Increasing the Probability of Detection and Evaluation of Buried Metallic Objects by Data Fusion GPR- Low Frequency Electromagnetic Sensor Array

Transcription

1 4th European-American Workshop on Reliability of NDE - Poster 4 Increasing the Probability of Detection and Evaluation of Buried Metallic Objects by Data Fusion GPR- Low Frequency Electromagnetic Sensor Array Adriana SAVIN, Alina BRUMA, Raimond GRIMBERG, Sorin LEITOIU, Rozina STEIGMANN, National Institute of R&D for Technical Physics, Iasi, Romania Cristian Claudiu COMISU, Technical University Gh.Asachi, Iasi, Romania Abstract. In present, for detection of buried metallic objects, are used two methods: Ground penetrating radar (GPR) - that functions usually at frequency in the range of hundred of MHz; this method presents a good spatial resolution but, due to the relatively small penetration depth it is efficient as far as depths in the range of few meters. Low frequency electromagnetic induction methods (EMI) that works in the range 1-50kHz and can detect metallic objects buried at depths much bigger but has the disadvantages of a spatial resolution inferior to GPR. To improve the spatial resolution of EMI, we propose the using for detection of a sensor array which, using the super resolution algorithms allows the simultaneous obtaining of a better signal to noise ratio and of a superior spatial resolution. The data fusion GPR-EMI leads to the increasing of the probability of detection and to improving the evaluation of shape and position of the buried metallic objects. Introduction Inverse electromagnetic scattering by objects that lie in free space or in layered media play an increasing role in a wide range of technological applications. It is for this reason that, during the years, many methodological approaches have been developed for a variety of problems. Among the numerous technological applications of the electromagnetic inverse scattering, Ground Penetrating Radar (GPR) also known as Georadar, is one of the most important. GPR is a near-surface remote sensing tool for detecting buried targets. Interesting applications fields of GPR are measurements for object location into the subsoil (i.e. piping, electrical phone cables, etc) or soil characterization. In all these applications, it is very important to quickly obtain measurements with a high level of precision in terms of location and dimensions of buried objects [1], [2]. As any other instrument based on electromagnetic scattering by objects, GPR presents in depth spatial resolution, which increases with the increasement of radar pulse fundamental frequency. Due to the scattering on soil inhomogeneities, working with a high central frequency will lead to the limit of penetration depth of the radar pulse, thus to the decrease of the investigation possibilities. There are other electromagnetic methods for detection of buried objects, these working at low frequencies from few Kilohertz up to tens of Megahertz [3], [4]. These methods have the advantage of high penetration depth, but a very low spatial resolution. 1

2 This disadvantage can be reduced by using sensor arrays for the scattered signal reception, and using super resolution procedures [5]. This paper proposes the simplification of the methods used for identification of objects buried in soil, using data fusion between the results delivered by GPR method, and those delivered by electromagnetic method with sensors array as reception and signal post processing with super resolution procedure, based on the maximum likelihood method. 2. Theory Background on GPR The GPR method is based on sending electromagnetic waves into the ground using a transmitting antenna, T x. A receiver antenna is used to collect the reflected signal from the interfaces between the materials and the scattering from inhomogeneities, having different electromagnetic properties within the materials (Figure 1). Figure 1. GPR functioning principle In free space, the radar impulse wavelength is c λ = (1) f where c=3x10 8 m/s is the light s speed in vacuum, and f is the fundamental frequency. In an amagnetic media (μ r =1), the wavelength of the radar impulse is λ0 c λ = = (2) ε r f ε r ε where ε r = is the relative permeability of the soil. ε 0 The radar impulse generated by the emission antenna is partially reflected by the soil s surface, the reflection coefficient being R. Another part of the impulse will be transmitted in the soil, the transmission coefficient being T, and will be reflected if a scattering object if is met in soil.. The computation of the scattered wave intensity, received by the reception antenna, can be made by different procedures: the analytical computation of the Maxwell s equations solutions using the Green s functions method [6] or the numerical computation using finite difference time-domain method (FDTD) [7]. The nature of the GPR forward problem classifies it as an initial value-open boundary problem. This means that, in order to obtain a solution one has to define an initial condition (i.e. excitation of the GPR transmitting antenna) and allow to the resulting fields to propagate through space reaching a zero value at infinity since, there is no specific boundary which limits the problem s geometry, and where the electromagnetic fields can take a predetermined value. The FDTD approach to the numerical solution of Maxwell s equations is to discretize both the space and the time continuously. 2

