Ground Penetrating Radar

1

Capabilities and limitations ofGround Penetrating Radar

Joshua Dee Smith, Brigham Young University, Electrical and Computer Engineering

Index Terms—Ground Penetrating Radar, Land Mine Detec-tion

Abstract—One hundred and ten million land minds havebeen distributed throughout the world across more than seventydifferent countries, and ninety percent of the victims of thesedevices are civilians. Plastic land mines are very prevalent andtend to attract younger victims with their bright colors [1].To help remove these hazards an imaging method that canidentify plastic and metal mines without making contact with theground is necessary. A Ground Penetrating Radar (GPR) systemis capable of identifying metal and plastic mines by launchingmicrowave radio frequencies into the ground to detect changes inpermittivity. Understanding these changes in permittivity allowsus to identify potentially dangerous objects hidden just below thesurface [2]–[8].

I. INTRODUCTION

According to the United Nations and their internationalresearchers, fifteen to twenty thousand individuals are maimedor die from exploding land mines each year. Of those injured,the majority are children similar to the boy shown in figure 1.The current cost to remove a land mine is anywhere from threehundred dollars to one thousand dollars per mine, and for everyfive thousand land mines successfully cleared, one worker willdie and two others will be severely injured. To make mattersworse, land mines can remain active for decades. Some landmines set during World War II are still active today [1].

Fig. 1. This figure shows a civilian victim of a land mine. This image wasprovided by Handicap International.

Land mines and improvised explosives are a weapon ofchoice by terrorists around the world today. Every day moreexplosives are being placed in roadways, walking paths, and

around important infrastructure to impede movement and in-crease fear. These explosives put our soldiers and the civiliansthey protect in danger each day. These new explosive devicesuse materials such as plastic and ceramics that cannot bedetected with metal detectors, but can still do significantdamage as shown in figure 2. These new materials, removalhazards, and the shear size of the problem have encouragedresearchers and private companies to develop methods whichsafely and accurately detect mines hidden in an area. GroundPenetrating Radar (GPR) systems provide a non-destructivemethod that can be safely deployed to quickly and accuratelydetermine the locations of mines hidden below the ground.

Fig. 2. This figure shows US tanks destroyed by land mines and otherimprovised explosives durring the Gulf War. This image was provided byhttp://www.group73historians.com/ .

GPR systems provide a safe method for locating landmines with precision. The system described in this paper wasdeveloped by NIITEK and uses a synthetic aperture radarantenna that is suspended above the ground in front of anarmored vehicle similar to the one shown in figure 3. With asuspended antenna, the truck will not accidentally detonate amine and endanger the operator. The GPR system can detecta mine made of metal or plastic in three dimensions whichgives the disposal team the best chances of safely removing ordisposing of the mine. The system can also paint the locationof a mine as well as tag the location of the mine usingcorrected GPR. To improve safety in extremely dangerousareas, a smaller version of the GPS system can be attached todrones and the be operated from a safe distance as shown infigure 4. The need to safely remove these mines is growingand developments in GPR are rising to meet this need.

2

Fig. 3. This figure shows the NIITEK GPR system mounted on the front ofa tractor. This image was provided by NIITEK at www.niitek.com.

Fig. 4. This figure shows a drone that has been fitted with a smaller versionof the GPR system. This image was provided by NIITEK at www.niitek.com.

II. BASIC DESIGN AND IMPLEMENTATION OF GPR

The theory behind GPR is similar to that of other radarsystems. A simple model consists of a two antenna system,one dedicated to transmission and the other to reception of thereturn signal. Coupled with the receive antenna is an analogto digital converter and a form of digital signal processingto improve the received data. The digitally adjusted returnsignal data and detected mine information are then given tothe operator through a digital display. A block diagram of theprocess is shown in figure 5.

