CLUTTER REDUCTION TECHNIQUES FOR GPR BASED BURIED LANDMINE DETECTION Tesfamariam, Gebremichael T. Institute of Telecommunications Signal Processing Group, Technische Uniiversitat Darmstadt Merckstarsee 25, Darmstadt, Germany gtesfa@spg.tu-darmstadt.de Dilip Mali Deptt. of Electrical and Computer Engg, EiT-M, Mekelle University, Mekelle, Ethiopia malidilips@gmail.com Abdelhak. M. Zoubir Institute of Telecommunications Signal Processing Group Technische Universitat Darmstadt Merckstarsse 25, Darmstadt, Germany zoubir@spg.tu-damstadt.de Abstract—Landmines are affecting the lives and livelihoods of millions of people around the globe. There are more than 110 million landmines in nearly 62 countries cause over 26,000 casualties each year. The costs of single landmine for buying and planting are about 3 - 10$ whereas it costs 300 - 1000$ for removal. Annually, about 80,000 APMs were being lifted, while about 2.5 million new APMs were being planted. Modern landmines are mainly non-metallic or contain small metal, so that conventional metal detectors cannot be used efficiently. Considerable efforts are put in GPR development for detection of shallow buried landmines. But, GPR also performs inadequately due to clutter, which dominates the data and obscures the main information. Clutter varies with surface roughness and soil conditions lead to measurement uncertainty and high false alarm rates. In this paper, advanced signal processing techniques are going to be used to reduce the clutter and we will compare background reduction techniques based on their test statistic. It is seen the width of the window has an impact on the estimation. It is found that running median of an appropriate window size is the best in reducing background clutter, running mean is the next. Keywords— Clutter Reduction, Signal Processing, Ground Penetrating Radar, Anti Personnel Mines, Ground Clutter Removal. I. INTRODUCTION A research conducted by US State Department indicates that landmines maim or kill an estimated 500 of people per week worldwide [1]. In addition to the direct casualties, the areas of arable land cannot be farmed due to the threat of landmines. According to the UN, in 64 countries around the world, there are estimated 110 million landmines still actively lodged in the ground. In practice, current demining techniques involve the use of explosive sniffing dogs, metal detectors, and mechanical prods [1], [2]. Metal detectors are very effective to locate metallic anomalies buried close to the ground surface, but they suffer from high false alarm rates [3]. This makes the demining of contaminated lands difficult, dangerous, slow and very costly processes. Ground penetrating radar (GPR) is an alternative technology for landmine detection that has been extensively researched, although it is not yet widely used in practice. GPR has the potential to be much more effective than metal detectors in locating plastic cased landmines which have little or no metal content [1], [2], [4]. Ground Penetrating Radar (GPR) is a geophysical method based on electromagnetic wave propagation commonly used for non-destructive subsurface imaging [5]. It was originally used for ground geological investigation and the identification of major discontinuities within rock beds [6]. It quickly became of common use to detect buried objects in civil engineering, forensic applications, archeology [6], and mine detection [7], [8]. Several studies have also been conducted to assess the geological hazards associated with fault zones and all of them showed that GPR can give a clear image of subsurface fractures in different geological environments. Impulse time-domain Ground Penetrating Radar (GPR) is capable of detecting shallowly buried landmines that have little or no metal content, provided that adequate contrast or discontinuity exists between the dielectric properties of the landmine and the surrounding soil. GPR radiates short-duration pulses of electromagnetic energy into the ground and records backscattered signatures composed of reflections from the target dielectric surfaces. In the process of acquiring GPR signatures, the receive and transmit antennas are moved over the ground surface at approximately constant velocity and height in such a way as to cross the center of the buried object, and the backscattered signals are collected in fixed-time intervals. As a result, three types of GPR data survey, A-scan, B-scan and C-scan are obtained [7], [8]. The signatures of shallowly buried landmines measured using GPR are normally obscured by a strong background signal comprised of reflections from the ground surface, surrounding noise and the antenna crosstalk, called clutter. Clutter may be defined as in [7] the clutter that affects the GPR system generally are” … those signals that are unrelated to the target scattering characteristics but occur in the same sample time window and have similar spectral characteristic to the target wavelet”. The statistics of this signal depend on the environmental conditions such as soil type, amount of moisture and composition and roughness of the ground surface and is usually assumed to have a slow spatial variation. In this discussion, we will only deal with a simulated ground penetrating radar data where buried mine-like objects of various characteristics and shapes buried at different depths in Proceedings of 2011 International Conference on Signal Processing, Communication, Computing and Networking Technologies (ICSCCN 2011) 978-1-61284-653-8/11/$26.00 ©2011 IEEE 133