Landmine Characterization Applying GPR Assessment and Modeling Approaches

Landmine clearance is a critical problem faced by many countries around the world, and the situation can be compounded by natural disasters or land development. Therefore, it is an urgent issue to detect landmines in the ground and remove them safely. For safe detection, non-touch-based detection methods are required. These methods involve the detection of landmines from the signals obtained by nontouch-based sensors, such as metal detectors and radars. Among those sensors, ground penetrating radars, or GPRs, is an attractive choice for landmine detection due to their advantages over other sensors. The GPR can be used as a stand-alone sensor or as a complementary sensor to a metal detector [1,2]. It can detect both metal and nonmetal landmines. Moreover, its weight can be made light, so that it can be installed in a handheld system or in a vehicle-mounted system in the form of an array of multiple antenna elements [3]. Identification of landmine from other targets or clutters is a vital task. For that reason, Wide varieties of signal processing techniques have been evolved and used to process GPR signals. Gader and Nelson et al. [4,5] have proposed a gradientbased method for landmine detection. Three features based on Singular Value Decomposition (SVD), Discrete Fourier transform DFT and Principal Component Analysis, (PCA), are extracted to each signal for landmine detection, which is performed by using the Mahalanobis distance method.


Introduction
Landmine clearance is a critical problem faced by many countries around the world, and the situation can be compounded by natural disasters or land development. Therefore, it is an urgent issue to detect landmines in the ground and remove them safely. For safe detection, non-touch-based detection methods are required. These methods involve the detection of landmines from the signals obtained by nontouch-based sensors, such as metal detectors and radars. Among those sensors, ground penetrating radars, or GPRs, is an attractive choice for landmine detection due to their advantages over other sensors. The GPR can be used as a stand-alone sensor or as a complementary sensor to a metal detector [1,2]. It can detect both metal and nonmetal landmines. Moreover, its weight can be made light, so that it can be installed in a handheld system or in a vehicle-mounted system in the form of an array of multiple antenna elements [3]. Identification of landmine from other targets or clutters is a vital task. For that reason, Wide varieties of signal processing techniques have been evolved and used to process GPR signals. Gader and Nelson et al. [4,5] have proposed a gradientbased method for landmine detection. Three features based on Singular Value Decomposition (SVD), Discrete Fourier transform DFT and Principal Component Analysis, (PCA), are extracted to each signal for landmine detection, which is performed by using the Mahalanobis distance method.
Egypt is one of the most contaminated landmine countries. The problem of landmines in Egypt has started up since the World War II in the northern part if western desert. The military operations carried out by the Allied Forces and the Axis Power from 1941-1943 left varieties of about 22 Million landmines and UXO in western desert nearly along the coast of the Mediterranean sea. Not only thousands of civilians killed and injured each year, but also the social, economic and environmental impacts of those mines are disgraceful [6]. Due to the risk to implement the research of the mines in the contaminated areas, it is often carried out measurements on test sites, or laboratory studies.
In this paper, the finger prints of various buried targets of different material were obtained through the application of our technique [7]. Initially, simulation models for different targets such as perfect electric conductor (PEC), plastic (PVC) and limestone were designed by using a plane wave electromagnetic source to find guideline results for comparison in case of real field measurements.
Secondly, our technique was applied to the test site measurements (GPR data). Our MATLAB code was designed to process the data which includes the removing shift in time, the removing the antenna coupling and the automatic gain control (AGC) to improve and enhance the data. Also the moving average and the bandpass filtering, processing were used to smooth and remove unwanted frequencies before using our technique. Subsequently, we obtained the finger print of various buried targets.

Methodology
The reflected power was compared in the illustrated technique to determine the locations of targets. The continuous wavelet transform was applied to offer very good time and frequency localization. As a *Corresponding author: Mahmoud A. Mohana, National Research Institute of Astronomy and Geophysics, Egypt, Tel: +20 2 255 43 111; E-mail: Mahamoud2746@ hotmail.com; Sherinmohamed2001@yahoo.com In this paper, our main objective is to validate the ability of Ground penetrating Radar to discriminate between various buried targets. Wavelets transform was used to obtain the spectrum distribution of every buried target. The difference between the distributions of the different target spectrum can be considered as a finger print for each one, the summations of powers at the locations of targets were calculated and compared.
Simulation models for different targets were made. The reflections of targets were analyzed by Daubechies Wavelets (db2) transform to differentiate between different targets and to get finger print for every target. This technique was applied for field measurement of different target type and the technique revealed that the variation of finger print for every buried target. We believe that with this t Also the summation of the multiple received reflections power was varied according to targets kind as shown in Figure 5a-5c which is used as a way to distinguish between the various targets too.
For further illustration, the reflected signal from each target was extracted separately by applying time gate of the Hammining window as shown in Figure 6a-6c. The finger prints of each target after using Hamming window were obtained by using the previous technique as shown in Figure 7a-7c.

