ESTCP live-site demonstrations in order to determine the capabilities of AGC to inform the MSD.

: Recent studies have demonstrated the potentially serious impact of magnetic soils on the detection of buried objects such as land mines and unexploded ordnance (UXO) using magnetometers and electromagnetic inductions sensors, two methods that are routinely used in clearance operations. Magnetic soils can cause both equipment malfunctions and an increase in false alarms due to the presence of anomalies that have a geologic or pedogenic origin. To improve the discrimination performance of land mine and UXO detection sensors, there is a need for a better understanding of the natural variability in magnetic characteristics of soils. This research addressed this need by develop a conceptual model for the prediction of magnetic characteristics of soils developed on a wide range of geological parent materials, of different ages, and in diverse climatological environments. The research provides the Army with a new approach to the non-intrusive geophysical characterization of subsurface materials and their spatial distribution; the prediction of location, frequency, and scale of subsurface heterogeneity. The conceptual model that we have developed will allow for a better prediction of sensor capabilities in many iron-oxide containing field soils.

In 2010, a toxic tar-like material (TLM) was discovered in the riverbed of a popular Congaree River boat launch site near downtown Columbia, South Carolina. In this article, we examine the novel infrastructural form that emerged to deal with this waste, a temporary cofferdam that was built in 2022 as an integral element in the Congaree River Sediment Cleanup project. In doing so, we consider what we call the short and long histories of the cofferdam’s construction. These histories provide a way of glimpsing into how post-industrial landscapes and infrastructures are shaped by an inexhaustible list of potentially impactful factors both within and beyond contaminated spaces. We argue that the short history of the cofferdam’s size, shape, composition, and age have been shaped by its long history of factors that include potential unexploded Civil War ordnance, the presence of threatened and endangered species, and state and non-state actor relationships. This case study invites us to think more broadly about the presence of coal tar as an index, object, and constituent of different environment infrastructural forms.