The detection of landmines by neutron backscattering: Exploring the limits of the technique

ARTICLE IN PRESS

0969-8043/$ - se

doi:10.1016/j.ap

�CorrespondE-mail addr

Applied Radiation and Isotopes 64 (2006) 706–716

www.elsevier.com/locate/apradiso

The detection of landmines by neutron backscattering: Exploringthe limits of the technique

G. Viestia,�, M. Lunardona, G. Nebbiaa, M. Barbuib, M. Cinauserob, G. D’Erasmoc,M. Palombac, A. Pantaleoc, J. ObhoXasd, V. Valkovicd

aDipartimento di Fisica dell’ Universita di Padova and INFN Sezione di Padova,Via Marzolo 8, I-35131 Padova, ItalybINFN, Laboratori Nazionali di Legnaro,Viale dell’ Universita 2, I-35020 Legnaro, Padova, Italy

cDipartimento di Fisica dell’ Universita di Bari and INFN Sezione di Bari,Via E. Orabona 4, I-70126 Bari, ItalydDepartment of Experimental Physics, RuXer Boskovic Institute, Bijenicka c.54, 10000 Zagreb, Croatia

Received 14 June 2005; received in revised form 30 November 2005; accepted 13 December 2005

Abstract

Neutron backscattering (NB) sensors have been proposed for Humanitarian De-mining applications. Recently, a prototype hand-held

system integrating a NB sensor in a metal detector has been developed within the EU-funded DIAMINE Project. The results obtained in

terms of performance of the NB systems and limitations in its use are presented in this work. It is found that the performance of NB

sensors is strongly limited by the presence of the soil moisture and by its small-scale variations. The use of the neutron hit distribution to

reduce false alarms is explored.

r 2006 Elsevier Ltd. All rights reserved.

Keywords: Back-scattered neutrons; Explosive detection; Humanitarian De-mining

1. Introduction

It is estimated that several millions of active mines havebeen deployed so far in about 70 countries, resulting inabout 2000 causalities per month, most of them (85%)being civilians (ICBL, 2003). The anti-personnel (AP) andanti-tank (AT) mines are mostly non-metallic or withminimum metal content, buried at a maximum depth ofabout 20 cm.

Localization and identification of landmines withclassical technologies is a time consuming, expensive andextremely dangerous procedure. In addition, it will take along time to de-mine the suspected areas in the affectedcountries. The mined areas are, indeed, close to thebattlefields, being consequently heavily polluted by metalpieces from the explosions of different ordnances. Thepresence of the metal clutters produces a large number offalse alarms in the metal detectors (MD) commonlyemployed in de-mining operations. Consequently, there is

e front matter r 2006 Elsevier Ltd. All rights reserved.

radiso.2005.12.017

ing author. Tel.: +39049 8277124; fax: +39 049 8277124.

ess: [email protected] (G. Viesti).

a strong need for a technological breakthrough in this fieldto definitively solve the land-mine problem within thedead-line established by the Ottawa treaty.In this respect, a device based on the neutron interroga-

tion can be used as a confirmation detector in the multi-sensor systems. Sensors using thermal or fast neutron-induced g-ray emission have been proposed and used in thepast for Humanitarian De-mining operations (Cinauseroet al., 2004; Lunardon et al., 2004; McFee et al., 1998;Cousins et al., 1998; Kuznetsov et al., 2004; Vourvopoulosand Womble, 2001). One of the major limitations in the useof such sensors is represented by their weight and size thatmakes them not usable as hand-held devices but only invehicle-mounted systems, with a specific limited impact onthe improvement of the Humanitarian De-mining opera-tions.A particular nuclear technique has shown in the past the

capability of being used in hand-held systems: the so-calledneutron back-scattering technique (NBT) (Bom et al.,2004; Borgonovi et al., 2000; Brooks and Buffler, 1999;Brooks and Drosg, 2005; Brooks et al., 2004; Csikai et al.,2004; Datema et al., 2001, 2002; Kiraly et al., 2004).

www.elsevier.com/locate/apradiso

ARTICLE IN PRESSG. Viesti et al. / Applied Radiation and Isotopes 64 (2006) 706–716 707

The principle of the technique is very simple: when a fastneutron source (like a 252Cf radio-isotopic source) is used toirradiate the soil, the yield of thermalized, backwardscattered neutrons depends on the hydrogen content of theirradiated volume. Therefore, to confirm the presence of themine, a NB sensor will verify the presence of anomaloushydrogen concentration in the target point, previouslyidentified by using, for instance, a common tool as theMD. From an operational point of view, it is mandatory tohave a unique hand-held system that integrates a NB probeinside a MD. Moreover, such system has to fulfil a numberof technical and operational requirements dictated by theend user needs. First, the total weight of the system headshould not be larger than 2kg, having dimensions typical ofa standard MD and, in any case, not exceeding an area of20� 30 cm2. In addition, the system should provide clearand simple information, as the usual man–machine interface(MMI) employed in commercial MD. Finally, its cost-effectiveness has to be verified in comparison with theoperation costs of the currently available techniques (i.e.MD as scanning and prodding as confirmation tools).

Such requirements imply a number of technologicalchallenges in the design of individual components, with thegoal of maintaining the performance of the single detectorswhen integrated in a unique sensor head.

The integration of a NB sensor with a MD head has beenthe major task of the DIAMINE project (Viesti et al.,2003). Within this project, the capability and the limits ofthe NBT have been studied in details in laboratoryconditions and using Monte Carlo simulations. The resultsobtained are presented in this paper.

