Magnetic Gradient and Neutron Backscattering Fusion for landmine Detection

Mohamed Elkattan Ahmed Salem A.S. Osman Aladin Kamel Hadia El-Hennawy Nuclear Materials GETECH, Atomic Energy Ain shams Univ. Ain shams Univ. Authority University of Leeds Authority Electronics depart. Electronics depart. Cairo, Egypt Leeds, UK Cairo, Egypt Cairo, Egypt Cairo, Egypt

emtiazegf@hotmail.com

Abstract—demining of abandoned minefields is a serious environmental problem for many countries worldwide. The main problem in landmine clearance is how to discriminate between a landmine, and other buried clutter. A single sensor for demining may have detection rate which is too low, or false alarm rate which is too high. Thus, combination of two or more complementary sensors will improve the detection performance. In this paper, we present an algorithm for the fusion between magnetic gradiometer and neutron backscattering data for landmine detection. Whereas magnetic gradiometer can distinguish the metal content of a landmine, nuclear backscattering technique focuses on detecting the explosive content of a landmine. Furthermore, the proposed algorithm was implemented on a microcontroller chip, to enhance the deminer decision toward any suspected object in the minefield.

Keywords-neutron backscattering; magnetic gradient; landmine; microcontroller, Decision-level fusion

I. INTRODUCTION Landmines are one of the most hazardous reminders in a

former area of armed conflict and often pose a deadly threat to the civil population in many countries. The extreme risk posed by mines implies that a mine sensor should have essentially no missed detections. This requirement and the clutter-rich environment in which mine sensors operates lead to high false alarm rates. Nonetheless, low false alarm rates are required because of the large land areas where mine clearance is necessary and the cost of clearing a putative mine. Current mine clearance technology predominantly relies on conventional sensors like metal detectors (MD). However, the commonly used metal detector offers only limited discrimination ability versus non mine objects. Therefore, even in the case where detection possibilities may be satisfactory, this may be achieved only at an enormous false alarm rate. This is justifying the evaluation of the advantages the integration of several sensors may offer. Until recently, the common choices for sensors include GPR, MD, and Thermal IR Cameras [1], [2]. A multisensor system acquires data according to different physical properties of landmine. This will allow combining complementary or orthogonal information about it. Whereas metal detector and GPR are not orthogonal sensors, as they both give the largest response signal in the presence of metal in the ground, thermal imaging suffering from extreme variability in performance as a function of dynamic environmental characteristics.

A nuclear technique based on the measurement of thermal neutrons produced from the moderation of fast neutrons by elastic scattering with hydrogen nuclei is effective for distinguishing explosive material, i.e., a landmine from other non hydrogenous materials. This approach can be considered as an effective way to discriminate between a landmine and other objects (e.g. stones and metals). However, this technique cannot stand alone for detecting of all landmines buried at different depths.

In this paper, we present a solution to this problem, by applying a multi-sensor fusion between the data recorded from the magnetometer and neutron backscattering. The diverse physical properties measured by the two sensors are attractive as a mean of improving mine detection process, along with decreasing the false alarm rate. The proposed approach is tested using different landmines buried at different depths. We also present a hardware implementation for the proposed approach using an inexpensive microcontroller chip.

II. NUCLEAR TECHNIQUES FOR LANDMINE DETECTION Several nuclear techniques have been suggested for the

detection of landmines and/or explosives, of which some are based on the use of neutrons. Techniques that can give information on the content of the landmine are thermal neutron analysis (TNA) [3], [4], and pulsed fast neutron analysis (PFNA), or a combination of the two [5]. Also, information on the content of the landmine can be gathered from the energy of neutrons that elastically interact in the hidden explosive, known as elastically backscattered spectrometry (EBS) [6]. Although some of these methods have shown great potential, due to the long acquiring time per hidden object, these techniques have only proven to be useful as a confirmation tool.

