Chapter Two

INNOVATIVE MINE DETECTION SYSTEMS

Researchers in physical, chemical, and biological sciences are

studying and developing methods that could reduce the false alarm
rate and maintain or increase the probability of detection for mine
clearance. New detection concepts involve searching for characteris-
tics other than mine metal content. A variety of techniques that
exploit properties of the electromagnetic spectrum are being
explored. In addition, research is under way to develop methods
based on acoustics of the mine casing. Biological and chemical
methods for detecting explosive vapors also are being explored, as
are methods for detecting bulk explosives based on chemical prop-
erties. Work also is under way to develop advanced prodders that
provide information about the physical characteristics of the object
being investigated.

This chapter describes the difficulties of predicting the performance

of innovative mine detection methods and our method for assessing
the potential of innovative systems. It then describes each type of
innovative technology and evaluates its potential to improve on
existing EMI detection systems. Table S.1 in the Summary provides
an overview of the technology evaluations.

METHOD FOR EVALUATING INNOVATIVE MINE

DETECTION SYSTEMS
Predicting the potential for an innovative mine detection system to
reduce the false alarm rate and increase the probability of detection
is an inherently difficult task. Research is being conducted by a myr-
iad of universities, government institutions, and private companies,

15
16 Alternatives for Landmine Detection

with different projects in various stages of development. However,

most of the technologies have not yet been field tested. This makes it
virtually impossible to assess operating characteristics (for example,
ROC curves) with any specificity, thus precluding defensible quanti-
tative performance comparisons.

Compounding the general lack of data to support quantitative evalu-

ations is a lack of comparability of the data that are available. The
performance of a landmine detection system depends on the types
and depths of mines present, the environment in which the system is
operated, and the human operator. For detectors that locate buried
objects, such mine properties as size, shape, and metallic content
substantially affect detector performance, as do the placements
(depth and orientation) of the mines. Detectors that search for
explosives can be sensitive to the type of explosive contained within
the mine. Detector performance is also tied to persistent environ-
mental attributes (such as soil type, terrain, vegetation, and clutter
density) and transient atmospheric conditions (wind, humidity, soil
moisture, and radio frequency or acoustic interference). Finally, such
human factors as individual operator tuning of the detector and
interpretation of the signals introduce additional sources of vari-
ability. Thus, results of reported field tests cannot be generalized,
with operating characteristics being conditional on the set of exper-
imental and field test conditions. As explained in Chapter One, such
variability is clearly illustrated in the results of the International Pilot
Project for Technology Cooperation field tests, shown in Figures 1.6
and 1.7. The primary implication is that even if published field
results were available for all current technologies, comparing per-
formance on these results alone would not be valid because of dis-
parities in the testing conditions and because of variability of detec-
tor performance in different environments.

Because of the barriers to credible quantitative comparisons of inno-

vative mine detection technologies, objective expert judgment is
required to interpret the available data on detector performance.
Without such judgment, it is easy to be misled. For example, many of
the existing field tests have been conducted in environments in
which confounding factors were minimized, yielding high probabili-
ties of detection and low false alarm rates. Nonpartisan judgment
thus was critical to the evaluations presented and to the stated tasks
of the study.
 Innovative Mine Detection Systems 17

To make the judgments about detector performance, we identified

two sets of prominent academic, government, and private-sector sci-
entists with expertise in landmine detection: (1) generalists, with
broad and long-standing experience in the landmine detection
community, and (2) specialists, currently at the forefront of develop-
ing specific cutting-edge detection technologies. Working with OSTP,
we appointed eight generalists to form the RAND S&TPI/OSTP Land-
mine Detection Technology Task Force. Based on input from the task
force members and our own independent literature review, we iden-
tified 23 leading specialists in current detection R&D efforts. Each
specialist then wrote a technical paper addressing the mode of
action, current capabilities and limitations, and potential future
performance of his or her specific technology. We requested that
each paper provide the following information:

brief description of the basic physical principles and mine fea-

tures (e.g., shape, explosives composition, metal content) ex-
ploited by the technology;
state of development of the technology (laboratory, bench, field);
current capabilities and operating characteristics (e.g., investiga-
tion times and examples of ROC curves from laboratory or field
tests);
known or suspected limitations or restrictions on applicability
(background clutter, mine type, environmental conditions, etc.);
estimated potential for improvement in the next two to seven
years;
outline of a sensible R&D program that could realize this poten-
tial, with rough projected costs to the extent possible; and
references to key technical papers describing any testing (espe-
cially field testing) of the technology.

The papers are published as technical appendixes (AW) to this

report.

The appendixes provided a structure for the task force to evaluate the
potential of each detection method. To conduct the evaluations, the
task force held several conference calls and met for two days at
RANDs Arlington, Va., office in May 2002. The summary evalua-
18 Alternatives for Landmine Detection

tion section included with each technology description below indi-

cates the results of the task force evaluations.

INNOVATIVE ELECTROMAGNETIC DETECTION SYSTEMS

A number of innovative methods are being explored that search for
buried mines based on changes in the electromagnetic properties of
the surface soil and shallow subsurface. These methods include
ground-penetrating radar (GPR), electrical impedance tomography
(EIT), x-ray backscatter, and infrared/hyperspectral systems.

Ground-Penetrating Radar
Description. GPR detects buried objects by emitting radio waves into
the ground and then analyzing the return signals generated by
reflections of the waves at the boundaries of materials with different
indexes of refraction caused by differences in electrical properties.
Generally, reflections occur at discontinuities in the dielectric con-
stant, such as at the boundary between soil and a landmine or
between soil and a large rock. A GPR system consists of an antenna
or series of antennas that emit the waves and then pick up the return
signal. A small computerized signal-processing system interprets the
return signal to determine the objects shape and position. The result
is a visual image of the object (see, for example, Figure 2.1) or an
audio signal indicating that its shape resembles a landmine, based
on comparison with a mine reference library.

The major design control in a GPR system is the frequency of the

radio wave. The scale at which GPR can detect objects is proportional
to the wavelength of the input signal, so the quality of the image
improves as the wavelength decreases and the frequency increases.
However, at high frequencies, penetration of the incident wave into
the soil can be poor. As a result, the designer must make a tradeoff
between quality of the image and required penetration depth. The
optimal design for maximizing image quality while ensuring suffi-
cient penetration depth changes with environmental conditions, soil
type, mine size, and mine position. Various alternative GPR designs
are being explored to optimize the tradeoff between penetration
depth and image quality under a wide range of conditions. Also criti-
 Innovative Mine Detection Systems 19

SOURCE: Ackenhusen et al. (2001).