3 In Figure 2 we present the simulation of GPR response having antenna set-up at 400MHz frequency which scans a monolayer soil having ε r =3, in which a parallelepiped void filled with air exists, having ε r =1, at 0.3 m depth under soil s surface. The void has 650x650x550mm 3 dimensions. The simulations have been made using the GPR MAX 2D free software [8] Figure 2. A GPR scan over a parallelepiped void having the dimensions of 650x650x550mm 3, placed under soil s surface, at a depth of 0.3m The wavelength in soil is 0.433m, inferior to the dimension of void. Examining the data presented in Figure 2, it can be observed that both surfaces of the void, superior and inferior, are visible. Also, it is observed that estimation of void dimensions along the scanning direction can be made. After that, we consider two ferromagnetic pipes, having 55mm outer diameter and 550mm length at the bottom of the void. These pipes can simulate the projectile of a launcher. The simulations have been effectuated for two distinct positions of the pipes: the pipes are separated between them, each being placed tangent to the lateral walls of the void (Figure 3a) and respectively, one by one placed at the middle of the void bottom (Figure 3b). a b Figure 3. GPR response for two ferromagnetic pipes with 55mm diameter a) pipes tangent to the void s lateral walls; b) pipes one at the side of other at the middle of void s bottom Examining the simulation results of a relatively simple problem as the one presented above, it can be observed that the interpretation of the results is difficult enough even in the absence of the measurement noise as well as the scattering on different layers and constituents of the real soil. For this reason we propose the using of other low frequency electromagnetic method, which cannot detect layers of the soil and its constituents, differing only by small 3

4 modifications of dielectric permittivity, but detecting metallic objects buried in soil and large voids from soil. 3. Electromagnetic Method for Determination of Soil Condition The method is based on an send-receiver transducer. The emission part is made from a plane rectangular spiral coil with 1m outer side, 0.06m step having 10 turns. The current density in the emission coil is 6A/mm2 and frequency 47kHz. The electric field created by the emission coil is presented in Figure 4. Figure 4. The electric field created by the emission coil The reception part is made from a 5x5 rectangular coils array, having 0.05x0.05mm2 dimensions and 100 turns. The physical realization of the transducer is presented in Figure 5a and the principle scheme is presented in figure 5b. a b Figure 5. Electromagnetic sensor array for determining soil condition a) physical realization; b) principle scheme The coils from the array are sequentially interrogated; the electromotive force induced in the array s coils is measured with a Lock-in Amplifier SRS850 and recorded into a PC. Noting with wij the weight value corresponding to the coils of whose center is indexed by the pair (i,j) we define the beamformer s output signal for the reception coil indexed (i,j) to be zij = wij yij (3) where z can be amplitude or phase output, y can be the amplitude or phase of e.m.f induced in the reception coil. w is the weight matrix which should assure the fact that, in 4

5 free space, the amplitude and phase of z for all the coils from array are approximately equals. We define the array s sensibility matrix as A = w* w H (4) where the superscript H denote the hermitic matrix. We can define a new matrix S, as being H ( ) 1 S A A A A The autocorrelation matrix of the response signal of the array can be defined as H R = z z (6) H = (5) xx The source locations are estimated to line at the points in space where the function S R xx (7) is maximal. The use of the super-resolution procedure simultaneous leads to the improvement of the signal to noise ratio, too. 4. Experimental results A soil region was examined using GPR type Utility Scan - Geophysical Survey Systems, Inc. (GSSI) - USA, with an antenna of 400MHz. The experimental data were processed afterwards by means of Radan 6.5 software, produced also by GSSI In figure 5 a, b, c, d we present the radar recordings for the following situations: Figure 5a the initial situation of the soil; Figure 5b the radar response in the case of a parallelepiped air void having the dimensions 650x650x550mm 3 at a 0.3 m depth; Figure 5c in the void, two ferromagnetic pipes having a diameter of 55 mm, were placed, tangent to the void s lateral walls; Figure 5d the two pipes are tangent in the central region of the void. In Figures 6 a, b, c, d, we present, in the same order, the results obtained with the electromagnetic transducer, based on a sensors array reception. Comparatively examining the data from Figures 5 and 6 it can be observed that the two methods for evaluation of soil condition are complementarily, meaning that the information provided by GPR can reveal the presence of layered soil and voids, while the information delivered by the sensors array transducer and using super-resolution procedure, can not indicate these, instead the depth of investigation is bigger. Due to the relatively high frequency of the GPR antenna, the examination speed is big enough, about 1m/s. Using the electromagnetic low frequency procedure, the examination speed is lowered. a b 5