Nuclear techniques that are used for the localization of anomalies (landmines) in the soil include X-ray [7], [8]. Gamma ray [9], and neutron back scattering (NBS) imaging. The first two techniques are based on changes of the density in the soil. The main problem with these techniques is to achieve an acceptable penetration depth due to attenuation and multiple scattering. Neutron back scattering technique relies on the thermalization of fast neutrons. In this technique a thermal neutron detector monitors slow neutrons which

600

emerge from the soil while irradiating the soil volume under investigation with fast neutrons. Above a hydrogen-rich anomaly, thermal neutron flux increases, as hydrogen is an extremely efficient moderator. Such thermal neutron hot spots may be caused by a landmine.

III. NEUTRON BACKSCATTERING

A. Principles Neutron Back Scattering (NBS) demining techniques relies

on the fact that landmines contain much more hydrogen atoms than the dry sand in which they may buried. The NBS method can therefore be applied most advantageously to dry soils such as occur in arid climates with desert surroundings. Hydrogen in the landmine can be found not only in the explosive chemicals, but also in the casing of the plastic landmines. To detect a landmine, soil is irradiated by fast neutrons. These neutrons lose energy by scattering from atoms in the soil and become thermal neutrons after number of collisions. Then, thermal neutron detector monitors the slowed neutrons coming back from the soil [11]. Thermalization process takes many fewer collisions when scattering from hydrogen as compared to other elements. Thus, concentration of thermal neutrons in hydrogenous regions of the soil will therefore be relatively high. The thermal neutron flux at the surface will show an increase above areas which are relatively rich in hydrogen. Such thermal neutron response may indicate the presence of a landmine.

NBS imaging uses a two-dimensional position sensitive neutron detector to produce two-dimensional neutron distribution or image. The advantage of using a two-dimensional image is that the sensitivity for mine detection is greatly enhanced in comparison to only monitoring the overall count rate. The landmine response is not a real image of the mine because scattering of the neutrons in the soil blurs the image resulting in hot spots which are more or less has circular shapes. Size of the spot has a week dependence on the mine size, but strongly dependant on the mine depth [12]. Deeper mines will give hot spots with larger diameters because neutrons that have been thermalized in the landmine will diffuse over larger distances, not only vertically toward the surface, but also horizontally.

The main advantage of NBS method is the high speed of operation. Landmines may be found within a second when a sufficiently strong neutron source is used. Additional advantage of NBS based devices over metal detectors is the insensitivities for rocks, stones, and other metal objects, like shells or cartridges of fire arms that used to be in the battle fields. Furthermore, existence of hydrogen in the casing of plastic landmines makes NBS technique an attractive method to detect the fully-metal-free mines, which are used increasingly nowadays.

Since NBS demining technique depends on that the landmines have much higher concentration of hydrogen atoms

than the soil, NBS method is sensitive to the soil moisture. Hydrogen content of a landmine is comparable to that of soil with a moisture level of more than 10%, resulting in a loss of contrast between the mine and its surroundings [13]. Moreover, small variations in the distance between the detector and the soil, i.e., the standoff distance, significantly influence the count rate, which in turn will affect the ability of detecting the landmine [14].

B. ESCALAD : Egypt SCAnning LAndmine Detector A scanning NBS device has been developed at the reactor

and neutron physics department of the nuclear research center of the Egyptian Atomic Energy Authority, with the help of Delft University of Technology and supported by IAEA. The detector is based on the Delft DUNBID landmine system [15].

Fig. 1 shows The Egypt SCAnning LAnd mine Detector (ESCALAD). It consists of 16 proportional counter tubes with resistive wires, mounted parallel in a flat encasing of 142 cm × 117 cm × 15 cm, perpendicular to the direction of scanning. 3He tubes are sensitive mainly to thermal neutrons through the reaction: 3He + n → p + t + 764 keV. The specifications of the tubes are: length 100 cm, diameter 2.5 cm and 3He pressure 10 bar. The neutron detection efficiency varies from 1 for thermal energy down to 10-4 for neutrons in the MeV energy range. The tubes are mounted in two banks of 8 tubes; tube pitch within a bank: 27.8 mm, bank separation: 237.6 mm between the adjacent tubes axes. This separation allows for the placement of radioactive neutron sources or of a neutron generator in the center of the detector [16]. Bank separation is a compromise between two effects:

•The detection sensitivity decreases for an increasing separation distance because of the increase of the path length: source–mine–detector;

• Bull; the ’fast neutron background’ increases for smaller separation distances leading to a worse signal-to-background ratio. This type of background is caused by high-energy neutrons, which come straight from the source.