NOTE: The bottom target is a metallic mine;
the top two targets are low-metal mines.

Figure 2.1Images of Landmines Produced by GPR System

cal in the design of a GPR system are signal-processing algorithms,

which filter out clutter signals and select objects to be declared as
mines.

GPR is a mature technology, but it has not yet been widely deployed
for mine detection. GPR was first used in 1929 to measure the depth
of an Austrian glacier (Olhoeft, 2002). The Army tested rudimentary
GPR techniques for mine detection in the 1940s. The first commer-
cial GPR systems were developed in 1972. Since then, use of GPR for
locating buried objects ranging from utility pipelines to archaeologi-
cal artifacts has proliferated. Although GPR is well established for
these other uses, understanding how different environmental factors
and mine characteristics affect its performance is far from complete.
Until very recently, GPR was unable to meet performance targets for
landmine detection established for military countermine operations
(see Appendix E).
20 Alternatives for Landmine Detection

The Army is currently completing development of a landmine detec-

tion system that combines GPR and EMI. This system is called the
Handheld Standoff Mine Detection System (HSTAMIDS) (see Figure
2.2). Field tested at Fort Leonard Wood, Mo., HSTAMIDS has
achieved probabilities of detection of 1.00 (with a 90-percent confi-
dence interval of 1.000.97) for metal antitank mines and 0.95 (with a
90-percent confidence interval of 0.970.93) for low-metal anti-
personnel mines, with an average false alarm rate of 0.23 per square
meter (see Appendix F). These test results cannot be extrapolated to
predict performance in other mined environments or even under
weather conditions different from those present on the day the test
was performed because of the significant effects of soil type and
moisture on GPR performance, as well as the variations in natural
clutter objects. Nonetheless, they do illustrate the potential for a
combined GPR/EMI system to achieve very high levels of perfor-
mance if the integrated sensor system can be optimized to account
for the effects of the local soil environment.

SOURCE: Minelab Countermine Divi-

sion, www.countermine.minelab.com/
countermine.asp.

Figure 2.2Prototype HSTAMIDS System

 Innovative Mine Detection Systems 21

Strengths. GPR has a number of advantages. First, it is complemen-

tary to conventional metal detectors. Rather than cueing exclusively
off the presence of metal, it senses changes in the dielectric constant
and therefore can find mines with a wide variety of types of casing
(not just those with metal). Generating an image of the mine or
another buried object based on dielectric constant variations is often
possible because the required radar wavelength is generally smaller
than most mines at frequencies that still have reasonable penetration
depth. Second, GPR is a mature technology, with a long performance
history from other applications. A fielded Army system (HSTAMIDS)
combining GPR with EMI for mine detection already exists and is
scheduled for production in 2003. Finally, GPR can be made light-
weight and easy to operate, and it scans at a rate comparable to that
of an EMI system.
Limitations. Natural subsurface inhomogeneities (such as roots,
rocks, and water pockets) can cause the GPR to register return signals
that resemble those of landmines and thus are a source of false
alarms. In addition, GPR performance can be highly sensitive to
complex interactions among mine metal content, interrogation fre-
quency, soil moisture profiles, and the smoothness of the ground
surface boundary. For example, Koh (1998) reports that because
excessive water causes rapid attenuation of radio waves, GPR will
perform poorly in wet soils for landmines buried below a depth of
about 4 cm. However, theoretical investigations by Rappaport et al.
(1999) indicate that increased soil moisture and interrogation fre-
quency may actually strengthen the return signal for nonmetallic
mines, but nonuniform soil moisture profiles (e.g., a wet surface and
dry subsurface) and rough ground surfaces present difficulties. For
the same mine, a given GPR can be very effective or ineffective,
depending on soil moisture and mine location; such complex inter-
plays make performance highly variable and difficult to predict. An
additional limitation is that unless the GPR system is tuned to a suf-
ficiently high frequency, it will miss very small plastic mines buried
at shallow depths because the signal bounce at the ground surface
(caused by the electrical property differences between air and soil)
will mask the return signal from the mine. Finally, the GPR system
designer must make a tradeoff between resolution of the return sig-
nal and depth, because high-frequency signals yield the best resolu-
tion but do not penetrate to depth.
22 Alternatives for Landmine Detection

Summary Evaluation. Current-generation GPR technology, such as

that embodied in HSTAMIDS, has the potential for high perfor-
mance. In addition, alternative approaches to GPR design (such as
the Wichmann system referred to in Appendix E) have the potential
to yield significant advancements over the available systems. How-
ever, the ability to model the radar response from different kinds of
landmines and natural clutter is essential for yielding the expected
performance gains. So far, such modeling is in its infancy. Ideally,
GPR systems would be able to provide high-resolution images to a
signal-processing system that could decide whether a buried object
is a root, rock, clutter object, or landmine. Ralston et al. (in Appendix
F) suggest development of a library of clutter signatures to aid in
this task.

Electrical Impedance Tomography

Description. EIT uses electrical currents to image the conductivity
distribution of the medium under investigation. Current implemen-
tations use a two-dimensional array of electrodes placed on the
ground, collecting conductivity data from stimulation of pairwise
combinations of electrodes. The data are then post-processed with
an algorithm that renders an image of the conductivity profile of the
subsurface volume. Both metal and nonmetal mines create anoma-
lies in the conductivity distribution that produce images, providing
information about the presence and location of mines.

Strengths. Because both metallic and nonmetallic mines create

conductivity anomalies, the technology is appropriate for detecting
all types of mines. Moreover, it is especially well suited for mine
detection in wet environments, such as beaches or marshes, because
of the enhanced conductivity of the moist substrate. The equipment
is relatively simple and inexpensive.

Limitations. The primary limitations are that the technology requires

physical contact with the ground, which might detonate a mine, and
that it cannot be used in such excessively dry, nonconductive envi-
ronments as desert or rocky surfaces. The technology is also sensitive
to electrical noise. Performance deteriorates substantially with the
depth of the object being detected for fixed electrode array size,
generally making it appropriate only for shallowly buried objects.
 Innovative Mine Detection Systems 23

The resolution is not as fine as that provided by other imaging tech-

niques, such as GPR.

Summary Evaluation. Because of its phenomenological limitations,

EIT is probably not well suited for broad humanitarian demining
needs. It has a potential niche for detecting nonmetallic mines in wet
environments, a task that confounds other technologies. Even for
this use, however, EIT is limited because it cannot detect at depth,
and often mines in moist environments, such as rice paddies, are
much deeper than those in shallow environments. A unique role for
EIT in humanitarian demining is not apparent from the available
information.

