A Generative Adversarial Network for Enhancing the Resolution of GPR B-Scans

Ground penetrating radar (GPR) is widely used to image the interiors of the structures such as tree trunks with internal defects, reinforced concrete columns, wooden utility poles, silos, and storage tanks. During the non-destructive testing or inspection of such structures, GPR is typically moved along a circular trajectory around these structures while emitting electromagnetic waves and receiving the reflected ones at each scanning point. The density of scanning points directly impacts the resolution of the GPR B-scans, with a large number of scanning points yielding high-resolution B-scans and images of the structures' interiors. That being said, performing a GPR survey on a large number of scanning points requires more accurate and reliable positioning systems to facilitate measurements (J. Qian, Y. H. Lee, K. Cheng, D. Qi, M. L. M. Yusof, D. Lee, and A. C. Yucel, “A Deep Learning-Augmented Stand-off Radar Scheme for Rapidly Detecting Tree Defects,” IEEE Trans. Geosci. Remote Sens., vol. 62, page 1–15, Jun. 2024), and increases the time for the data collection. To address this challenge, conventional interpolation methods have been applied to smooth and refine the features in B-scans (R. J. Oliveira, B. Caldeira, T. Teixido, J. F. Borges, and A. Carneiro, “Increasing the Lateral Resolution of 3D-GPR Datasets through 2D-FFT Interpolation with Application to a Case Study of the Roman Villa of Horta da Torre (Fronteira, Portugal),”, Remote Sens., vol. 14, issue 16, Aug. 2022). However, such methods often fail to capture subtle features related to reflections from internal defects or irregularities, limiting their effectiveness in resolution enhancement. Therefore, a more robust method is needed to improve the resolution of GPR B-scans.