The total sensitive detector area is about 100 cm × 44 cm. The detector is mounted on a frame allowing it to move or

scan horizontally above the ground at a standoff distance of 10cm. The neutron counts are processed during the movement and are used to build an image of the thermal neutron flux, which is emerging from the ground. The coordinate of a neutron hit in the direction of scanning with respect to a ground related reference frame is determined from the position of the tube being hit in the encasing and the position of the encasing with respect to the ground. The coordinate along the tubes is determined by charge division performed by the electronics. Counts are stored according to the measured positions in a 32 × 256 array of pixels with a pixel size of 3.3 cm × 3.4 cm. The first dimension denotes the direction along the tube thus perpendicular to the scan direction; whereas the second one denotes the direction of scanning. The overall background count rate is about 500 c/s for a 252Cf source with an emission rate of 1.6 × 10-5 n/s, and a detector–ground-

601

surface distance of 10 cm. The maximum count rate possible with the current electronics is 2 × 105 c/s, which corresponds to maximum source strength of about 6 × 107 n/s. Each tube in turn spends a time of 3.4/ � s above a row of pixels, with � (cm/s) is the scan speed. The total measurement time for the detector tubes to pass completely over a row of pixels, i.e., the pass–over–time is: 16 (3.4/�).

Figure 1. Egypt Scanning Landmine Detector ESCALAD.

IV. MAGNETIC GRADIENT TECHNIQUE Magnetic detection is one of the standard methods of

environmental investigations used in locating ferro-metallic objects, especially when the signal-to-noise ratio of the magnetic anomalies is high. Measurement of perturbations in the direction and/or strength of the Earth’s magnetic field are used to locate underground ferro-magnetic objects. Gradient of the perturbations were also used for the detection.

An advantage of the gradient measurements is that it removes to some extent the undesirable influence of disturbing fields, compensates for the influence of the regional field and does not depend on diurnal variations in the Earth’s magnetic field. Recent development of magnetic survey systems has made it possible to rapidly detect much smaller objects using magnetic surveys, than it was possible before [17]. However, these advantages produce a large amount of data that need an automatic interpretation technique that can help in real time detection of landmines.

V. EXPERMINTAL SETUP In order to investigate the ability of combining both NBS

technique, and Magnetic Gradient technique, a test lane was conducted at the Nuclear Research Centre of Egyptian Atomic Energy Authority in Egypt. The layout of the test lane is illustrated in Fig. 2. The measurements were done in an open air in desert sand. The sand was very dry to depths of over 30 cm with silt and flint stones. The moisture content of the soil was very low. In this test, seven objects were used “see Fig. 3”. A separation distance of 2 meters is made between each object and the other. Table 1 lists the description and burial depth of the objects. The depth of the buried mine is defined as the distance between the soil surface and the top of the mine. The

sand was just left as it was after burying the mines and no effort was done to level the sand surface.

As it could be seen, both antipersonnel and antitank landmines with both metal and plastic casing used in the test, with a piece of wood that works as a “Clutter, i.e. Hydrogen Content Object” in the NBS images. Also a steel cylinder is used as a clutter object for the magnetometer response. For the plastic casing landmines, a small piece of metal was added to each one of them, representing the “detonator” of the landmine.

Whereas, ESCALAD used as the NBS devise, the magnetic instrument used in this experiment is a proton gradiometer with a resolution of 0.01 nT. This device uses two sensors to measure either vertical or horizontal gradient of the magnetic field. In this test, the instrument was used to measure the vertical gradient and the distance between the two sensors was 60 cm. The distance between the lower sensor and the ground level was 60 cm making the measurement level from the ground to be 90 cm. The measurements were made at a sampling interval of 30 cm.

