Effects of Soil Condition on Landmine Discrimination Using GPR Sensor

It is well known that a huge number of buried landmines in the post-war countries are still active, and that many civilians are killed or injured by landmine related accidents every year. Therefore, detection and clearance of buried landmines is a very important issue. Metal detectors are the almost universally available fielded devices for landmine detection, but they have great difficulty in detecting landmines that are made of plastic or with low metal content. On the other hand, Ground Penetrating Radar (GPR) is an emerging technique for landmine detection that can detect plastic or low metal content landmines. However, the GPR performs inadequately due to the ground clutter, because return from the shallowly buried landmine and that from ground surface overlap in time. Furthermore, the GPR also receives returns from other subsurface objects such as rocks, tree roots, or metal fragments in the ground, which yields high levels of false alarms. Generally speaking, the procedure for locating buried landmines using the GPR is divided into two stages, detection and identification. During the detection stage, all buried objects, including the desired landmines together with other objects such as stones, metal fragments, etc., are detected and their locations are specified. In the identification stage, the detected objects are classified and the landmines are differentiated from the other objects. The detection stage requires signal-processing techniques for radar target detection, whereas the identification stage requires pattern classification techniques. Compare with the detection process, the identification process is usually more complicated because considerable target information needs to be deduced from the total GPR data for reliable and accurate identification. Therefore, although many techniques for landmine detection when applied to GPR data have been proposed and designed to date, the development of reliable identification techniques is more important in reducing high incidence of false alarms. In this context, we here focus on the identification problem. In the identification problem, selection of features used for target classification plays a key part because the classification performance depends strongly upon these features chosen. Therefore, in this research, we propose some kinds of features for landmine discrimination and evaluate the effects of soil condition on the performance through Monte Carlo simulations using a dataset generated by a 2-Dimensional Finite Difference Time Domain (2D-FDTD) method.