Ultrawideband Synthetic Aperture Radar Unexploded Ordnance Detection

Abstract

Airborne ultrawideband (UWB) synthetic aperture radar (SAR) can perform wide-area detection of unexploded ordnance (UXO) to locate former bombing ranges efficiently. Two main issues in UWB SAR UXO detection, feature extraction, and discriminator design are considered. A space-wavenumber distribution and moment invariants-based method is proposed to extract the multi-aspect feature of UXO with both amplitude and spatial distribution information. Based on the extracted feature, a support vector machine (SVM) with hypersphere classification boundary, referred to as HS-SVM, is used as the UXO discriminator, which can be trained with a small training set of only UXO samples. Furthermore, the problem of HS-SVM kernel choice is studied, and the hidden Markov model (HMM) kernel is proved to be better than the Gaussian kernel. The efficiency of the proposed feature extraction method and the HMM kernel HS-SVM is validated using real data collected by a UWB SAR system.