Figure 2. Layout of the conducted Test Lane. Circles indicate the locations of the objects

Figure 3. The used objects in the Experiment- their description in Table 1

A B C

D

E F

G

602

TABLE 1 : DESCRIPTION OF THE BURIED OBJECTS

Object Type Size cm

Casing material

Explosive Content

G

Buried Depth

cm

A VS50 APM 9 x 4.5 Plastic 50

RDX 10

B Dummy APM 9 X 4 Bakelite

100 TNT

Simulant

15

C T-80 ATM

20.4 x 10.8 Plastic

2400 compB

20

D T-71 ATM 31.5 X 10 Metal 6000

TNT 30

E Cube of wood

10 X 10 X 10 wood

No Explosive Content

on surface

F Steel Cylinder 10 X 3 Steel

No Explosive Content

30

G PMN APM 11.2 x 5.6 Plastic 240

TNT 15

VI. DATA ANALYSIS

A. Magnetic Gradient Fig. 4 shows the recorded vertical magnetic gradient over

the buried objects. Many anomalies were recorded but not correlated directly to the landmines. Generally, the magnetic signature of hazardous objects is related to the physical properties of the ferromagnetic materials that are used as a casing or as the detonator of many landmines. The magnetic signature is usually controlled by the contribution result of remnant and induced magnetization. If both magnetization directions are vertical, then the anomaly will be symmetrical and has a peak located directly above the source. In Egypt, the direction of the Earth’s magnetic field is inclined and produces asymmetrical anomalies that are very difficult to interpret. Correction of the direction of the Earth’s magnetic field is usually made using a technique known as reduction to the pole. However, this reduction is correct if there is no remnant magnetization.

Figure 4. Vertical magnetic gradient magnetic profile over the buried objects.

Alternative to the reduction to the pole, analytic signal approach is used to enhance the shape of the magnetic anomaly over their sources. The approach of the analytic signal of magnetic anomalies, which was initially used in its complex function form and makes use of the properties of Hilbert transform [18]. The amplitude of the analytic signal is defined as the square root of the squared sum of the vertical and two orthogonal horizontal derivatives of the magnetic field. This transformation produce positive anomaly with peak located

nearly above shallow object. The appeal of the method is that analytic signal anomalies can be easily computed. Another reason for the appeal of the method is that the locations and depths of the sources are found with only a few assumptions about the nature of the source bodies.

Fig. 5 shows the analytic signal of the vertical gradient profile. The analytic signal transformation enhanced the position of the anomalies with respect to the actual location of the objects. The locations of the peaks over some objects were found slightly shifted. But most of the objects were detected. The red circles in the figure indicate the detected objects. Only C and E were not sensed by the magnetometer. It is not surprise as these objects are made of plastic and wood. However some anomalies not related to the objects were also detected. The black circles in the figure indicate those anomalies. See, for examples, the anomaly between B and C and the anomaly between D and E. These anomalies are most probably related to scattered debris and would represent stumbling block in enhancing the delectability using magnetic systems.

Figure 5. Analytic signal of the vertical gradient magnetic profile.

B. Neutron Backscattering For ESCALAD NBS device, the image of the intensity

distribution of the back-scattered thermal neutrons over the soil surface is formed while scanning. Fig. 6 shows the scanned NBS over the buried objects. The area of hot spot depends on mine burial depth, mine size and detector standoff distance. Generally the raw data consists of three components: 1) signal from the mine (if present), 2) a contribution caused by neutrons scattered from the soil, and 3) a contribution from fast neutrons, which hit the detector without having entered the soil. The latter two components constitute to the background in the image. A row of image pixels perpendicular to the scan direction will be called “pixel raw”, while a row of image pixels along the scan direction will be called “pixel line”.

