Discrimination of buried objects using time-frequency analysis and waveform norms

Ground Penetrating Radar (GPR) are widely used to probe the sub-surface. Recently, various time-frequency analyses has been proposed to discriminate buried land mines from other clutter objects and thus reduce GPR false alarm rates. This paper examines the possibility for discrimination and assesses it experimentally. The approach uses the Choi-Williams time-frequency transform to analyse ultra-wideband signal returns from a range of shallow buried objects. Single Value Decomposition is performed on isolated object time-frequency signatures. The signatures are evaluated using a set of waveform norms that discriminate in time, frequency and energy content. The results indicate that this approach could improve land mine detection rates and reduce false alarms.