DMRF-UNet: A Two-Stage Deep Learning Scheme for GPR Data Inversion Under Heterogeneous Soil Conditions

Traditional ground-penetrating radar (GPR) data inversion leverages iterative algorithms that suffer from high computation costs and low accuracy when applied to complex subsurface scenarios. Existing deep learning-based methods focus on the ideal homogeneous environments ignore interference due clutters noise in real-world heterogeneous environments. To address these issues, a two-stage neural network (DNN), called DMRF-UNet, is proposed reconstruct permittivity distributions of objects GPR B-scans under soil conditions. In first stage, U-shape DNN with multi-receptive-field convolution (MRF-UNet1) built remove inhomogeneity soil. Then, denoised B-scan MRF-UNet1 combined noisy be inputted second (MRF-UNet2). MRF-UNet2 learns inverse mapping relationship reconstructs distribution objects. avoid information loss, an end-to-end training method combining loss functions two stages introduced. A wide range scenarios are generated evaluate performance. The test results numerical experiment real measurement show permittivities, shapes, sizes, locations accuracy. comparison existing demonstrates superiority methodology for