Figure 6. Measured signals of the lane with the tested objects by ESCALAD system

The first step in the processing of the raw image is the application of a “fast neutron background” correction. The fast neutron background is an intrinsic system property. It has been measured by placing the detector about 1.5 m above the soil and registering the count rate distribution along each tube. The correction is done by summing the fast neutron measurement

603

along the scan direction and subtracting the result from the pixel rows in the raw image after normalization to the measurement time. The effect of this correction is to remove the central high intensity band and to increase relatively the intensity at the edges of the detector. The next step in the image analysis concerned with the using of a proper filter to enhance the mine signal over the background fluctuation. The process consists simply of moving the center of the filter mask of size 3 x 3 pixels from point to point in an image. At each point (X, Y), the response of the filter at that point is the sum of products of the filter coefficients and the corresponding neighborhood pixels in the area spanned by the filter mask.

Fig. 7 shows the filtered image for scanning of lane test. The obtained image shows clearly the effectiveness of ESCALAD system for detecting of landmine with different amount of explosives encased in plastic or steel casing. The only object that has not been detected is a piece of steel was buried at a depth of 30 cm, while it was detected by the magnetic gradiometer. Also ESCALAD detected the wood cube, resulting in a Hot spot that look like a landmine signature.

Figure 7. Filtered image of the lane with the tested objects by ESCALAD system

VII. FUSION FOR LANDMINE DETECTION For the aim of obtaining fusion results as good as possible,

the choice of an adequate combination method has to be carefully done. The fusion level defines the level at which the information of the different sensors is combined. The level of fusion mainly depends on the types and similarity of sensors to be fused [19]. When the sensors are alike, the fusion is done at the data level to take all data into account. When sensors are very different, decision-level fusion is more suitable and is also computationally more efficient.

The focus of this paper is to apply decision fusion to real landmine data from both the neutron backscattering, and the magnetic gradient techniques. For our algorithm, we first determine the peaks triangles in the magnetic analytic signal data, and define them as “suspected areas”. Then we take the corresponding area in the neutron backscattering data to determine the existence of a hot spot. Since the hot spot dimensions are variable, in our algorithm, we calculate the angle Ө between the “maximum count”, and the “minimum count” in the suspected area “see Fig. 8”. If this angle is more than a certain threshold, the suspected neutron backscattering area considered to have a hot spot. The landmine existence decision is taken if the suspected area has both a peak in the magnetic analytic signal data, and a hot spot in the neutron backscattering data. Otherwise, this location is announced to have no buried landmine. The proposed scheme is illustrated in Fig. 9.

For the validation of the algorithm, it was tested using the magnetic analytic signal, and neutron backscattering data measured over the tested objects. By having the analytic signal data as an input, the algorithm detects the existence of 12 peaks, corresponds to 12 suspected areas. Afterwards, the existence of hot spots in these suspected areas was examined. The results of the algorithm are shown in Table 2.

Figure 8. The angle Ө between the maximum count rate and the mimimum count rate for the suspected area contains VS50 APM landmine

Figure 9: Fusion Algorithm Scheme

It can be noticed that the algorithm detected the first two landmines successfully. Furthermore, the algorithm detected the third landmine, but at this time based on the area under the anomaly between B and C, and the existence of a “hot spot” under this area. This can be considered as a false result even there is a landmine there, as the used magnetic anomaly peak is not the real magnetic landmine signature. The same situation happened in the case of the anomaly between C and D, as the algorithm detects the existence of a landmine, whereas, there is no one there.

The algorithm couldn’t detect the T-71 landmine; even it has a strong anomaly in the magnetic analytic signal data, because its NBS signature is very weak. The same situation happened in the case of PMN landmine, where the algorithm failed to detect it. The suspected areas where the cube of wood, and the steel cylinder were buried there, resulted to be no landmine areas, which eliminates the false alarms of both the magnetic and the neutron backscattering techniques. The rest of the suspected areas that have anomalies related to scattered debris, with no neutron backscattering hot spot, resulted to be no landmine areas.

