Automatic subsurface unexploded ordnance detection using a wideband full-polarization ground penetrating radar and entropy-based polarimetric signatures

Abstract

An automatic buried unexploded ordnance (UXO) classification methodology based on the H-α entropy-based polarimetric features extracted from wideband (50–600 MHz) fully polarimetric ground penetrating radar (GPR) data is presented. This method first established a background H-α feature of the whole test site for determining the classification boundary of random and isotropic scatters. Then UXO-like elongated targets are detected by the presence of sufficient H-α feature appearing in the linear scattering classification zone. The detail processing and detection procedure will be discussed. The performance of the proposed UXO classifier will be evaluated using receiver operating characteristic (ROC) curve based actual field data collected at the US government Fort Ord UXO test site setup for evaluating different UXO detection technologies.