2. General characteristics of the NBT

Within the DIAMINE project, it has been establishedthat sensors based on the NBT are characterized byintrinsic limitations that have to be clearly explored beforefielding such tools. A landmine can, indeed, be detectedonly when the signal due to the hidden object is signifi-cantly larger with respect to the background due to:

(i)
the primary ionizing radiation (fast neutrons and g-rays)emitted from the source that hits directly the detector;
(ii)
the secondary neutrons scattered from the nuclei ofthe soil.
The first kind of background depends on the type andposition of the source employed (typically a 252Cf radio-isotopic source) and on the detector intrinsic sensitivity tog-rays and fast neutrons. In order to optimize the finalperformance of the system, such background can beeffectively minimized by the proper selection of the thermalneutron detector and of the source type and geometry.Particular measurement techniques might also by effective:as an example, a significant improvement in the signal-to-noise ratio has been obtained by using a time-tagged 252Cfsource, as reported in the work of (Craig et al., 2000).

On the other hand, the background associated with thesoil moisture sets an intrinsic limitation to the method. Thelandmine detection is possible, indeed, only when thethermalization capability of the buried mine is significantlygreater than that of the surrounding soil, the latter beingessentially due to the soil moisture. Since the thermaliza-tion capability is mainly determined by the hydrogencontent inside a given material, the condition for thedetection lies in the hydrogen density difference betweenthe landmine and the soil. In this respect, as discussed indetail in (ObhoXas et al., 2004), each type of mine can becharacterized by a specific average H density, determinednot only by the explosive charge but also by the externalcase, made often from plastic materials. Consequently, it ispossible to define for each type of mine a critical value ofthe soil moisture for which the detection is impossible,when the contrast between the buried mine and the baresoil is absent.A specific analysis of this problem, reported in (ObhoXas

et al., 2004), brought us to the conclusion that the best useof the NBT is in countries where the soil moisture isgenerally lower than about 10% in weight, so that a largenumber of land-mines can be safely detected. Thiscondition is expected to be valid in case of aridic soils thatare characteristic of countries as Afghanistan, Ethiopia,Eritrea, Egypt, and Somalia, where a large part of the land-mine problem is localized. Also in such countries, however,a detailed determination of the soil moisture is necessary,when planning the use the NBT.As an example, we have performed moisture measure-

ments of soil samples taken in some Afghanistan locationsduring the summer 2002. Some samples exhibited a verylow soil moisture value (o1% in weight), documenting anideal condition for the use of the NBT. On the contrary,other samples exhibit a higher soil moisture values (about8% in weight), which is close to the suggested 10% limit forthe use of the NBT. This seems to demonstrate that also inthose countries generically defined as ‘‘aridic’’, the use ofthe NB sensor might be limited to some type of soils and/orspecifically dry seasons and/or locations. It is worthmentioning that the knowledge and the monitoring of thesoil characteristics during de-mining operations seems to bea relevant issue for a number of tools, including thosebased on electro-magnetic induction (Das et al., 2002).This means that the hydrogen content of the soil in a

given mine filed needs to be monitored, when the NBT isemployed. This can be easily obtained, as an example, bymonitoring the absolute ‘‘background’’ count rate in acalibrated NB detector, when a well-defined geometry isused. The problems related to the ‘‘background’’ due to thesoil moisture will be further discussed in Section 6.

3. NB sensor description

The NB detector employed in this work makes use of alarge area (20� 20 cm2) multi-wire proportional counter(MWPC) as a neutron detector (Fioretto et al., 2004). The

ARTICLE IN PRESS

Fig. 1. Yield of the backscattered neutrons as a function of the source-

sample distance for a 1.3 kg sample (a) and as a function of the sample

weight placed at a source-soil distance of 10 cm (b). High density poly-

ethylene (HDP) samples are used. The background due to direct radiation

hitting the detector, as measured in a sample out run, has been subtracted.

An exponential function is also shown in panel (a), as suggested in (Csikai

et al., 2004). The line in panel (b) is a linear fit to the experimental data.

G. Viesti et al. / Applied Radiation and Isotopes 64 (2006) 706–716708

MWPC detector consists basically of 3 parallel electrodes:a plane of wires as anode and the two cathodes coated witha 3m thick B4C layers, 97% enriched in 10B. The anodewires are grouped to obtain 10 independent portionshaving an area of 2� 20 cm2 each. In this system, theefficiency for low-energy neutrons is determined by theconversion via the 10B(n,a) reaction, which is estimatedto be about 16% (Arnaldi et al., 2003). The detectorelectrodes are light (700 g) and have been specificallydesigned to minimize the metal content, so that theperformance of a standard MD is not lowered when it isintegrated with the NB sensor. During laboratory testswith the NB sensor working in stand alone mode, theMWPC electrodes were enclosed in a gas-tight, sealed G-10box and operated at atmospheric pressure with a mixtureof Ar (85%) and CO2 (15%). A steady gas charge willallow operation during an 8 h shift without showing asignificant deterioration of the MWPC signal size.

4. Laboratory tests

The laboratory tests have been performed in two steps.The first step was devoted to the study and optimization ofthe global features of the detector, the second one to theverification of the response of the sensor in a simulation ofthe in-field conditions.