604

TABLE 2 : FUSION ALGORITHM RESULTS

The Data Fusion algorithm, is embedded on EASYPIC v7

Kit, 40 MHz osc./clock input, with PIC 18F452 Microcontroller, used for implementation “see Fig. 10”. The microchip has 1536 bytes of RAM, and 32Kbyte flash memory used as a central processing unit to acquire data from the magnetometer, and neutron backscattering sensors, through the PC, process the data to find the anomaly peaks, calculate the suspected areas, compare it with the corresponding area in the neutron backscattering data, finding the hot spots, and generate the results of the suspected areas, with the location of the landmines, present it on the text LCD. The algorithm consumes 88 % of the available RAM, and 99 % of the flash memory. The Kit using only 5 volts as a processing power, and verify if the suspected area has a landmine or not, beside its position if any, in around 5 seconds, from acquiring the data, which make the proposed approach more suitable for real time application.

Figure 10. EASYPIC v7 Kit, with PIC 18F452 Microcontroller, used for implementation.

VIII. COCLUSION In this paper, decision fusion algorithm was applied to

magnetic and neutron backscattering data resulted from scanning over real landmine objects. The results are promising in discriminating the landmine signature from the surrounding clutter. Furthermore, the proposed algorithm was implemented on an inexpensive microcontroller chip that has low power consumption. The algorithm shows an acceptable accuracy in detecting the landmine location, in a few seconds, using the available memory. This accuracy, along with small detecting time makes the suggested approach suitable for real time detection of buried landmines.

the results of the proposed algorithm concludes to the potential advantage of using an integrated system containing nuclear technique represented on ESCALAD system and conventional technique that depend on the detection of metal parts only represented in magnetic gradiometer.

However, the fusion algorithm results suggest utilization of more sensors that are able to detect different landmine physical properties. Using more sensors will be very effective in decreasing the false alarms that mislead the proposed fusion algorithm. A confirmation technique like elemental analysis gamma ray sensor can be used for more investigation of such false responses, and would help greatly in the demining process and reduce the risk of hazardous objects.

ACKNOWLEDGMENT The authors are greatly indebted to Prof. Riad Megahid,

Laboratories for Detection of Landmines and Illicit Materials, Nuclear Research Centre, Atomic Energy Authority, for his kind cooperation. The authors are grateful to Prof. Fouad Soliman, Exploration Division, Nuclear Materials Authority, for helping during practical parts of this work.

REFERENCES

[1] Nada Milisavljevic, and Isabelle Bloch, “Sensor fusion in anti-personnel mine detection using a two-level belief function model,” IEEE Trans. Sys.Man. cyper. Part C: Applications and Reviews, vol. 33, no. 2, pp. 269-283, May 2003.

[2] Toshio Fukuda, Yasuhisa Hasegawa, Kazuhiro Kosuge, Kiyashi komoriya, Fumihisa Kitagawa, and Tomohiro Ikegami, “Environment-adaptive antipersonnel mine detection system-advanced mine sweeper,” in Int. Conf. Int. Robots, and Systems, Oct. 2006, Beijing, China.

[3] T. Cousins, T. A. Jones, J. R. Brisson, J. E. McFee, T. J. Jamieson, E. J. Waller, F. J. LeMay, H. Ing, E. T. H. Clifford, E. B. Selkirk, “The development of a thermal neutron activation (TNA) system as a confirmatory nonmetallic land mine detector,” J. Radioanal. Nucl. Chem., vol. 235, no. 1-2, pp. 53-58, 1998.

[4] G. Viesti, “Exploring the limits in the detection of landmines by neutron induced reactions,” Zagreb, Croatia, IAEA/PS/RC-77, presented in 1st Research Coordination Meeting on Applications of Nuclear Techniques to Anti-Personnel Landmines Identification, Nov. 1999.

[5] G. Vourvopoulos, P. C. Womble, and J. Paschal, “PELAN : a pulsed neutron probe for uxo, ied and landmine identification,” Zagreb, Croatia, IAEA/PS/RC-799, presented in 1st Research Coordination Meeting on Applications of Nuclear Techniques to Anti-Personnel Landmines Identification, Nov. 1999.