This particular system has been developed by a privatecorporation with assistance from the American government.A sample set of data has been recored by the military ona specially designed dummy mine field using this system,and many research groups are using this data to developalgorithms for detecting mines with greater probabilities andlower constant false alarm rates. The specific SNR for varyingground types is not given directly in the literature, but Zhuand Wilson describe the system as having extremely low self-noise, low coupling, and an extremely high signal to noiseratio (SNR) [9], [10]. Information on the system specifications

Fig. 5. This figure shows a block diagram for a basic GPR system.

is limited but an explanation is given of how test data isacquired. The antenna array is 1.2 meters wide and contains 24individual v-dipole antennas spaced 5 centimeters apart. Thiscreates twenty four individual channels that sample 416 rangepoints below the surface for every 5 centimeters of lateralmovement. A sample is taken roughly every 8 picoseconds.The bandwidth of this system can range from 200MHz to7GHz and this is adjusted to help account for varying groundconditions [9], [10]. This sampling provides the user with athree dimensional block of data from which signal processingcan be used to detect mines and other explosive devices. Anexample of how the system is arranged is shown in figure 6,and an example of the system return for a land mine is shownin figure 7.

Fig. 6. This figure shows how the NIITEK array gathers data as it movesthrough a mine field.

If I were to design a similar system I would begin byevaluating the targets. Lets suppose anti-personal land mineis about 5 centimeters in diameter on average, and an anti-tank mine is roughly 10 centimeters in diameter on averageaccording to UN statics. If we apply the general equation fordetermining the radar cross section (RCS) of a spherical target,

σ = π ∗ r2, (1)

3

Fig. 7. This figure shows the GPR return for a metal land mine [9].

then we can calculate a simple approximation for the RCS ofa generic anti-personal land mine and an anti-tank land mine.This would be a very rough approximation and a better methodwould need to include the permittivity and permeability ofthe materials in question. These properties are often whatcreate the strongest effects with GPR return signals. Once wehave established the size of our smallest target we could usethis information to determine an appropate wave length. Afterchoosing an appropriate wave length, the necessary frequencycould then be calculated as,

f =c

λ. (2)

The Niitek system uses a series of 24 v-dipole antennas in theirarray. We might assume the effective area of the antenna isdetermined using the figure 8 assuming the v-dipole antennais using two quarter wave length sections to create a singlehalf wave length antenna. From this assumption the antennagain could be found using,

G =4πAe

λ2. (3)

It would be impractical to bury mines too deep, this forcesmost to be buried close to the surface. If the average mine isburied with in 25 centimeters of the surface, then we should seta maximum range to be at least half a meter to ensure detectionof even overly deep land mines. We know that the systemwill receive data from 416 range bins per antenna for every 5centimeters of lateral movement. The individual measurementsare taken in about 8 picoseconds. A pulse width of 0.25nanoseconds could be sufficient to transmit an excitation signalunder proper conditions, and this pulse width is used to definethe system’s bandwidth using,

B =1.2

τ. (4)

Fig. 8. This shows a v-dipole and the method I used to estimate its effectivearea.

We will assume the system losses are small based on reportsgiven by the manufacturer, the department of defense, andother researchers who have access to field data. AccordingDaniels [11], [12], a GPR systems main factor in determiningthe range of the system is the path loss. When sending anelectromagnetic signal into the ground, the losses will be solarge, in comparison to air, that the effects will be in the nearfield instead of the far field. The target of interest will bewithin centimeters instead of kilometers. The total path lossfor a given range is given in equation 5. [13], [14]

LT = Le + Lm + Lt1 + Lt2 + Ls + La + Lsc, (5)

where LT is the total path losses, Le is the loss in theantenna efficiency, Lm is the antenna mismatch losses, Lt1

is the transmission losses from air to the material, and Lt2 isthe retransmission losses. In this equation Ls is the antennaspreading loss, La is the attenuation loss, and Lsc is the targetscattering loss. The maximum range a GPR can achieve isdirectly related to the systems operating frequency and thelosses related to the propagation path of the radar. A simpleexample of a GPR system showing its range, footprint, andspreading is shown in figure 9.

Antenna losses, Le, and antenna mismatch losses, Lm, aredirectly related to the antenna used in the system and the wayit is connected to the transmitter and receiver. These lossesare often small and can be controlled with proper design andconstruction.