In the first step, a weak, sealed 252Cf source (5�104 neutron/s) was placed on the detector G10 box,centered with respect to the MWPC active area. Samplesof high density polyethylene (HDP) of different weightwere placed in air at different distances from the source, tostudy the sensor response as a function of the hydrogenquantity and the measuring geometry.

Typical results are reported in Fig. 1 in terms of netcounts obtained by subtracting a sample-out measurementfrom the sample-in run. As expected, it is found that thedetector response depends mainly on the solid angle underwhich the sample is irradiated and is proportional to thequantity of hydrogen present in the sample. Moreover, it isseen that, as suggested in (Csikai et al., 2004) the data inFig. 1(a) are reproduced by using an simple exponentialfunction.

In these experimental conditions, the background rate,which is due only to the primary radiation (g-rays andneutrons) emitted from the 252Cf source, depends on thesource-detector distance, while the signal rate depends onthe number of fast neutron hitting the sample. Conse-quently, the optimization of the signal-to-noise ratio isreached when the source is located close to the sample andfar from the detector. As an example, in our geometry thesignal-to-noise ratio (S�N)/N measured with (S) andwithout (N) the HDP sample is ðS �NÞ=N ¼ 1 when thesource is close to the detector and about 0.5 kg of HPD areplaced in air at a distance of 10 cm from the detector. Thisratio improves up to the value ðS �NÞ=N ¼ 6 when thedetector-sample distance is still 10 cm but the source is nowpositioned in contact with the sample.

Taking into account the above results in planning the useof a hand-held NB sensor with a typical stand-off distanceof about 10 cm, we can define a ‘‘standard geometry’’ whenthe source is in contact with the lower detector surface anda ‘‘confirmation geometry’’ when the source is loweredfrom its standard position to the soil surface. Asdemonstrated by laboratory tests, the best signal-to-noiseratio is obtained in the latter condition.In a realistic de-mining scenario, the ‘‘confirmation

geometry’’ might be used only after the suspect point hasbeen identified by a preliminary MD scan in which thesource is maintained in its ‘‘standard geometry’’. It is worthmentioning that the weight of a sealed radioactive source asthe one used in our sensor is limited to a few grams. It istherefore possible to design a source driving system thatwould place the source in contact with the soil withouttriggering the buried explosive device.

ARTICLE IN PRESS

Table 1

Experimental net count rate versus burial depth of the dummy APM, as

measured in laboratory conditions (12 cm stand off distance)

Burial depth (cm) Experimental net count rate (counts/s)

1.7 2.9470.06

5 0.9670.02

10 0.4870.01

15 0.1970.01

Statistical uncertainties are reported. For details see the text.

G. Viesti et al. / Applied Radiation and Isotopes 64 (2006) 706–716 709

In the second step, laboratory tests were performed usinga dummy anti-personnel landmine (APM). The dummylandmine is a right cylinder having 80mm diameter and34mm height. The external container is made by acrylicmaterial having a weight of 95 g. The TNT simulantconsists of powdered cyanuric acid (C3H3N3O3), crystallineoxalic acid (C2H2O4 � 2H2O) and powdered graphite toobtain a mixture which simulates the elemental composi-tion of TNT: [H:C:N:O] ¼ [5:7:3:6]. The weight of thefilling powder is 100 g. The dummy landmine wasmanufactured at the Cape Town University (ZA) anddelivered to the participants to the IAEA CoordinatedResearch Program on Humanitarian De-mining. TheTNT simulant composition was provided by the WesternKentucky University (USA).

During the laboratory measurements, the NBT sensorwas placed on a large soil box containing sand (with ameasured moisture of 3% in weight) in which the dummyAPM was buried at different depths.

Typical results are shown in Fig. 2 in terms of the netcount rate (S�N) versus the dummy APM depth, inconfirmation geometry and stand off distance of about12 cm. Here S refers to the irradiation with the buriedlandmine, whereas N refers to the irradiation of the baresoil. Data are also reported in numerical form in Table 1. Itappears that the measured count rate is rather small (3–0.5counts/s) because of the weak 252Cf source (5� 104 neutron/s) employed and the efficiency of our NB detector. However,the detection of an APM buried up to a depth of 10–15 cmin an almost dry soil seems in principle possible with a60–120 s irradiation by using a stronger source (for instancea 2.5� 105 neutron/s) in ‘‘confirmation geometry’’.

It is worth mentioning that the statistical accuracy of themeasurement, that will determine the detection probability,

Fig. 2. Experimental net count rate versus burial depth of the dummy

APM, as measured in laboratory conditions (12 cm stand off distance).

The background due to direct radiation hitting the detector, as measured

in a sample out run, has been subtracted. Statistical uncertainties are

within the marker size. An exponential function is also shown, as

suggested in (Csikai et al., 2004).