X-Ray Backscatter
Description. Traditional x-ray radiography produces an image of an
object by passing photons through the object. X rays have a very
small wavelength with respect to mine sizes, so in principle they
could produce high-quality images of mines. Although pass-through
x-ray imaging of the subsurface is physically impossible, the
backscatter of x rays may still be used to provide information about
buried, irradiated objects. X-ray backscatter exploits the fact that
mines and soils have slightly different mass densities and effective
atomic numbers that differ by a factor of about two.

There are two basic approaches to using backscattered x rays to cre-

ate images of buried mines. Methods that collimate (i.e., align) the x
rays employ focused beams and collimated detectors to form an
image. The collimation process increases size and weight and dra-
matically reduces the number of photons available for imaging.
Thus, high-power x-ray generators must be used as sources. The
large size, weight, and power requirements of such systems are not
amenable to person-portable detectors. Alternatively, uncollimated
methods illuminate a broad area with x rays and then use a spatial
filter to deconvolve the system response. They may be suitable for
person-portable detection.

Strengths. To readily distinguish mines from soils, it is necessary to

use low-energy incident photons (60200 keV). In this energy range,
cross sections are roughly 10 or more times larger than is possible
with most other nuclear reactions that would be applicable to mine
24 Alternatives for Landmine Detection

detection. In addition, because of the reduced shielding thickness

needed to stop low-energy photons, uncollimated systems can be
made small and relatively lightweight. Largely because of the medical
imaging industry, compact x-ray generators are now obtainable.
Low-energy isotopic sources have been readily available for a long
time. Practical imaging detectors are becoming more widespread,
although it may be necessary to custom build for mine detection
purposes. The medical imaging industry is likely to drive further
advances in x-ray imaging hardware.
Limitations. In the required energy range, soil penetration of x-ray
backscatter devices is poor. This limits detection to shallow mines
(less than 10 cm deep). If source strengths are kept low enough to be
safe for a person-portable system, the time required to obtain an
image may be impractically long. In addition, the technology is
sensitive to source/detector standoff variations and ground-surface
fluctuations. Further, to image antipersonnel mines, high spatial
resolution (on the order of 1 cm) is required. This may be difficult to
achieve in the field. Finally, the technology emits radiation and thus
will meet resistance to use because of actual or perceived risks.
Summary Evaluation. X-ray detection using the uncollimated
imaging approach may be useful for handheld confirmatory detec-
tion of antipersonnel landmines. In fielded systems, images of mines
are likely to be fuzzy but should still allow mines to be distinguished
from most diffuse or elongated false alarms. On the whole, however,
x-ray backscatter does not offer particular innovations or likely
avenues of improvement relative to other technologies and is un-
likely to yield substantial improvement in detection capabilities.

Infrared/Hyperspectral Systems
Description. Infrared/hyperspectral methods detect anomalous
variations in electromagnetic radiation reflected or emitted by either
surface mines or the soil and vegetation immediately above buried
mines (see Figure 2.3). The category encompasses technologies of
diverse modes of action, including active and passive irradiation
using a broad range of electromagnetic wavelengths.
Thermal detection methods exploit diurnal variations in tempera-
tures of areas near mines relative to surrounding areas. For example,
 Innovative Mine Detection Systems 25

SOURCE: Space Computer Corp., www.spacecomputer.com.

NOTE: The left pane shows the infrared image, while the right pane
shows the visible image. Mine locations are denoted with x.

Figure 2.3Infrared Image of Mines

mines or the soil above them tend to be warmer than surrounding

areas during the day but lose heat more quickly at night. Laser illu-
mination or high-powered microwave radiation can be used to
induce these differential temperature profiles.

Nonthermal detection methods rely on the fact that areas near mines
reflect light (either natural or artificial) differently than surrounding
areas. Anthropogenic materials tend to preserve polarization
because of their characteristically smooth surfaces, allowing dis-
cernment of surface mines. Moreover, the physical activity of
emplacing mines changes the natural soil particle distribution by
bringing small particles to the surface, which in turn affects the way
in which the soil scatters light. Systematic changes in vegetation
moisture levels immediately above buried mines also may be lever-
aged.

Strengths. These methods are attractive because they do not involve

physical contact and can be used from a safe standoff distance. They
are lightweight and are effective at scanning wide areas relatively
quickly. When deployed from airborne platforms, they are particu-
26 Alternatives for Landmine Detection

larly effective for detecting surface mines. Collecting and processing

the signals temporally (as opposed to in snapshots) tends to im-
prove performance by tracking diurnal cycles.

Limitations. The methods, particularly thermal imaging, have been

used in several prototype multisensor systems, but extreme variabil-
ity in performance as a function of dynamic environmental charac-
teristics has precluded their use for close-in detection and accurate
identification of mine locations. Despite maturity of the sensor, the
algorithms to process the signals in an informative way are relatively
undeveloped and are not linked to physical phenomena. Thermal
signatures currently are not well understood, and a comprehensive
predictive model does not exist. Moreover, waves at the frequencies
used by the methods cannot penetrate soil surfaces, and the local-
ized hyperspectral anomalies produced by mine emplacement are
ephemeral and are quickly eliminated by weathering. Thus, the tech-
nologies are able to detect buried mines under only limited transient
conditions.

Summary Evaluation. With the possible exception of methods that

would simulate solar heating as a means to enhance the thermal sig-
natures of buried targets, infrared/hyperspectral methods are not
particularly suitable for close-in buried mine detection. The underly-
ing phenomena are not sufficiently characterized, and natural
processes quickly erase the detectable surface anomalies. The tech-
nology has demonstrated ability and expected future promise for
airborne minefield detection, especially for surface mines, but it is
not expected to be useful for close-in detection of buried mines.

ACOUSTIC/SEISMIC SYSTEMS
Acoustic/seismic methods look for mines by vibrating them with
sound or seismic waves that are introduced into the ground. This
process is analogous to tapping on a wall to search for wooden studs:
materials with different properties vibrate differently when exposed
to sound waves. These methods are unique among detection meth-
ods that identify the mine casing and components in that they are
not based on electromagnetic properties.
 Innovative Mine Detection Systems 27

Description
Acoustic/seismic mine detection systems typically generate sound
(above ground) from an off-the-shelf loudspeaker, although there are
many possible configurations. Some of the acoustic energy reflects
off the ground surface, but the rest penetrates the ground in the form
of waves that propagate through the soil. When an object such as a
mine is buried, some of the energy reflects upward toward the
ground surface, causing vibration at the surface (see Figure 2.4).
Specialized sensors can detect these vibrations without contacting
the ground. A variety of different kinds of sensors (laser Doppler
vibrometers, radars, ultrasonic devices, microphones) have been
tried.