Transmission losses, Lt1, and retransmission losses, Lt2, arelosses related to the coupling of the radar system to the ground.Common GPR systems would be lying flat on the groundand this would allow the antenna to couple directly with theground minimizing negative effects caused at the surface. Inland mine detection it is important to maintain an air gapbetween the antenna and the ground. In figures 3 and figure 4we see that the radar antenna is lifted off the ground to reducethe chance of unintentional detonation of land mines. This gaprequirement forces this GPR system to couple the antenna tothe ground through the air. Indirect coupling adds an additionalloss media while introducing reflections and fringing from theinteraction with the surface of the ground. These additionallosses can be mitigated using proper signal processing.

4

Fig. 9. This figure shows a simple example of a standard GPR coupleddirectly to the ground.

The spreading loss, Ls, is the loss related to the ratio of thepower received versus the power transmitted. A GPR systemassumes that the beam travels directly below the transmittingantenna in a form more like a pencil beam, but in actualitythe beam will spread as it interacts with the soil, clutter, andtargets. This spreading causes the return signal to be weakerthan expected. The loss associated with this weakening indescribed as

Pr

Pt=

GtGrσ

(4πR2)2 . (6)

In this equation Gt is the gain of the transmitting antenna, GR

is the gain of the receiver, σ is the radar cross section of ourtarget, and R is the range to our target. Using this equation wecan define our spreading loss as the ratio of power receivedversus power input, and if we were detecting point chargesthen R would be raised to the fourth power. This equationchanges slightly for different shaped targets [11], [12].

Target scattering losses, Lsc are dependent upon the geom-etry of the target, its material composition, and its location.Traditional land mines were made of metal and thus providean excellent return signal for our radar system if located abovea threshold depth. Our ability to detect plastic land mines andother explosive devices depends heavily upon the dielectricproperties of the materials in which they were buried. A sharpcontrast in relative permittivity will give a stronger return thana very weak or gradual change in permittivity.

Material attenuation losses, La describe the losses associ-ated with traveling through the different layers of materialsunder the ground. Different soil types have different lossparameters which are directly related to the material’s relativepermittivity, permeability, and loss tangent. This parameter isalso affected by the users choice in operating frequency, highfrequencies give better resolution at reduced ranges and lowfrequencies give reduced resolution at greater ranges. Danielsgives a useful table of common soil elements and their lossesat a reference frequency [11], [12].

Approximations for these losses were chosen based on

Parameter Decibel ValueRCS for Anti-T* -54.1416 dBRCS for Anti-P* -66.1784 dB

Wavelength -26.0206 dBFrequency 135 dB

Antenna Gain 55.9868 dBMaximum Range -6.0206 dB

System Temperature 49.5424 dBPulse width -192.041 dBBandwidth 193.625 dB

System Losses 3 dBAntenna Efficiency -4 dBAntenna Mismatch -1 dB

Losses Antenna to Ground -2 dBLosses Ground to Antenna -2 dB

Spreading Losses -20 dBTarget Attenuation loss 1 dB

Target Scattering Anti-T* -54.1416 dBTarget Scattering Anti-P* -66.1784 dB

Transmit Power 0 dB - 98.068 dBTABLE I

SYSTEM PARAMETERS THAT I WOULD USE TO FIND LAND MINES WITHTHE NIITEK SYSTEM. *ANTI-P STANDS FOR ANTI-PERSONAL AND

ANTI-T SANDS FOR ANTI-TANK

example situations given by Daniels [11] for use my system.To determine the depth to a detected target, the velocity

of propagation, travel time, and the relative permittivity ofthe surrounding material is needed. Electromagnetic wavespropagate at the speed of light in a vacuum, but when theypass through material with electric properties, these waves willtravel with a relative speed given as

νr =c√εr. (7)

Using this relative velocity, νr, and the time delay betweentransmission and reception, we can calculate the depth of theidentified target.

d = νrt

2(8)

A table of my system specifications based on the systemin question is shown in table II.