depends on the type of mine as well as on the absolutevalue of the background. The latter is directly related to thelevel of moisture in the soil. Since it is not easy to vary in acontrolled and reproducible way the moisture level in thelaboratory soil box, the experimental data of Fig. 2 havebeen used to validate the Monte Carlo calculationspresented in the next section. Results from extensiveMonte Carlo simulations have been employed, in turn,to study the viability of the technique by predictingthe performance of the NB sensor in a large number ofdifferent conditions.Finally, a further group of measurements has been

devoted to the study of the hit distributions (i.e. thenumber of counts as a function of the position in theMWPC) to verify the possibility of obtaining informationabout location and shape of the buried object. A typicalbackground subtracted hit distribution for the dummyAPM buried at 5 cm depth in a position corresponding tothe centre of the MWPC detector (i.e.10 cm) is shown inFig. 3. Gaussian fit to the experimental data provides thecentroid (12.272.0 cm) and the width (21.678.0 cm[FWHM]) values of the experimental distribution. It isfound that the centroid of the distribution determinescorrectly the mine position. On the contrary, the widthvalue is very large compared with the 8 cm diameter of thedummy APM. Similar results have been obtained also forother depth values up to 15 cm. It is particularly interestingto note that the width of the hit distribution was found tobe independent from the depth and the size of the buriedobject. This fact was experimentally studied by using notonly the dummy land mine but also HPD samples.On the other hand, the ‘‘background’’ hit distribution

measured in case of the bare soil, shown in Fig. 4, is quitedifferent from the one obtained for the buried landmine,being more flat. Consequently, even if the hit distributionmeasured with the buried APM does not show acorrelation with the real size of the hidden object, itappears that a bump-like structure with a well-definedwidth might provide evidence for the presence in the soilof a localized hydrogen anomaly associated with thelandmine. This feature seems to be very interesting indiscriminating landmine signals from false alarms due to adispersed soil moisture, as it will be further discussed in thenext sections with the help, also in this case, of MonteCarlo calculations.

ARTICLE IN PRESS

Fig. 4. Experimental hit distribution of the bare soil. Predictions from

Monte Carlo simulations are also reported (line). For details see the text.

Fig. 3. Experimental hit distribution of the dummy land-mine buried at

different 5 cm depths in a position corresponding to the centre of the

MWPC detector (i.e. 10 cm). The background due to the bare soil has been

subtracted. The line is a Gaussian fit to the experimental data.


5. Monte Carlo simulations

An extensive Monte Carlo simulation campaign hasbeen performed to determine the performance of the NBsystem as a function of type of buried land-mine, detectorstand-off distance, measuring geometry, mine burial depthand soil moisture.

The production and transport of radiation has beensimulated by means of the standard program GEANT 3.21(GEANT, 1994), interfaced to the MICAP subroutine inwhich the cross-sections data from ENDF-B VI compila-tion are used (Palomba et al., 2003). Specifically for thesimulation reported here, a routine has been implementedfor the evaluation of the time of detection of neutrons withrespect to the beginning of the simulation, after thestatistical neutron emission from the source and transportthrough the set-up. A detection time coordinate is recorded

for all detected events. Consequently, a time-dependentdata acquisition can be fully simulated.In the Monte Carlo simulation, a large soil sample was

considered with the soil density set at the value 1.5 g/cm3.The soil composition was taken from (Chilton et al., 1984).The soil moisture will be indicated in the following as

percent water content in weight with respect to the totalweight of the dry soil plus water. The water is supposed tobe uniformly dispersed in the soil volume.The landmines were simulated by assuming generally an

external plastic case made by poly-ethylene and a TNTcharge. The adopted poly-ethylene chemical composition isCnH2n with density 0.92 g cm�3, whereas the one of TNT isC7H5N3O6 with density 1.654 g cm�3. The geometricalparameters of landmines were taken from current data-base. As an example, the small APM PMA2 was assumedto be a thin (0.4 cm) poly-ethylene cylinder having 6.2 cmexternal diameter and 3.5 cm height, filled with 100 g TNT.The rather large TMA-3 AT mine was assumed to be aright cylinder (25 cm diameter, 8 cm height) filled with6.5 kg of TNT without outer plastic case.Simulations were performed as a function of:

(1)
the mine burial depth (5, 10, 15 and 20 cm below thesoil level);
(2)
the soil humidity (0% (dry soil), 5%, 10%, 15%, 20%in weight; in some selected cases it was also set at 2.5%,7.5% and 12.5% in weight);
(3)
the stand off distance, defined by the position of theboron converter placed at 5, 10 and 15 cm above thesoil level;
(4)
the type of buried mine: APM (PMA1, PMA2 andPMA3) and ATM (TMA3) were studied. Simulationruns without buried landmines were used to estimatethe soil background.
The neutron source was a standard sealed 252Cf source(with an external case having 8mm diameter and 10mmlength) with an energy spectrum simulated by using a Max-well distribution with nuclear temperature T ¼ 1:42MeV.The emission from the source was assumed to be isotropic.The source intensity was assumed to be 2.3� 105 neutron/s.When the neutrons are hitting the detector, the 10B (n,a)7Lireaction inside B4C converter produces a–7Li-chargedrecoils that are tracked inside the B4C layer. Each chargedparticle exiting the converter is recorded as a valid count,assuming that the MWPC has 100% efficiency for chargedparticles independently of their kinetic energy.More than 200 different calculations have been per-

formed using both standard and confirmation geometries,as previously defined.

5.1. Simulation results

In a first step, a quantitative comparison between theMonte Carlo calculations fully reproducing the experi-mental conditions and the laboratory measurements with


the dummy APM was performed. It is found that theexperimental data from Fig. 2 are rather well reproducedby the Monte Carlo prediction, with an average deviationof 20% between experiment and simulation. This result wastaken as a validation test of the Monte Carlo simulations.

Secondly, the two different geometries (standard andconfirmation) were compared, to confirm the findings ofthe laboratory tests. As an example, Monte Carlo predic-tions in term of net counts (S�N) are reported in Fig. 5 forthe standard geometry in case of an AT TMA-3 and anAP PMA-2 mine for a 60 s long irradiation and a 10 cmstand-off distance. Dry soil and 5% soil moisture wereconsidered.

