A full waveform inversion with noise models in GPR detection

Noise in engineering is inevitable, and existing methods often eliminate weak target signals along with the noise, resulting in suboptimal accuracy in ground penetrating radar (GPR) detection. A full waveform inversion (FWI) with noise models is proposed. Noisy FWI is achieved by incorporating comprehensive noise models of electromagnetic propagation during GPR operations as prior information in the forward modeling of FWI. The mathematical models corresponding to the characteristics of thermal noise, antenna noise, surface and subsurface wrap-around noise, and multipath interference noise during GPR detection are first summarized and incorporated into the theoretical framework of noisy FWI. Then, the feasibility and robustness of noisy FWI are verified through numerical simulations. Finally, the measurements of concrete moisture content demonstrate the effectiveness of the noise FWI. This method improves GPR detection accuracy in noisy environments, providing a novel approach for subsurface characterization in geotechnical and civil engineering.