Adaptive boxcar background filtering for real-time GPR utility detection

Ground Penetrating Radar (GPR) investigation tools are increasingly used for real-time detection of underground utilities. Background noise is an annoying problem, because it sometimes masks the reflection from objects of interest. This study introduces an efficient background removal algorithm, which is so simple that it can be incorporated into GPR logging devices. The algorithm is based on the recently published outliers-out algorithm for stacking seismic data. Experiments conducted on both synthetic and field GPR data show that proposed background removal algorithm yields much better results than the commonly used average trace subtraction algorithm in a relatively comparable computational time.