Researchers have field tested acoustic/seismic methods for landmine

detection on approximately 300 buried antitank and antipersonnel

RANDMR1608-2.4
5
x 10
5

Mine

4 Blank
Amplitude (m/s)

120 170 220 270

Frequency (Hz)

SOURCE: Rosen et al. (2000).

Figure 2.4Amplitude of Surface Vibration of Ground over a Mine

(solid line) and a Blank (dashed line) in Response to Sound Waves
28 Alternatives for Landmine Detection

mines and several hundred square meters of clutter locations at

Army field sites in Virginia and Arizona (see Appendix G). Initial tests
focused on antitank mines and yielded high probabilities of detec-
tion and low false alarm rates. For example, in one test, the acoustic
system identified 18 of 19 mines buried in dirt and gravel, yielding a
probability of detection of 95 percent. There was only one false alarm
in the test, even though the test site was seeded with clutter items
that had confounded a GPR system (Rosen et al., 2000). When the
system was modified with advanced signal-processing algorithms,
the false alarm rate dropped to zero.

Strengths
Acoustic/seismic sensors are based on completely different physical
effects than any other sensor. For example, they sense differences in
mechanical properties of the mine and soil, while GPR and EMI sen-
sors detect differences in electromagnetic properties. Thus, acoustic/
seismic sensors would complement existing sensors well.

Acoustic/seismic systems also have the potential for very low false
alarm rates. In experiments to date, false alarms from naturally
occurring clutter, such as rocks and scrap metal, have been ex-
tremely low (although such hollow clutter items as soda bottles and
cans would cause false alarms because the resonance patterns of
these objects are similar to those of mines). An additional strength is
that, unlike GPR systems, these sensors are unaffected by moisture
and weather, although frozen ground may limit the sensors capabil-
ity.

Limitations
The greatest limitation of acoustic/seismic systems is that they do
not detect mines at depth because the resonant response attenuates
significantly with depth. With current experimental systems, mines
deeper than approximately one mine diameter are difficult to find.

Also problematic is the slow speed of existing systems. Speed cur-

rently is limited by the displacement sensor, which senses the vibra-
tions at the surface caused by the sound waves. These displacements
are very small (less than 1 m) and are thus difficult to measure
 Innovative Mine Detection Systems 29

quickly in the adverse conditions of a minefield. The required scan

time for locating antipersonnel mines may range from 125 to 1,000
seconds per square meter (see Appendix H). However, a number of
methods are being investigated to speed up the detection process.
For example, an array of N sensors will speed the system by a factor
of N. Small prototypes of such arrays have been developed and can
be expanded and improved with further work.

An additional limitation of existing systems is that moderate to heavy

vegetation can interfere with the laser Doppler vibrometers that are
commonly used to sense the vibrations at the ground surface. A new
type of sensor could be developed, however, to overcome this flaw.

Summary Evaluation
Significant progress has been made in the past five to ten years in
developing acoustic/seismic mine detection systems. Interactions
between the seismic waves and buried mines and clutter are much
better understood, as are the seismic sources and displacement
sensors. The systems show great potential, but more research is
needed to make them practical. The development of an array of dis-
placement sensors that is fast, can penetrate vegetation, and can
function in the adverse conditions of a real minefield would be
especially useful.

EXPLOSIVE VAPOR DETECTION TECHNIQUES

Each detection technology discussed above searches for the casing or
mechanical components of a mine. Additional research is taking
place to develop methodsboth biological and chemicalthat
identify the presence of explosive vapors emanating from mines.
Ideally, such sensors would determine whether explosive vapors are
present above an anomaly located by a metal detector or other
device. Although each method has a different theoretical basis, all are
designed to sense low concentrations of explosive compounds or
their derivatives in soil or the boundary layer of air at the soil surface.
Determining the performance potential of each chemical- or biologi-
cal-sensing technology requires an understanding of how explosives
migrate away from landmines as well as knowledge of the chemical
and physical principles of the sensor.
30 Alternatives for Landmine Detection

When a mine is buried in the soil, it almost always will gradually

release explosives or chemical derivatives to the surrounding soil
through either leakage from cracks and seams or vapor transport
through the mine casing (in the case of plastic mines). While typically
about 95 percent of the explosive will adsorb to the surrounding soil,
the remaining 5 percent will travel away from the mine, mostly
through dissolution in water in the soil pores (see Appendix Q). Some
of this explosive will migrate to the ground surface in vapor form.
One of the key issues in detecting explosive vapors and residues is
that the concentrations available for detection are extremely low (see
Appendix Q). Thus the sensor must be able to operate at a very low
detection threshold. The analyte that is the focus of most explosive
detection research is 2,4-dinitrotoluene (2,4-DNT), which is a
byproduct of trinitrotoluene (TNT) manufacturing that is present as
an impurity in military-grade TNT. TNT has a very short half-life in
soil (about a day at 22C) because it is easily biodegraded and has
very low vapor pressure; 2,4-DNT is much less easily biodegraded
and has a higher vapor pressure, so it is the dominant chemical
present in the explosive signature from most landmines. In
experiments from a landmine test site reported in Appendix Q, the
2,4-DNT and 2,4,6-TNT concentrations in air above the soil were 200
10 15 g per milliliter and 1 1015 g per milliliter, respectively.
Figure 2.5 summarizes the ranges of concentrations of 2,4-DNT and
TNT vapors likely to be found in surface soils above landmines. To be
effective, an explosive vapor detection system must be sensitive to
concentrations as low as 1018 g per milliliter if the soil is very dry or
as low as 1015 g per milliliter if the soil is moist.

Biological Methods
Biological detection methods involve the use of mammals, insects, or
microorganisms to detect explosives. Like chemical sensors, these
methods rely on detection of explosive compounds rather than on
detection of metal or changes in the physical properties of the sub-
surface. Thus, they have the potential for reducing false alarm rates
from metal clutter. Each of the different methods operates on a dif-
ferent set of principles and is at a different stage of development. The
oldest involves using trained dogs, which were first shown to be
 Innovative Mine Detection Systems 31

RANDMR1608-2.5

Range of explosive concentrations in vapor

above plastic mines
Very dry soil Moist soil

Range of explosive concentrations in vapor 1 10 100

above metal mines
Very dry soil Moist soil 1 10 100 micrograms
10-6g/ml
nanograms (parts per
1 10 100 thousand)
10-9g/ml
(parts per
1 10 100 picograms million)
10-12g/ml
femtograms (parts per
1 10 100 billion)
10-15g/ml
(parts per
attograms
trillion)
10-18g/ml
(parts per
quadrillion)

SOURCE: Adapted from a figure by R. Harmon, Army Research Office.