6 c d Figure 5. GPR real recordings: a) initial situation of the soil; b) void in soil; c) pipes tangent to the void s lateral walls; d) pipes one at the side of other at the middle of void s bottom a b 6

7 c d Figure 6. Electromagnetic transducer recordings: a) initial situation of the soil; b) void in soil; c) pipes tangent to the void s lateral walls; d) pipes one at the side of other at the middle of void s bottom The electromagnetic low frequency procedure based on sensors array proves its utility in condition in which GPR signals emphasize complex structures in soil, which can be metallic objects buried at relatively high depths, their presence, shape and orientation being confirmed by this procedure. 5. Data fusion The notion of data fusion is referred to the fact that we have data about a certain object, phenomenon or process, delivered by different types of sensors which measure different parameters and, through combination of all these information, after certain rules, we obtain a much more correct and complete characterization, of the studied item. Data fusion can be realized if fulfil the following conditions: the information delivered by sensors fixed or mobile shall have same referential the data delivered shall be scaled, working with relative measures, brought for example in the range 0 1, to can be compared the rules for data fusion must take into account the physical significance of the 7

8 information delivered by sensors and by the correlation between these and the item. In the situation presented in this paper, the GPR data provides at one scan, information about the amplitude of reflected wave and the time between the moment of radar impulse emission and the one when the antenna receive the impulses came form the scattering on diverse inhomogeneities from the soil. Thus, the GPR information is presented as a matrix which groups the position of antennas (along one direction), time of flight and respective the amplitude of received impulse. Knowing the dielectric permittivity of soil, the time of flight can be converted in distance along vertical direction, in the scale of depth at which the scatter is placed. Grouping more scans, a 3D image can be obtained, in which the amplitude of the impulse received by the antenna is commuted into a coloured scale. In the electromagnetic low frequency method, the information consists in the amplitude of the scattered electromagnetic field (real and imaginary components) in the different coils that form the array. To evaluate correctly the shape and the position of scatter, the inversion operation is necessary. The solution of the inverse problem will not be unique, in the actual stage of knowledge in this domain, due to the absence of some initial information that are the constraints for the inversion operation. Thus, we propose the simultaneous use of the two procedures any time when doubts appear at the interpretation of GPR data, the fusion being made through the interpretation of human operators. 6. Conclusions The GPR method allows the evaluation of soil s condition and detection of its inhomogeneities, depending on the penetration depth and spatial resolution, the low frequency electromagnetic method allows a good evaluation of the position and orientation of metallic objects buried in soil, if the super-resolution method is used. Combining the information delivered by the two methods, the probability of detection can be improved as well as the one for the interpretation of information, especially for the detection of buried dangerous objects. 7. Acknowledgement This paper is partially supported by Romanian Ministry of Education, Research and Youth, under CNMP Contract no /2008 SysArr. References [1] D J Daniels, Ground Penetrating Radar 2nd ed, London, 2004 [2] L.P.J. Peters, J.J. Daniels and J.D. Young, Ground penetrating radar as a subsurface environmental sensing tool, IEEE Proc 82 (1994), pp [3] Method for imaging with low frequency electromagnetic fields, US Patent , dec.1994 [4] H. Huang,, I.J. Won, Conductivity and Susceptibility Mapping Using Broadband Electromagnetic Sensors, Journal of Environmental and Engineering Geophysics: Vol. 5, No. 4, pp [5] R. Grimberg, L. Udpa, S. S Udpa, Electromagnetic transducer for the determination of soil condition, Applied Electromagnetics and Mechanics, IOS Press, 28, (1-2), (2008), pp [6] R. Grimberg, L. Udpa, A. Savin, R. Steigmann, V. Palihovici, S.S. Udpa, 2D Eddy Current Sensors array, NDT and E International, Elsevier, 39, 4, (2006), pp [7] A. Giannopoulos A. The investigation of transmission-line matrix and finite-difference time-domain methods for the forward problem of ground probing radar, D.Phil. thesis, Dept. of Electromagnetics, University of York, UK,1997 [8] GPR MAD 2D/3D User s manual, ver.2.0, 8