In such conditions, the signal from the small PMA-2 mineseems to be generally very small. A positive signal outside thestatistical uncertainties is evidenced up to 20cm burial depthonly in dry soil (Fig. 5(a)). Already in the case of 5% soilmoisture (Fig. 5(c)), the signal uncertainties are above thebackground level only for APM buried up to about 10 cmdepth. For larger soil moisture values, the Monte Carlosimulations predict that only shallow APMs are detectable.

The situation improves when a bigger TMA-3mine isconsidered, because of its larger hydrogen content asshown in Fig. 5(b) (dry soil) and Fig. 5(d) (5% soilmoisture). However, also in the latter case limitationsappear for mines buried at 20 cm depth.

Results obtained by using the confirmation geometry areshown in Fig. 6. In this case, the large TMA-3 seems to bedetectable in dry (Fig. 6(b)) and in 5% (Fig. 6(d)) moisture

Fig. 5. Results from Monte Carlo simulations: net counts (S�N) detected for a

row) buried at different depth in dry soil (a) and 5% soil moisture (c). The low

exponential function is also shown, as suggested in (Csikai et al., 2004).

soils up to 20 cm depths. Limitations appear when the 10%soil moisture case is considered (not shown) for minesburied at 20 cm depth. An improvement is also seen in thecase of the PMA-2 mine that is predicted now to bedetectable up to 15 cm depth.The problems due to the presence of the soil moisture in

the detection of AP and AT mines have been explored indetails. It is worth mentioning that the signal due to thehidden landmine is small compared with the backgrounddue to the moisturized soil. In a 60 s irradiation, the signaldue to the presence of a PMA-2 mine is of the order of200–300 counts for soil moisture values up to 10%,whereas the corresponding net counts due to a TMA-3AT mine are of the order of 1000–2000 in the sameconditions. Such signals have to be compared with thepredicted increase of the background due to the soilmoisture of about 700 counts per 1% variation. This meansthat small, local variation of the water inside the soil mightmask the signal due to the landmine, especially in the caseof small AP devices. This point is discussed in detail inSection 6.

5.2. Simulated hit distributions

As already pointed out in Section 4, the experimental hitdistribution for the bare soil and for the buried landminesare somewhat different. Consequently, also the spatialdistribution of the neutron counts has been studied in anextensive way by Monte Carlo method.

60 s irradiation in case of ‘‘standard geometry’’ and a PMA-2 mine (upper

er row refers to a TMA-2mine: dry soil (b) and 5% soil moisture (d). An

ARTICLE IN PRESS

Fig. 6. Results from Monte Carlo simulations: net counts (S�N) detected for a 60 s irradiation in case of ‘‘confirmation geometry’’ and a PMA-2 mine

(upper row) buried at different depth in dry soil (a) and 5% soil moisture (c). The lower row refers to a TMA-2mine: dry soil (b) and 5% soil moisture (d).

An exponential function is also shown, as suggested in (Csikai et al., 2004).

Fig. 7. Comparison of the experimental hit distribution of the dummy

land-mine buried at 5 cm depths in a position corresponding to 10 cm with

the results from Monte Carlo simulations (line).


A direct comparison between results from laboratorytests and Monte Carlo predictions is reported in Figs. 4 and7. It seems that the Monte Carlo predictions account forthe flat distribution due to the bare soil (shown in Fig. 4)and for the bump-like structure due to the hidden object(shown in Fig. 7).

Moreover, detailed calculations performed for the PMA-2 and TMA-3mines at different depths in dry soildemonstrated that the width of the background subtractedhit distribution is very weakly dependent on the depth ofthe mine and even on the size of the buried object, inagreement with the experimental findings reported inSection 4.

The shape of the background hit distributions due to thebare soil was also studied in details as a function of the soilmoisture. Typical distributions for dry, 5%,10% and 15%soil moisture are reported in Fig. 8. Simulation resultsexhibit a well-defined transition from the flat distribution,characteristic of the dry soil, to a more pronounced bellshapes at higher soil moisture values. To extract quanti-tative parameters, the hit distributions were fitted by usinggaussian functions determining the width of the bump. Theextracted FWHM values decrease by increasing the soilmoisture, being the minimum value of 3772 cm, deter-mined for the 15% case. Consequently, it appears that thewidth values characterizing the bare soil distributions aremuch larger compared with the ones measured in the caseof buried landmines. This fact might be used in determin-ing the detection of a hidden landmine, in the assumptionthat the distribution of the moisture in the real soil might

result to be homogeneous as assumed in the Monte Carlosimulation.

6. Discussion

The introduction of new techniques in HumanitarianDe-mining operations relies on a number of constrains ofdifferent nature. It is indeed required that new tools have todetect ATM and APM buried up to a depth of 20 cm. Theperformances of the new tools need to be qualified in terms

ARTICLE IN PRESS

Fig. 8. Hit distributions predicted by Monte Carlo calculations (confirmation geometry) for different soil moisture values: dry soil (a) and 5% (b), 10%

(c), 15% (d) soil moisture. Gaussian fits are also reported.

G. Viesti et al. / Applied Radiation and Isotopes 64 (2006) 706–716 713

of detection efficiency and false alarm rate, so that theimprovement in operation efficiency can be quantified withrespect to the current situation. Once the performance ofthe system has been ascertained, other issues as the cost-effectiveness, the operational and human aspects becamerelevant.

