Noise suppressing and direct wave removal in GPR data based on shearlet transform

Abstract

Ground penetrating radar (GPR) is often used to detect buried objects and evaluate structural condition. However, the direct wave and random noise often influence the arrival-time detection and the target-position location. We present a new application of Shearlet transform (ShT) to GPR data processing for direct wave removal and random noise suppression. ShT is a non-adaptive geometric-analysis technique, which has the properties of multi-directions and multi-scale, so it can show the optimal representations of signals in higher dimensions. The original GPR data is transformed to the ShT domain. The direct wave and the remaining GPR signal are effectively separated. While we eliminate the direct wave, the GPR signal is not damaged. The Shearlet coefficients of the GPR signal are relatively large, whereas random noises are relatively small. So we can use the threshold algorithm depending on different scales and directions in the ShT domain to suppress random noise. The GPR signal can be preserved very well and SNR is enhanced.