VIBWNet: An Efficient GPR Clutter Suppression Method for Tunnel Linings With Double-Layered Rebar Mesh and Heterogeneous Concrete

Ground-Penetrating Radar (GPR) frequently encounters substantial clutter interference, which hinders the precise identification of concealed defects in tunnel linings, particularly in scenarios involving double-layered rebar and heterogeneous concrete. In this paper, an efficient clutter suppression method is proposed to address this challenge. Initially, the method employs two-dimensional variational mode decomposition (2D-VMD) for data augmentation, thereby enhancing the representation of rebar-related clutter signals. Subsequently, the VIBWNet network is proposed to suppress clutter, enabling the restoration and enhancement of defect signals. Additionally, this study incorporates heterogeneous concrete models during dataset construction to ensure a closer match with real-world heterogeneous concrete conditions, thereby substantially enhancing the network’s generalization capability when processing real-world measured data. Furthermore, evaluations on sandbox and concrete block experiments confirm the method possesses strong generalization capability for measured data, demonstrating its effectiveness in suppressing complex clutter while accurately restoring defect signals. Finally, quantitative ablation studies and visual comparisons of network variants are conducted to elucidate the specific contributions of each component, while comparative analyses against recognized unsupervised baselines demonstrate the superiority of the proposed method.