Magnetic Anomaly (Peak)

Nuclear Backscattering

( angle Ө)

Fusion Decision

Object Type

1 0.33o No No Landmine

2 76.35o Yes VS50 APM

3 73.80o Yes Dummy APM

4 83.26o Yes T-80 ATM

5 52.71o Yes No Landmine

6 00.38o No T-71 ATM

7 00.89o No Cube of wood

8 0.34o No No Landmine

9 0.83o No Steel Cylinder

10 0.18o No No Landmine

11 0.17o No PMN APM

12 0.03o No No Landmine

605

[6] J. Gsikai, “landmine identification by elastically backscattered Pu-Be neutrons,” Zagreb, Croatia, IAEA/PS/RC-799, presented in 1st Research Coordination Meeting on Applications of Nuclear Techniques to Anti-Personnel Landmines Identification, Nov. 1999.

[7] S. Shope. G. J. Lockwood, J. C. Wehlburg, M. M. Selph, J. M. Jojola, B. N. Turman, and J. A. Jacobs, “Real-time x-ray backscatter imaging of landmines,” Zagreb, Croatia, IAEA/PS/RC-799, presented in 1st Research Coordination Meeting on Applications of Nuclear Techniques to Anti-Personnel Landmines Identification, Nov. 1999.

[8] E. T. Dugan, and A. M. Jacobs, “X-ray lateral migration radiography,” presented in the Landmine Basic Research Tech. Rev. Conf., Aug. 2000.

[9] L. Zhang, and R. C. Lanza, “CAFNA, coded aperture fast neutron analysis for contraband detection: preliminary results,” IEEE Trans. Nucl. Sci., vol. 46, no. 6, pp. 1913-1915, 1999.

[10] F. D. Brooks, A. Buffler, and S. M. Allie, “Detection of plastic land mines by neutron back scattering,” in 6th Int. Conf. Appl. Neutron Science, Crete, Greece, Jan. 1999.

[11] G. Nebbia, “DIAMINE (detection and imaging of anti-personnel landmines by neutron backscattering technique),” presented at Proc. 5th Int. Symp. Technol. Mine Problem, Monterey, CA, Apr. 2002.

[12] Victor Bom, A. Mostafa Ali, A. M. Osman, A. M. Abd El-Monem, W. A. Kansouh, R. M. Megahid, and Carel W. E. van Eijk, “A feasibility test of land mine detection in a desert environment using neutron back

scattering imaging,” IEEE Trans. Nucl. Sci., vol. 53, no. 4, pp. 2247-2251, Aug. 2006.

[13] J. Obhodas, D. Sudac, K. Nad, V. Valkovic, G. Nebbia, and G. Viesti, “The soil moisture and its relevance to the landmine detection by neutron backscattering technique,” in Nucl. Instrum. Methods Phys. Res. B, vol. B213, pp. 445-451, Jan. 2004.

[14] C. Datema, V. R. Bom, and C. W. E. van Eijk, “Experimental results and monte carlo simulations of a landmine localization device using the neutron back scattering method,” in Nucl. Instrum. Methods Phys. Res. A, vol. A488, no.1-2, pp. 441-450, Aug. 2002.

[15] V. R. Bom, C. W. van Eijk, and M. A. Ali, “DUNBID, the delft university neutron backscattering imaging detector,” in Appl. Radiat. Isot., vol. 63, no. 5-6, pp. 559-563, Nov. / Dec. 2005.

[16] Victor R. Bom, R. M. Megahid, A. M. Osman, and M. Musa, “ESCALAD, the egyptian scanning land mine detector, first field test results,” in 9th Int. Conf. Appl. Nucl. Appl., Crete, Greece, Jun. 2008.

[17] Salem, A., Ravat, D., Gamey, T. J., and Ushijima, K., “Analytic signal approach and its applicability in environmental magnetic applications,” Journal of Applied Geophysics, vol. 49, pp. 231-244, 2002.

[18] Blakely R. J., Potential theory in gravity and magnetic applications: Cambridge University Press., 1995.

[19] E.Waltz, J.Llinas, Multisensor Data fusion, Artech House, 1990.

606