Study of background subtraction for ground penetrating radar

In this paper, study of various background subtraction or ground bounce removal methods for GPR or for subsurface utility mappinghas been performed. Mean, median, SVD, PCA, ICA and Training methods have been evaluated and tested on number of GPR data sets collected with a horn antenna for variety of nonlinear rough surfaces and for various soil moisture contents. A performance evaluation criterion for analyzing the different algorithms has been introduced. Each algorithm relative benefits and drawbacks has been determined. In preprocessing stage, GPR imaging influence by the various ground bounce removal techniques. Based on this dependency, variety of ground surfaces have been analyzed by indeginousiely developed GPR unit and collected data for bare rough ground surface.