These are system parameters that could fall within the rangeof those used on the NIITEK system and these values predictSNR values between 150 dB to over 250 dB. The SNR valueswere calculated using

SNR =PtG

2λ2σ

(4π)3R4kT0FBL

, (9)

where Pt is the transmit power, G is the antenna gain, λ is thewavelength, σ is the appropate RCS, R is the desired range,k is the Boltzmann constant, T0 temperature related noise, Fis the system noise, B is the bandwidth, and L is the totalpropagation losses. These large SNR values are undoubtedlyover predicted, but they show that the system is capable ofproducing SNR values large enough to be useful in advanced

5

signal processing. This particular treatment would be good fora situation where the soil is relatively dry and homogeneouslymix throughout our maximum range. If instead we were eval-uating an area with rocks, buried debris, and additional clutter;an entirely different system configuration would be necessary.This is why the NIITEK system has certain parameters suchas the pulse width or transmit power that can be configuredin the field.

III. RADAR RETURN

The GPR system assumes the return measured by theantenna is caused by reflections coming from directly belowthe system. Figure 9 show that the signal spreads from thetransmitter in a cone shape. This pattern allows the radar to seea target without being directly over the device, but it adds anartifact in the return image. When the system detects a targetthat is at an angle with the detector, as seen with position 1in figure 10 for example, the system assumes that the targetis directly below the transmitter, but at a deeper value. As thesystem moves over the target, the prediction of an actual depthimproves until our antenna is found directly over the target.Once directly over the target an accurate depth value can becomputed. An example of this process and its resulting signalresponse are shown in figure 10, and a real image with thissame artifact is shown in figure 11.

Fig. 10. This figure shows an example of a standard GPR as it is movedacross a target of interest. The hyperbolic pattern is caused by the false depthprediction associated with a travel time and velocity of travel.

The data extracted from measuring and processing the returnsignal is used to generate two and three dimensional imagesof targets contained in the soil structure. Unfortunately theseimages are not easily evaluated and the level of expertisenecessary to interpolate the data increases as the signal tonoise ratio (SNR) decreases. Increasing the SNR availablein the GPR signal will result in a higher detection rate andthus increase the longevity of the system and its users, and toimprove SNR the operator can adjust the bandwidth to matchcurrent ground conditions. Another approach to improvingdetection involves the use of algorithms to identify targetsin the data. Algorithms such as feature extraction developedby Quan Zha [9], [10] would simplify the evaluation of dataand can be used to automate the entire process. Automateddetection allows GPR equipped drones to work in a more

Fig. 11. In this figure: a) is an example of the return from a metal mine,and b) is the return from a plastic mine. Notice in these images that the mineis centered under the arc of the hyperbolic pattern created as the GPR wasmoved across the signal. [10]

autonomous capacity, and would allow civilians trained inmine disposal this additional tool without much additionaltraining. The high signal to noise ratio achieved in the Niiteksystem provides an opportunity to develop innovative methodsfor automated detection of land mines.

IV. DEVELOPMENT OF ALGORITHMS TO IMPROVEDETECTION

Thanks to the systems high SNR, a number of differentalgorithms have been developed in recent years to improvea systems ability to accurately identify both metal and non-metal land mines in real time [9], [10]. One method marks thehighest return amplitude measurement of the 416 range binsthat is found closest to the surface from each of the twentyfour antennas. As the system moves forward it compares thehigh return points in three dimensions using a polynomial fitto look for hyperbola shapes. If the high return points forma hyperbola in two different directions centered at a rangepoint, then the system has detected a mine. Figure 12 givesan example of this process in 2 dimensions.

V. CONCLUSION

The presence of land mines and improvised explosive de-vices on the battle field endangers our soldiers and this wouldbe reason enough to peruse this technology, but unlike soldiers,land mines do not go home and they can not differentiatebetween civilian and solider. These lethal devices need to besafely and efficiently removed from current battlefields as wellas those of old. Continuing improvements to GPR specificallydesigned to find these devices will help to achieve this goal.