NOTE: To be effective in all environments, a vapor detector ideally would
sense concentrations as low as 1018 g per milliliter in dry soil and 1015 g per
milliliter in moist soil.

Figure 2.5Range of Concentrations of 2,4-DNT and TNT in Boundary

Layer of Air Near Soil Above Landmines

capable of smelling landmines in the late 1970s (Johnston et al.,

1998). Methods employing insects and microorganisms are newer
approaches that have not yet been fielded.
Dogs and Rats. Description. Mine dog detection teams have long
assisted in humanitarian demining efforts. For example, more than
200 mine detection dogs currently are at work in Afghanistan. These
dogs can detect mines about 95 percent of the time under favorable
weather and soil moisture conditions (Horowitz et al., 1996).
Dogs have a keen sense of smell, originating from their ancestral
survival needs to find food, determine territorial boundaries, and
sense the presence of enemies (see Appendix T). By offering dogs a
reward of food or play, they can be trained to signal when they smell
mines. In the mine detection context, dogs walk ahead of their han-
dlers, noses to the ground, and sit at the first scent of a mine (see
Figure 2.6). A manual deminer then follows and investigates the area
with a probe. In another application, known as Remote Explosive
32 Alternatives for Landmine Detection

Scent Tracing mode, dogs sniff at filters that have collected vapors
near suspected mine locations. If a dog identifies a filter as contain-
ing explosives, then a deminer returns to the location from which the
vapor was sampled to look for a mine.
Currently, dogs are capable of detecting explosive vapors at concen-
trations lower than those measurable by the best chemical sensors,
so the lower limit at which they can detect explosives is uncertain
(Phelan and Barnett, 2002). One recent study recognized that avail-
able laboratory chemical analytical methods are far from the
sensitivity limits of the dog. Nevertheless, it attempted to determine
the detection threshold for dogs by diluting soil contaminated with
explosives to varying levels, two of which were 10 and 100 times
lower (based on extrapolation, not detection) than the current
chemical detection limit (Phelan and Barnett, 2002). The researchers
tested the ability of three different teams of trained dogs (one from
the United States, one from Angola, and another from Norway) to
identify explosives in samples from the various dilutions. They found
that a few of the dogs could correctly identify samples containing an
estimated 1016 g per milliliter of TNT or DNT. However, perfor-
mance varied by many orders of magnitude depending on the indi-
vidual dog, how it was trained, and the manner in which the training

SOURCE: RONCO Consulting Corp.,

www.roncoconsulting.com.

Figure 2.6Mine-Detecting Dog in Bosnia

 Innovative Mine Detection Systems 33

was reinforced. Further, detection performance of the dogs used in

this study also appears to have been influenced by environmental
conditions associated with the testing location and procedures fol-
lowed, including the inadvertent use of TNT-contaminated soil
samples as clean controls in testing at least one group of dogs.

As an alternative to using dogs or in conjunction with using dogs,

researchers at the University of Antwerp have trained African giant
pouch rats to detect mines. The rats are trained using food rewards
to signal the presence of explosives by scratching the ground surface
with their feet. Field tests of the use of rats in mine detection have
begun.

Strengths. Canines are proven to work exceptionally well in many

scenarios and under many environmental conditions. The olfactory
sensitivity of some, but not all, dogs is higher than the best currently
available mechanical detection methods. Advantages of using rats
include the possibility that they could be deployed in large numbers
and that they do not weigh enough to trigger mines, which reduces
the possibility of injury.

Limitations. Dog performance varies widely depending on the indi-

vidual dog, how it was trained, and the capabilities of the handler.
Further, dogs may need to be retrained periodically because they can
become confused if they discover behaviors other than explosives
detection that lead to a reward. An additional limitation is that when
trained to detect high levels of explosives, dogs may not automati-
cally detect much lower levels and may need to be specially trained
for this purpose. Like other methods that rely on vapor detection,
performance of mine detection dogs can be confounded by envi-
ronmental or weather conditions that cause explosive vapors to
migrate away from the mine or that result in concentrations of
vapors that are too low even for dogs to detect. Rats likely would have
similar limitations.

Summary Evaluation. Canines are proven performers and a valuable

asset in demining. However, continued investigation of the sensitiv-
ity of canine olfaction and how this varies with the dog and with
training is necessary to understand the factors that affect reliability.
Additionally, the vapor and particle signature of the mine in the field
must continue to be investigated to better understand performance
34 Alternatives for Landmine Detection

potential for canines. Additional research to explore the potential for

deployment of African giant pouch rats in demining also is war-
ranted.

Bees. Description. By lacing sugar with a target chemical and placing

the sugar in the bees natural foraging area, bees can be trained to
associate the chemical odor with food and to swarm over any loca-
tion containing the target odorant. Entomologists have trained bees
to detect a variety of explosives and have been researching ways to
use trained bees in humanitarian demining. There are two suggested
strategies. The first involves monitoring the movement of bees
trained to detect explosives and keeping track of the locations where
they swarm. The second involves sampling the beehive for the pres-
ence of explosives, which can be transported to the hive on the bees
mop-like hairs.

Several field tests have been conducted to investigate the potential

use of bees in mine detection (see Appendix S). The most recent and
comprehensive test involved placing DNT in petri dishes, covering
the DNT with sand, and placing the dishes in a flat, open space for
subsequent detection by the bees. In these experiments, bees proved
capable of detecting DNT concentrations that were estimated to be
0.713.0 ppb (approximately 1012 g per milliliter). Earlier lab testing
had indicated that bees could detect concentrations down to 20 ppt.

Strengths. Trained bees detect explosives and therefore are not lim-
ited by the same types of false alarms that plague metal detectors.
They also potentially could search a relatively large area in a short
time.

Limitations. As for chemical and bacterial detection systems, more

needs to be understood about the fate and transport of explosives in
the subsurface before the full potential of trained bees to detect
landmines can be understood. To date, no field trials using actual
mines have been conducted. Further, bees can only work under lim-
ited environmental and weather conditions. They do not work at all
at temperatures below 40F. In addition, all tests to date have been in
clear, open fields; whether bees would perform in forested or other
heavily vegetated environments is unknown. The inability to track
bee movements also currently poses difficulties.
 Innovative Mine Detection Systems 35

Summary Evaluation. Continued investigation of the use of bees in

mine detection is warranted. Experiments under more realistic con-
ditions could give a better indication of the potential of this method.
However, clear decision points should be established for determin-
ing whether to continue with research funding, if the method con-
tinues to look promising, or to terminate it, if there are insurmount-
able obstacles.

