Layered Media Inversion Network Applied in Ground Penetrating Radar

Ground-penetrating radar (GPR) is a mainstream detection tool for layered media inversion. However, due to the multiple reflections and complicated refraction effect, traditional inversion method can't achieve accurate parameter inversion. For the issue, a transformer-based layered media inversion network with GPR is proposed in the paper. The network utilizes velocity spectrum as the data set to enhance feature identification and utilizes transformer-based self-attention mechanism as the backbone to reallocate attention resources based on the degree of feature importance. It focuses more attention on reflected features adaptively and maps the velocity spectrum to the permittivity image. Extensive simulated results demonstrate that the proposed network has the superiority in reconstructing the underground structure and inversing the multilayered parameters.