In this work, we are discussing essentially the first point,i.e. the capability of the NBT system to detect buried objects.The prototype developed in our project was specificallyoriented to satisfy the operational requirements of Humani-tarian De-mining that influenced heavily our specific detectordesign. However, several points in the following discussionapply generally to all types of NB sensors.

The first general problem is the limited impact that suchtechniques might have on the Humanitarian De-mining,due to the a priori requirement of using NB devices only inaridic countries, having soil moisture below the 10% limit.If this point is accepted, the second problem is thecapability of sensors based on the back-scattering techni-ques of detecting landmines up to a depth of 20 cm. For agiven neutron source, the distribution of the thermalizedneutrons inside the soil decrease exponentially with thedepth, as demonstrated in (Csikai et al., 2004). Byparameterizing the exponential decrease of the signal fromthe buried object by using the function S ¼ A� exp�ðBxÞ,where x is the burial depth, it is possible to fit theexperimental data and predict the size of the signal atdifferent depth. As an example, our experimental data inFig. 2 are reproduced by using a dumping parameterB ¼ 0:29. Similar numerical results (B ¼ 0:2920:36) havebeen obtained by the Debrecen group (see Csikai et al.,

2004 and references therein) in case of their ‘‘twindetectors’’ operated with 252Cf or Pu-Be sources (withoutany moderator or Cd cover). The dumping parameterdepends on the specific design of the sensor and on theneutron source used.As previously discussed, the capability of detecting

landmines in laboratory condition depends on the signal-to-background ratio. Results reported for some of theproposed back-scattering systems are presented in Table 2.It seems that in controlled laboratory conditions theproposed sensors are close to the requirement of detectingboth APM and ATM in dry sand up to a burial depth of20 cm. However, their performance decreases as soon asthe background due to the bare soil increases for higher soilmoisture values.This observation opens a further problem which is

related to the relatively weak signal from the buried object,when compared to the noise generated by the bare soil inrealistic field conditions. This is a general problem and doesnot depend, in large part, on our specific sensor design, i.e.on the detection efficiency of the MWPC neutron detectoror on the activity of the sealed neutron source. As discussedin Section 5.1, the signal from the hidden objects might becompletely masked by fluctuations of the background dueto variations of the soil moisture in the vicinity of thesuspect point. As an example, it might happen that thebackground is measured in a region of the minefield havingmoisture higher than the one characterizing the point inwhich the mine is buried. This would result in the reductionof the net counts in the background-subtracted spectrum,giving rise to possible false negatives.

ARTICLE IN PRESS

Table 2

Performance of backscattering systems in detecting landmines at different depths in laboratory condition

Detector Ref. Test bed Moisture APM Limiting

depth (cm)

ATM Limiting

depth (cm)

HIDAD-D Brooks et al. (2005) Sand Dry 23

HIDAD-D Brooks et al. (2005) Sand 12% w/w 7

SHELL detector Datema et al. (2002) Sand 4% w/w 10 20

DIAMINE This work Sand 3% w/w 15 20

Results are obtained by using laboratory tests and Monte Carlo simulations.


From a practical point of view, such small-scalevariations of the soil moisture have been clearly seen insome minefield locations in Croatia, where the soil moisturewas accurately measured during the DIAMINE project(ObhoXas et al., 2004). In some cases, the small-scalevariation of the soil moisture was measured in samplesof about 100� 100 cm2 area for pixel size of about10� 10 cm2. As reported in (ObhoXas et al., 2004) the fieldmeasurements demonstrate that the soil moisture is astrongly varying function, with differences of about a factor2 for points having relative distance of less than 100 cm.

The effects due to the soil moisture variations wereaddressed by using the experimental soil moisture datainside specific Monte Carlo simulations. Typical AP andAT mines were assumed to be buried at a given depthin positions randomly selected inside a 100 cm� 100 cmsample area. The point at which the background issupposed to be measured was also selected randomly inthe same sample area with the condition of being at least50 cm from the mine position. Then, the soil moisturevalues corresponding at the mine and at the backgroundpositions were determined from the experimental field dataand the corresponding counts in our NB sensor weredetermined by interpolating from a data base constructedby using the results from the Monte Carlo simulationsdescribed in Section 5. In this way the net signal wasobtained by subtracting the two count values and the‘‘detection’’ of the mine is determined by the statisticalsignificance of the net counts. This procedure allowsextracting the probability of false alarms (FA) associatedwith the selected soil moisture distribution, the type ofmine and the burial depth.

The result of this exercise is rather clear. In the case of aPMA2 buried at 5 cm depth in a soil for which the moisturedistribution is assumed as the one measured in a costallocation near Zadar (Croatia) with an average moisture ofabout 5% in weight, the probability of FA is evaluated tobe about 6%. This value increases up to 24% for the samemine buried at 10 cm depth. The FA probability becomesacceptable only if we consider soil with average moisture ofabout 1% in weight, obtained by scaling down themeasured distribution. The FA probability is less impor-tant in the case of ATM, since its signal is much larger. Asan example, this probability is not larger than 1%considering a TMA-3mine buried up to 20 cm when theaverage soil moisture is about 2.5% in weight.

The above results are certainly raising questions aboutthe possibility of fielding a device that determine thedetection of the mine by using a simple procedure for thebackground subtraction, when small APMs have to bedetected.A possible solution for the reduction of the FA

probability due to the soil moisture variation might takeadvantage from the information contained in the hitdistribution, as discussed in previous section. However,field measurements of the hit distribution associated withthe small-scale variation of the soil moisture are needed todemonstrate definitively the possibility of discriminatingagainst false alarms.

