A Method for Detecting Shallow Cavities in Roadbeds Based on Deep Learning and Ground Penetrating Radar Measurement Data

In the course of road use, cavity disease in the roadbed area is often a direct threat to the structural stability of the road. This is because unfavourable geological conditions in various roadbed areas often end up forming cavities, making it an important issue to investigate how to intelligently detect cavities in roadbeds. To this end, an experimental road has been built specifically for radar detection and a large number of cavity models of varying sizes and depths have been embedded in the roadbed area of the experimental road. Using a known real model and a software simulation model built partly by the time-domain finite difference method as the object of study, a deep learning framework was used to construct a mapping relationship between ground-penetrating radar images and subsurface dielectric constant maps, identifying the limitations of the current study, which has served to advance the application of deep learning in the analysis of road ground-penetrating radar images.