Test-site measurements
An experimental test site has been constructed within the campus of the National Research Institute of Astronomy and Geophysics (NRIAG) in the Fayoum as described in The experimental test site has been surveyed by using the GPR instrument of Geophysical Survey System Inc. (GSSI) model SIR 20H attached to the antenna of 1.5 GHz central frequency.
fact that the summation of reflected power from metallic targets is the maximum between all targets, so that the reflected power distribution at the target locations was studied. We found that a significant difference between power distributions of the different target spectrum, which varies depending on the target type that can be considered as a finger print for every target. Also the multiple received reflections were varied depending on target kind. This technique enabled us to discriminate between different targets.

A commercial software CST [8]
Microwave Studio program was used as a FDTD simulator for observing the electromagnetic scattering from a determined landmine (metal or plastic). Table 1 shows the numerical simulation parameters for different landmines models and limestone model as shown in Figure 1 in which plane wave is used as electromagnetic source. These models were simulated with operating frequency from 400 MHz up to 1400 MHz; central frequency 900 MHz. The plane wave source was used instead of antennas for well describing and well controlling the electromagnetic radiation or wave from targets. It facilitates the simulation of an incident wave from a source located a large distance from the observed object in combination with far field monitors.
For the FDTD numerical simulation, exciting the model with a plane wave source requires several initial conditions to be satisfied such as: 1. 1. The surrounding space should consist of a homogenous material distribution 2. 2. Boundary conditions must be defined as the direction of incidence 3. 3. Other excitation ports must not be located on boundary conditions [9].  Figure 2, there were three various reflections. The first one was the direct coupling, the second one was the ground reflection and the last one was the reflection from different targets. By using the fact that the reflection of the metal target is much greater than the other reflections, then by comparing the reflected power of all targets as shown in Figure 3, we were able to confirm the location of the targets which indicates the accuracy of our results. By applying our technique on these signals, the finger print of each previous target was detected. By comparing Figure 4a-4c at the same location determined in Figure 3, we have found that there are different signal image for each target. The significant difference in the signal image was considered as a finger print to this specific target. By calculating the relative summation of power distribution at target locations, we found that the distribution of power from the metallic target was five times, limestone target and it was eighty double the plastic target which enables us to easily differentiate between them.   The one of the profiles which was carried out on a test site, that contains a set of targets made of different materials was selected as illustrated in Table 2. Figure 8 shows the GPR section of this profile. Initially, we needed to apply processing methods on the GPR data of the experimental test site. Our MATLAB code was designed to process the GPR data which included the static correction of time shift, background removal and adding the automatic gain control (AGC) to improve and enhance the data. Also the moving average and the bandpass filtering, processing were used to smooth and to remove unwanted frequencies. Figure 9 shows the final GPR section after processing.
Secondly, We applied our technique to that profile after processing. By using the reflected signal of each target, the different finger prints of real field measurements were obtained. Figure 10a-10c, show the fingerprint of each buried targets. Also by studying the summation of power distribution, we found a great ability to differentiate between           the various targets where the percentages of relative summation vary in iron and aluminum and plastic.
Figure 11a-11c shows another method to discriminate between the targets by calculating the sum of multiple received reflections which varied according the target type too. The results of real measurements were very close to the results of simulation models, which confirm the credibility of our own technique.

Conclusion
Many of researches were provided for the detection of landmines sites to save human life and to obtain optimum utilization of the contaminated areas. Given the gravity to implement the researches of the mines in the affected areas, it is often carried out measurements on test sites, or laboratory studies. In this paper, a novel method for landmine detection is presented. We were able to differentiate between the various targets using our own technique, which relies on applying wavelet transform on the reflected signals from the targets and then calculate the summation of power distribution at target locations which gave a high performance to differentiate between them. The significant difference between the distributions of the different target spectrum can be considered as a finger print for every target, also the multiple received reflections varied according target type.
We achieve a success to differentiate and to obtain Finger prints of almost targets by using illustrated technique. The finger prints of trusted simulation models were obtained to bring out the set of specific concepts in order to rely on them in real field measurements.
Data are obtained by using a GPR of pre-equipped test site. They were processed in order to reduce unnecessary signals in order to isolate signals for landmines. The finger prints of various buried targets were obtained by using illustrated technique too.