A Deep Learning-Based GPR Forward Solver for Predicting B-Scans of Subsurface Objects

The forward full-wave modeling of ground-penetrating radar (GPR) facilitates the understanding and interpretation GPR data. Traditional solvers require excessive computational resources, especially when their repetitive executions are needed in signal processing and/or machine learning algorithms for data inversion. To alleviate burden, a deep learning-based 2D solver is proposed to predict B-scans subsurface objects buried heterogeneous soil. constructed as bimodal encoder-decoder neural network. Two encoders followed by an adaptive feature fusion module designed extract informative features from permittivity conductivity maps. decoder subsequently constructs fused representations. enhance network's generalization capability, transfer employed fine-tune network new scenarios vastly different those training set. Numerical results show that achieves mean relative error 1.28%. For predicting B-scan one object, requires 12 milliseconds, which 22,500x less than time required classical physics-based solver.