Deep Neural Network-Based Permittivity Inversions for Ground Penetrating Radar Data

Yintao Ji,Fengkai Zhang,Jing Wang,Zhengfang Wang,Peng Jiang,Hanchi Liu,Qingmei Sui
DOI: https://doi.org/10.1109/jsen.2021.3050618
IF: 4.3
2021-01-01
IEEE Sensors Journal
Abstract:Ground penetrating radar methods have been widely used for geological surveys, nondestructive inspections, advanced detections, and other subsurface structural detection processes. In this study, in order to fulfill the inversion processes for the aforementioned applications and reconstruct the permittivity of geo-structures with different actual sizes, a Deep Neural Network based inversion network was proposed to invert the relative permittivity of geo-structures from GPR B-scans. A network referred to as the Permittivity Inversion Network (PINet) utilizes a time dimension compression operation to address attenuation induced issues. In addition, it employs a global feature encoder in order to extract the global information, as well as automatically learn the spatial alignments between the GPR data and the permittivity images. This study constructed a universal dataset containing models with different actual sizes and target abnormities with different shapes and permittivity, for the purpose of training and validating the PINet. The inversion network was first validated using simulated GPR data. The inversion results demonstrated that the PINet was capable of accurately reconstructing permittivity images from GPR data with different central frequencies. Moreover, the performance of the PINet was verified using real GPR data. The position, rough shape and permittivity of the targets can be reconstructed.
What problem does this paper attempt to address?