Bacteria. Description. During the 1990s, researchers engineered a

strain of bacteria that fluoresce in the presence of TNT (see Appendix
R and Kercel et al., 1997). A regulatory protein in the bacteria recog-
nizes the shape of the TNT molecule and fluoresces whenever the
molecule is present (see Figure 2.7). In principle, a bacterial mine
detection process would involve spraying bacteria on the mine-
affected area, possibly using an airborne system. The bacteria would
be allowed to grow for several hours. Then, a survey team would
return to search for fluorescent signals. The search could be con-
ducted either from an airborne system or using a handheld fluores-
cence detector.

One field trial using bacteria has been conducted. Five targets con-
taining from 4 oz to 10 lb (100 g to 5 kg) of TNT were placed in a

SOURCE: http://attic.jcte.jcs.mil/documents/forerunner_07_01.pdf.

Figure 2.7Bacteria Fluorescing in the Presence of TNT

36 Alternatives for Landmine Detection

quarter-acre field site. The bacteria detected all five targets, but there
were also two false alarms. Based on this single field trial, it is not
possible to determine the lowest concentration of explosive that
bacteria are capable of detecting.

Strengths. Like chemical sensors, bacteria can be engineered to be

highly specific to the explosive of concern. The regulatory protein
that causes the bacteria to fluoresce recognizes only TNT and struc-
turally similar molecules. Thus, this method has the potential to
reduce false alarms from clutter objects. An additional advantage is
that it may allow coverage of a large area in a relatively short time. In
theory, the unit cost of this method should decrease as the size of the
search area increases.

Limitations. The limited research to date has revealed possible envi-

ronmental limitations of this method. Bacteria are highly sensitive to
environmental conditions. The existing strain used to locate TNT
cannot survive at extreme temperatures. In addition, the method
functions only in moist soil because dry soil quickly absorbs the
bacteria. Another limitation is that the potential for false alarms is
unknown. For example, the two false alarms in the single field test
could have resulted from the migration of explosives away from the
targets or from some other chemical in the environment that trig-
gered the fluorescent response. An additional problem in experimen-
tal trials was that the fluorescence detector missed some of the sig-
nals from the bacteria. Finally, the performance potential of this
method will be limited by the fate and transport of explosives in the
subsurface. If the explosives migrate away from the mine, then the
bacterial signal may occur at a distance from the mine. In addition to
these operational limitations, public concerns about introducing
genetically engineered organisms into the environment may limit the
application of bacteria in mine detection.

Summary Evaluation. The potential for using bacteria in mine

detection remains largely untested, except for the single field trial
referenced in Appendix R. Nonetheless, continued investigation is
warranted as long as clear decision points for terminating or
continuing funding are established. For example, if research shows
that environmental confounding factors (such as moisture condi-
tions and temperature) preclude the use of bacteria in all but an ideal
 Innovative Mine Detection Systems 37

environment, then research may need to be halted unless bacterial

strains can be bred to overcome these limitations.

Chemical Methods
Description. A variety of possible nonbiological mechanisms for
detecting low concentrations of explosives in air or in soil samples
have been investigated in recent years (see Table 2.1). Most of these
investigations resulted from DARPAs Dogs Nose program, which
sponsored R&D leading to the development of highly sensitive odor
detection devices. Some of the techniques were patterned after the
mammalian nose. For example, one approach uses arrays of
polymer-based sensors that detect explosive vapors (and other
volatile chemicals) based on the amount of swelling in the polymers

Table 2.1
Chemical and Physical Methods for Sensing Explosive Vapors

Approximate
Detection Limit
Sensor Category Description (g explosive per ml air)
Fluorescent Measure a change in fluorescence 1015
wavelength on the tip of a polymer-
coated glass fiber or on an antibody
biosensor that occurs in response to
the presence of explosives
Electrochemical Measure changes in electrical 1012
resistance of arrays of polymers upon
contact with explosive vapors;
alternatively, measure changes in
electrical properties in coupled
electrode pair during reduction or
oxidation of explosives
Piezoelectric Measure shift in resonant frequency 1011
of various materials (thin polymer,
quartz microcrystal, or other) due to
mass change upon exposure to
explosive vapor
Spectroscopic Compare the spectral response of a 109
sample with that of a reference
material
38 Alternatives for Landmine Detection

caused by exposure to the analyte of concern (Freund and Lewis,

1995; Doleman and Lewis, 2001; Hopkins and Lewis, 2001). These
swelling differences, in turn, lead to easily detected changes in
electrical resistance of the polymers. Like mammalian noses, these
sensors rely on low-level responses of multiple different types of
receptors (many polymer types in the case of the mechanical
sensors, and many types of receptor proteins in the case of the
mammalian nose).

Of the various vapor sensors, a system using novel fluorescent poly-

mers is closest to being deployable and currently has the lowest
detection limit, as shown in Table 2.1 (Yang and Swager, 1998; Cum-
ming et al., 2001). The sensor consists of two glass slides, each cov-
ered by a thin film of the fluorescent polymer. When a sample of air
containing explosives passes between the slides, some of the explo-
sive binds to the polymer and in the process temporarily reduces the
amount of fluorescent light that the slides emit. A small photo-
multiplier device detects the reduced light emission, and electronics
signal the operator that explosives are present. Nomadics Inc. has
developed a man-portable prototype of this system and has con-
ducted limited field tests (see Figure 2.8). The current system can

SOURCE: Nomadics Inc., www.nomadics.com.

Figure 2.8Nomadics Sensor for Detecting Explosive Vapors

 Innovative Mine Detection Systems 39

detect explosive vapor concentrations as low as 1015 g per milliliter,

and additional development work is expected to lower this threshold
(Cumming et al., 2001).

Advantages. Vapor and residue sensors detect explosives and there-

fore could serve as complements to detection devices that rely on
physical features of the mine. In addition, most of the methods have
the potential to be engineered as small, lightweight, easily trans-
portable, and simple-to-operate systems with relatively low power
requirements. The Nomadics prototype already available is compa-
rable in size to a typical metal detector and, like a metal detector, can
operate at a walking pace. It has an extremely low detection thresh-
old (1015 g per milliliter).