6

Fig. 12. This is an example of how polynomial fitting could be used to finda land mine. The highest high return points are marked per antenna. Thesepoints are then compared using a polynomial fit algorithm. If they form ahyperbola then a mine is detected.

REFERENCES

[1] N. E. Walsh and W. S. Walsh, “Rehabilitation of landmine victims : theultimate challenge,” Bulletin of the World Health Organization, vol. 81,pp. 665 – 670, 09 2003.

[2] J. Leckebusch, “Ground-penetrating radar: a modern three-dimensionalprospection method,” Archaeological Prospection, vol. 10, no. 4, pp.213–240, 2003. [Online]. Available: http://dx.doi.org/10.1002/arp.211

[3] L. Peters, J. Daniels, and J. Young, “Ground penetrating radar asa subsurface environmental sensing tool,” Proceedings of the IEEE,vol. 82, no. 12, pp. 1802–1822, 1994.

[4] A. Neal, “Ground-penetrating radar and its use in sedimentology:principles, problems and progress,” Earth-Science Reviews,vol. 66, no. 34, pp. 261 – 330, 2004. [Online]. Available:http://www.sciencedirect.com/science/article/pii/S0012825204000054

[5] J. Milsom and A. Eriksen, Ground Penetrating Radar. JohnWiley & Sons, Ltd, 2011, pp. 185–209. [Online]. Available:http://dx.doi.org/10.1002/9780470972311.ch10

[6] S. Lambot, E. Slob, I. van den Bosch, B. Stockbroeckx, and M. Van-clooster, “Modeling of ground-penetrating radar for accurate charac-terization of subsurface electric properties,” Geoscience and RemoteSensing, IEEE Transactions on, vol. 42, no. 11, pp. 2555–2568, 2004.

[7] J. Bourgeois and G. Smith, “A fully three-dimensional simulation ofa ground-penetrating radar: Fdtd theory compared with experiment,”Geoscience and Remote Sensing, IEEE Transactions on, vol. 34, no. 1,pp. 36–44, 1996.

[8] A. K. Benson, “Applications of ground penetrating radar inassessing some geological hazards: examples of groundwatercontamination, faults, cavities,” Journal of Applied Geophysics,vol. 33, no. 13, pp. 177 – 193, 1995. [Online]. Available:http://www.sciencedirect.com/science/article/pii/0926985195900403

[9] Q. Zhu and L. Collins, “Application of feature extraction methods forlandmine detection using the wichmann/niitek ground-penetrating radar,”Geoscience and Remote Sensing, IEEE Transactions on, vol. 43, no. 1,pp. 81–85, 2005.

[10] J. Wilson, P. Gader, W. H. Lee, H. Frigui, and K. Ho, “A large-scale systematic evaluation of algorithms using ground-penetrating radarfor landmine detection and discrimination,” Geoscience and RemoteSensing, IEEE Transactions on, vol. 45, no. 8, pp. 2560–2572, 2007.

[11] D. Daniels and I. of Electrical Engineers, Ground penetratingradar, ser. IEE radar, sonar, navigation, and avionics series.Institution of Electrical Engineers, 2004, no. v. 1. [Online]. Available:http://books.google.com/books?id=7QRTAAAAMAAJ

[12] D. Daniels and T. I. of Electrical Engineers, Surface-PenetratingRadar, ser. Iee Radar, Sonar, Navigation and Avionics Series,6. Institution of Electrical Engineers, 1996. [Online]. Available:http://books.google.com/books?id=hFx5QgAACAAJ

[13] L. Conyers, Ground-Penetrating Radar for Archaeology, ser.Geophysical Methods for Archaeology. AltaMira Press, 2013.[Online]. Available: http://books.google.com/books?id=eVdi3NLvgk4C

[14] L. Conyers and D. Goodman, Ground Penetrating Radar: AnIntroduction for Archaeologists. AltaMira Press, 1997. [Online].Available: http://books.google.com/books?id=hjBmAAAAMAAJ