UWB-GPR Data Processing for Identification of Anti-personnel Landmines under Rough Ground Surface

A process for finding buried landmines by using GPRs is divided into two stages, detection and identification. In the detection stage, all buried objects including the desired landmines together with other clutter objects are detected. In the identification stage, the detected objects are classified and the landmines are differentiated from the other objects. In this research, we focus on the identification stage and propose a GPR data processing for identification of shallowly buried anti-personnel landmines. After suppressing the ground surface reflection, we extract features related to waveform correlation from the GPR data. As an identification algorithm, an approach based on a matched filter is employed. In order to check the identification performance, Monte Carlo simulations are carried out using dataset generated by a 2D-FDTD method. Results are shown in the form of ROC curves and AUC values, and effects of soil condition (i.e. ground surface roughness, inhomogeneity, and moisture in soil) on identification performance are assessed.