7. Conclusions

A new hand-held landmine sensor was studied within theDIAMINE project, based on the integration of a NBsensor with a MD. During the project, important resultshave been achieved from the hardware side. The sensorhead has been designed and major components have beentested in laboratory conditions by using dummy landmines, demonstrating the possibility to detect APMs whenthe sensor is used in the so-called confirmation mode.In the meantime, an investigation of the general features

of the NBT has been performed by using a combination oflaboratory measurements and Monte Carlo simulations.The results foresee the possibility of APM detection forburial depths less than 10–15 cm and soil moisture less than10% in weight. The detection possibility of AP land-minesup to the limit of 20 cm depth needs to be furtherinvestigated. Limitation of the NBT sensor use to a well-defined range of soil moisture values (below 10% in weight)seems to be an intrinsic feature of the technique.Furthermore, specific soil moisture measurements per-

formed within the DIAMINE project in mined areas inBalkans (ObhoXas et al., 2004), have revealed that inseveral type of soils the moisture exhibits a large localvariability. Such effects need to be further studied, since thelocal variation of the soil moisture might cover or mimicthe signal of the buried land mine.Results from laboratory measurements indicate that the

hit distribution of the 20� 20 cm2 sensor of the DIAMINEhand-held prototype will not provide quantitative informa-tion on the shape of buried object. Nevertheless, it seemsthat the analysis of the hit distribution width can be used to


discriminate between local or diffuse hydrogen concentra-tions in the soil. This fact might be used to reduce the falsealarm probability due to the small-scale soil moisturevariation. Field tests are needed to prove such possibility.

The use of the hit distribution information requires,however, an increase in the acquired statistics that can beobtained either by longer measuring times or improving theperformance of the existing prototypes in terms of both thedetector efficiency and neutron source activity. In the lattercase the use of higher activity neutron source althoughtechnically feasible, might rise further problems in thepractical use of a hand-held NB sensor in HumanitarianDe-mining operations because of the related increase incost, weight and radiation hazard.

Acknowledgments

This work was supported and performed in part withinthe project DIAMINE, funded by European Union underthe contract IST-2000-25237. The DIAMINE Consortiumintegrated research bodies as the Italian Istituto Nazionaledi Fisica Nucleare, the Institute of Physics of the SlovakAcademy of Science, the JRC-IRMM Geel, Belgium, withthe companies LABEN SpA, CAEN SpA and NeuriCamSpA from Italy and Plein&Baus GMBH and VALLONGMBH from Germany. LABEN was in charge of theproject coordination. This work was also part of theCoordinated Research Project ‘‘Application of NuclearTechniques to Anti-personnel Landmines Identification’’of the International Atomic Energy Agency under theResearch Agreement ITA 10985. Thanks are due to Prof.F. Brooks of the Cape Town University for providing thedummy APM in the framework of the IAEA CRP. We areindebted to the INTERSOS NGO for providing Afghani-stan soil samples. Several useful discussion with Prof. J.Csikai are acknowledged. We are also indebted to Prof. J.Csikai for providing us unpublished data.

References

Arnaldi, R., Colla, A., De Marco, N., Ferretti, A., Gallio, M., Musso, A.,

Oppedisano, C., Piccotti, A., Poggio, F., Scomparin, F., Sigaudo, F.,

Travaglia, G., Vercellin, E., Cortese, P., Dellacasa, G., 2003, RPC as

thermal neutron detector for Humanitarian De-mining, in detecting

environmental, industrial and biomedical signals. In: Proceedings of

the International Workshop, Bari, Italy, 2002, World Scientific,

Singapore, 2003, pp. 166–180.

Borgonovi, G.M., Ginaven, R.O., V. J. Orphan, V.J., 2000, Landmines

and unexploded ordnance detection, in a remotely controlled multi-

sensor platform for Humanitarian De-mining, Report of the Advisory

Group Meeting held 3–7 April 2000 at the IAEA Headquarters,

Vienna, Austria, IAEA publication IAEA/PS/AG-1093.

Bom, V.R., Datema, C.P., van Eijk, C.W.E., 2004. The status of the Delft

University Neutron Backscatter Landmine Detector (DUNBLAD).

Appl. Radiat. Isotopes 61, 21–25.

Brooks, F.D., Buffler, A., 1999. Detection of plastic land mines by

neutron backscattering, In: Sixth International Conference on

Applications of Neutron Science, June 1999, Crete, Greece (unpub-

lished).

Brooks, F.D., Drosg, M., 2005. The HYDAD-D anti-personnel landmine

detector. Appl. Radiat. Isotopes 63, 565–574.

Brooks, F.D., Drosg, M., Buffler, A., Allie, M.S., 2004. Detection of anti-

personnel landmines by neutron scattering and attenuation. Appl.

Radiat. Isotopes 61, 27–34.

Chilton, A.B., Shultis, J.K., Faw, R.E., 1984. Principles of Radiation

Shielding. Prentice-Hall, Englewood Cliffs, NJ.

Cinausero, M., Lunardon, M., Nebbia, G., Pesente, S., Viesti, G.,

Filippini, V., 2004. Development of a thermal neutron sensor for

Humanitarian Demining. Appl. Radiat. Isotopes 61, 59–66.

Cousins, T., Jones, T.A., Brisson, J.R., McFee, J.E., Jamieson, T.J.,

Waller, E.J., Lemay, F.J., Ing, H., Clifford, E.T., Selkirk, E.B., 1998.