Limitations. Perhaps the greatest obstacle that these systems must

overcome is the need for a probability of detection of one if they are
to replace manual prodders for confirming the presence of mines.
No deminer will exchange a conventional prodder for a device that
has less than near-perfect probability of detection because of the
obvious safety concerns involved. The detection sensitivity of current
technologies, with the exception of the fluorescent polymer ap-
proach, is not low enough to provide for reliable detection of metal-
encased mines in dry soil, and even this method may not perform
well in the driest of environments. Another problem is that the pres-
ence of explosive residues in soil from sources other than landmines
will trigger false alarms. This limitation would apply primarily to
battlefield environments because dispersed explosives remaining
after weapons are fired will biodegrade over time. Naturally occur-
ring chemicals that react with the polymers also may cause false
alarms. For all of these systems, the effects of variations in environ-
mental conditions, especially soil moisture, on performance is not
well understood. Further, the location at which explosive vapors are
present at the highest concentration is often displaced from the mine
location. Current understanding of explosives fate and transport
from buried mines is insufficient to allow for the reliable location of a
mine based on measurement of the extended explosive vapor signal.

Summary Evaluation. Explosive vapor and residue detectors have

the potential to be used as confirmatory sensors for landmines if the
probability of detection can be increased to near one. Whether this is
possible cannot be determined without additional basic research.
40 Alternatives for Landmine Detection

Research is necessary to establish the lower limits of detection for the

different types of vapor sensors. Also needed are further investiga-
tions to allow quantitative modeling of the amounts and locations of
explosives available for detection at the surface under different envi-
ronmental conditions.

BULK EXPLOSIVE DETECTION TECHNIQUES

Biological and chemical methods for detecting explosive vapors cur-
rently are limited by incomplete knowledge of how explosive vapors
migrate in the shallow subsurface. An additional category of explo-
sive detection technologies overcomes this limitation by searching
for the bulk explosive inside the mine. Methods being explored for
this purpose include nuclear quadrupole resonance (NQR) and a
variety of methods that use the interaction of neutrons with compo-
nents of the explosive. These technologies emerged from interest in
detecting bulk explosives in passenger baggage for the airline indus-
try and in investigating the potential presence of explosive devices in
other settings.

Nuclear Quadrupole Resonance

Description. NQR is a radio frequency (RF) technique that can be
used to interrogate and detect specific chemical compounds, includ-
ing explosives. An NQR device induces an RF pulse of an appropriate
frequency in the subsurface via a coil suspended above ground (see
the prototype in Figure 2.9). This RF pulse causes the explosives
nuclei to resonate and induce an electric potential in a receiver coil.
This phenomenon is similar to that exploited by magnetic resonance
imaging (MRI) used in medical testing, but NQR uses the internal
electric field gradient of the crystalline material rather than an exter-
nal static magnetic field to initially align the nuclei.

Strengths. NQR has a number of features that make it particularly

well suited for landmine detection. The primary attraction of NQR is
its specificity to landmines: In principle, it signals only in the pres-
ence of bulk quantities of specific explosives. Unlike many other
technologies, its false alarm rate is not driven by ground clutter but
rather by its signal-to-noise ratio (SNR). The SNR increases with the
 Innovative Mine Detection Systems 41

SOURCE: A. Hibbs, Quantum Applied Sci-

ence and Research Inc.

Figure 2.9Prototype Handheld NQR Mine Detector

square root of the interrogation time and also increases linearly with
the mass of the explosive. Thus, with sufficient interrogation time,
NQR can achieve nearly perfect operating characteristics (probability
of detection near one with probability of false alarm near zero). This
makes NQR more attractive as a confirmation sensor used to interro-
gate only those locations identified by other detectors (e.g., GPR,
EMI) as likely mine locations. Interrogation times of 0.53.0 minutes
may be sufficient for performance that leads to high probability of
detection (more than 0.99) and low probability of false alarm (less
than 0.05). The NQR signal from cyclotrimethylenenitramine (RDX)
is particularly large, implying high performance and small interroga-
tion times (less than three seconds) for detection of mines containing
RDX. Another positive feature of NQR is that it is relatively robust to
diverse soil conditions; for example, because it requires bulk con-
centration of explosives to declare, it is not misled by trace explosive
residues as can be the case with vapor-sensing techniques.

Limitations. The major weakness of NQR is the fact that, because of

its nuclear properties, TNT, which comprises the explosive fill of
most landmines, provides a substantially weaker signal than either
RDX or tetryl, posing a formidable SNR problem. Moreover, TNT
42 Alternatives for Landmine Detection

inherently requires longer interrogation times because its nuclear

properties preclude interrogation more frequently than once per five
to ten seconds. Another significant limitation is the susceptibility of
NQR to RF interference from the environment. This is especially
problematic for TNT detection because the frequencies required to
induce a response from TNT (790900 kHz) are in the AM radio band.
When present, radio signals overwhelm the response from TNT.

An additional weakness is that NQR cannot locate explosives that are

encased in metal because the RF waves will not penetrate the case.
This is not a major weakness because a large majority of antiperson-
nel mines have plastic cases, and EMI detection can successfully
detect those with metal cases. NQR also cannot detect liquid explo-
sives, but very few antipersonnel mines use liquid explosives.

NQR is very sensitive to the distance between the detection coil and
the explosive. Therefore, the detection coil must be operated very
close to the ground, which can be problematic in rough or highly
vegetated terrain. Moreover, current implementations require sta-
tionary detection for optimal results; detection in motion substan-
tially degrades the SNR.

Summary Evaluation. NQR is an explosive-specific detection tech-

nology that offers considerable promise as a technique for reducing
false alarms as compared with such conventional detection
approaches as EMI. It offers opportunities for improvement not
addressed by competing technologiesmost notably that its potency
derives from unique explosive signatures of mines. In principle, this
specificity affords it the possibility of a zero false alarm rate against
nonmetal mines. Currently, the most promising role of NQR is that of
a confirmation sensor used in conjunction with a conventional
scanning sensor or as part of an integrated multisensor detection
system.

Neutron Methods
Description. Neutron interrogation techniques involve distinguish-
ing the explosives in landmines from surrounding soil materials by
probing the soil with neutrons and/or detecting returning neutrons.
Differences in the intensity, energy, and other characteristics of the
 Innovative Mine Detection Systems 43

returning radiation can be used to indicate the presence of explo-

sives.

Only three of the many possible reactions involving neutrons or

gamma rays have reasonable potential for landmine detection. The
first, thermal neutron analysis, is the only nuclear technique that is
currently fielded by a military: The Canadian military uses it as a
vehicle-mounted confirmatory detector for antitank mines (see
Appendix N). Size and weight limits imposed by physics preclude it
from being person-portable or being able to detect small antiper-
sonnel mines in practical applications. The second method, known
as fast neutron analysis, has similar limitations. The third method,
neutron moderation, is the only one of the three with the potential to
yield a person-portable detector for antipersonnel mines. Neutron
moderation discerns buried materials with low atomic numbers (e.g.,
hydrogen).