The development of a thermal neutron activation (TNA) system as a

confirmatory non-metallic land mine detector. J. Radioanal. Nucl.

Chem. 235, 53–58.

Craig, R.A., Peurrung, J., Stromswold, D.C., 2000. Mine detection using

timed neutron moderation. The UXO/Countermine Forum, April

2000 (see http://availabletechnologies.pnl.gov/images/UXOcon-

f2a.pdf).

Csikai, J., Doczi, R., Kiraly, B., 2004. Investigations on landmine

detection by neutron-based techniques. Appl. Radiat. Isotopes 61,

11–20.

Das, Y., McFee, J.E., Cross. G., 2002. Soil properties database for

humanitarian demining: a proposed initiative. 17th World Congress of

Soil Sciences, 14–21 August 2002, Bangkok, Thailand.

Datema, C.P., Bom, V.R., van Eijk, C.W.E., 2001. Landmine detection

with the neutron backscattering method, IEEE Trans. Nucl. Sci. 48,

1087–1091.

Datema, C.P., Bom, V.R., van Eijk, C.W.E., 2002. Experimental results

and Monte Carlo simulations of a landmine localization device using

the neutron backscattering method. Nucl. Instrum. Methods A 488,

441–450.

Fioretto, E., Barbui, M., Giangrandi, S., Cinausero, M., Prete, G.,

Nebbia, G., Viesti, G., 2004. Neutron back-scattering sensor for the

detection of land mines. Nucl. Instrum. Methods B 213, 457–459.

GEANT, 1994. Detector Description and Simulation Tool. CERN,

Program Library Long Write-up W5013, also http://wwwinfo.cern.ch/

asdoc/geant_html3/geantall.html.

ICBL (The International Campaign to Ban Land-Mines), 2003. The

landmine monitor report 2003: toward a mine-free world. http://

www.icbl.org/lm/2003/.

Kiraly, B., Sanami, T., Doczi, R., Csikai, J., 2004. Detection of explosives

and illicit drugs using neutrons. Nucl. Instrum. Methods B 213,

452–456.

Kuznetsov, A.V., Evsenin, A.V., Gorshkov, IYu., Osetrov, O.I., Vakhtin,

D.N., 2004. Detection of buried explosives using portable neutron

sources with nanosecond timing. Appl. Radiat. Isotopes 61, 51–57.

Lunardon, M., Nebbia, G., Pesente, S., Viesti, G., Barbui, M., Cinausero,

M., D’Erasmo, G., Palomba, M., Pantaleo, A., Filippini, V., 2004.

Detection of landmines by using 14MeV neutron tagged beams. Appl.

Radiat. Isotopes 61, 43–50.

McFee, J., Aitken, V., Chesney, R., Das, Y., Russell, K., 1998. A

multisensor, vehicle-mounted, tele-operated mine detector with data

fusion. In: Proceedings of the 1998 Conference on Detection and

Remediation Technologies for Mines and Minelike Targets III,

Orlando, USA, SPIE 3392.

Palomba, M., D’Erasmo, G., Pantaleo, A., 2003. A The Monte Carlo code

CSSE for the simulation of realistic thermal neutron sensors for

Humanitarian De-mining. Nucl. Instrum. Methods A 498, 384–396.

ObhoXas, J., Sudac, D., NaX, K., Valkovic, V., Nebbia, G., Viesti, G.,

2004. The soil moisture and its relevance to the landmine detection by

neutron backscattering technique. Nucl. Instrum. Methods B 213,

445–451.

Vourvopoulos, G., Womble, P.C., 2001. Pulsed fast/thermal neutron

analysis: a technique for explosives detection. Talanta 54, 459–468 and

references therein.

Viesti, G., Lunardon ,M., Nebbia, G., Barbui, M., Cinausero, M.,

Fioretto, E., Prete, G., Pantaleo, A., D’Erasmo, G., Palomba, M.,

http://availabletechnologies.pnl.gov/images/UXOconf2a.pdf

http://availabletechnologies.pnl.gov/images/UXOconf2a.pdf

http://wwwinfo.cern.ch/asdoc/geant_html3/geantall.html

http://wwwinfo.cern.ch/asdoc/geant_html3/geantall.html

http://www.icbl.org/lm/2003/

http://www.icbl.org/lm/2003/

ARTICLE IN PRESSG. Viesti et al. / Applied Radiation and Isotopes 64 (2006) 706–716716

Abbrescia, M., Iaselli, G., Loddo, F., Paticchio, V., Ranieri, A.,

Trentadue, R., Colla, A., Musso, A., Piccotti, A., Poggio, F.,

Dellacasa, G., Lazzizzera, I., Lecca, P., ObhoXas, J., Sudac, D.,

NaX, K., Valkovic, V., Brusegan, A., Lobo, G., Jurkovic, M., Hlavac,

S., Catarsi, F., Franchi, G., Chianella, M.A., Galimberti, D., Pavesi,

L., Zorat, A., Koester, A., Plein ,M., Merz, A., Schneider, H., Vallon

,G., 2003. The DIAMINE landmine detection system. In: Duggan,

J.L., Morgan, I.L. (Eds.), 17th International Conference on Applica-

tion of Accelerator in Research and Industry, November 2002,

Denton, USA, AIP CP680, pp. 924–927.

The detection of landmines by neutron backscattering: Exploring the limits of the technique

Documents