Strengths. The physical properties of neutron moderation allow the

technology to use low-strength source radiation, which reduces
shielding required to protect workers from radiation exposure. Thus,
designing a handheld system may be possible. Costs of a production
imager are expected to be moderate.

Limitations. Neutron activation methods can, at best, measure rela-

tive numbers of specific atoms but cannot determine what molecular
structure is present. Because neutron moderation is most sensitive to
hydrogen, hydrogenous materials, particularly water, produce many
false alarms. Thus, to detect landmines successfully, it is necessary to
use the response from the neutrons to generate a visual image of the
area under investigation. Simulations show that the method will
work in soil with 10-percent moisture or less and may be usable
when moisture content is as high as 20 percent. Ground-surface
fluctuations and sensor height variation also contribute to false
alarms in nonimaging systems. Imaging can reduce these effects,
although some degradation of the image is expected.

With sources having sufficiently low strength to be practical for

handheld use, a few seconds will be required to acquire an image.
This makes neutron moderation imaging more suitable for confir-
mation than for primary detection. Also, there is a perceived (more
than actual) radiation hazard associated with nuclear techniques
44 Alternatives for Landmine Detection

that must be overcome by the users. Broad-area, low-power elec-

tronic neutron sources, under development by Defence R&D
Canada, could reduce this perceived risk.

Summary Evaluation. The majority of neutron technologies have

physical limitations that preclude them from being portable. Only
one technologyneutron moderation imagingmay be useful for
handheld confirmation of antipersonnel landmines. In fielded sys-
tems, images of these mines are likely to appear as fuzzy blobs, but
that will still allow mines to be distinguished from most diffuse or
elongated false alarms generated by moisture. On balance, however,
neutron moderation imaging is very unlikely to yield substantial
improvements in detection speed beyond what is capable with other
confirming detectors.

INNOVATIVE PRODDERS AND PROBES

The last step in mine detection has long been probing manually to
determine whether a signaled item is a mine or just harmless clutter.
The probe operator learns through experience to feel or hear the dif-
ference between a mine casing and other buried objects. Probing is
dangerous. Deminers often inadvertently apply sufficient force to the
mine to detonate it. This excess force does not always cause the mine
to explode, but when it does, the probe itself can become a deadly
projectile. New concepts are being explored to improve the safety of
probing and to help discriminate mines from clutter.

Description
Research to improve prodders and probes has followed two lines of
investigation: (1) development of probes that would signal to the
deminer when too much force is being applied and (2) development
of smart probes that provide information about the characteristics
of the item being investigated. These latter probes are intended to
provide information about some physical, electromagnetic, or
chemical characteristic of the object being investigated in order to
identify it. The only such probe engineered to date delivers an
acoustic pulse to determine whether the object is plastic, metal, rock,
or wood (see Appendix W). Performance of this smart probe was
mixed in limited testing: In a Canadian test, it correctly identified 80
 Innovative Mine Detection Systems 45

percent of the mines, but in a U.S. test it identified only 69 percent of

the mines. An improved version identified 97 percent of the mines in
a Canadian test. An alternative type of smart probe that is in the
research stage sprays focused jets of cold or heated water at regular
intervals to detect mines by sound or thermal signature.

Strengths
Probing is an established step in manual demining. Improved probes
could decrease the risks to deminers by providing feedback about the
nature of the object being investigated. In addition, theoretically, a
probe could deliver any of a number of different detection methods
(acoustic, electromagnetic, thermal, chemical, etc.), and the proxim-
ity of the probe to the landmine could improve performance. For
example, methods based on identifying explosive vapors likely would
perform better in close proximity to the mine, where vapor concen-
trations are much greater than those on the surface. However, such
advanced probes have yet to be developed.

Limitations
Any improved probe must essentially identify mines 100 percent of
the time to be accepted by the demining community. This is because
deminers view their current conventional probes as 100-percent
accurate. Some field testing has borne this out: A conventional mili-
tary prodder in one field experiment correctly distinguished between
38 mines and 119 rocks (see Appendix W). In addition, instrumented
probes may not be useful in dense ground or in ground with exten-
sive root structures.

Summary Evaluation
Because the probe is likely to remain part of the deminers tool kit for
the foreseeable future, limited efforts to develop more sophisticated
probes should continue. To be useful in speeding up demining oper-
ations, probes must be able to identify rapidly and accurately
whether the detected item is a mine or another object. Research
should continue on instrumented probes with detection devices that
analyze the material content of the item under investigation. To
46 Alternatives for Landmine Detection

increase probe safety, work should continue to develop probes that

indicate the level of force being applied.

ADVANCED SIGNAL PROCESSING AND SIGNATURE

MODELING
This chapter described a diverse range of innovative sensors for
obtaining signals that indicate the presence of landmines. Underly-
ing nearly all of these technologies are algorithms that translate these
signals into information that can be used (by either a human opera-
tor or an automated system) to make a declaration decision. Tradi-
tionally, these algorithms have been very simple and have focused on
detecting anomalies in the subsurface, as opposed to providing
information about the size, shape, or chemical content of the object.
For example, conventional handheld EMI detectors provide infor-
mation only about the strength of the return signal. Typically, the
frequency of the audible tone increases as the received signal
strength increases. The operator then decides what signal strength to
use as an indicator of the possible presence of a mine. Without the
additional step of discriminating background clutter from legitimate
targets, the number of false alarms is large and hampers the detec-
tion and clearance process.

The primary goal of advanced signal-processing algorithms is to

maximize the use of information generated by the sensor to help
discriminate targets from background clutter. Recent efforts have
combined information from physical models specific to the
particular sensor technology, statistical analyses of the generated
signals, and spatial information to achieve this goal (see Appendixes
U and V). Advanced algorithms leveraging statistical models have
been shown to reduce the false alarm rate for EMI, GPR, and
combined sensors in certain settings (Collins et al., 2002; Lewis et al.,
2002; Witten et al., 2002). Because of fundamental physical limita-
tions of the technologies, no amount of signal processing will elimi-
nate all false alarms from EMI and GPR systems. Nonetheless, such
advanced algorithms are important to improving mine detection
technologies. For nearly all types of sensors, advanced algorithms
could be or are being developed that make efficient use of the gener-
ated signal to improve operating characteristics. More important,
advanced signal-processing and signal fusion methods are crucial to
 Innovative Mine Detection Systems 47

the development of next-generation mine detection systems dis-

cussed in the next chapter.