Automatic Detection and Depth Estimation of Subsurface Voids using Ground Penetrating Radar and Convolutional Neural Network

Summary Subsurface cavities are a major challenge that may potentially cause many problems, such as structural instability, safety risks, and environmental damage. Because of these significant impacts, cavities require early detection and monitoring to avoid material and human losses. Investigation and manual processing of these structures using traditional methods can be difficult and time-consuming. Therefore, automated approaches using deep learning algorithms for identifying subsurface anomalies have recently emerged, providing promising pathways for realtime cavity detection. This study proposes a framework based on a machine learning model capable of detecting and estimating subsurface cavities and void depth by passing a series of radargrams through a pre-trained Convolutional Neural Network (CNN). The model was trained on 1408 augmented B-scans collected with 200 and 400 MHz antennas. The results show impressive (100 %) success in classifying cavities, demonstrating the ability of the framework to recognize cavity structures in radargrams automatically. Furthermore, a correlation method between image pixels and spatial information was carried out to estimate the depth of the cavity directly from the radargram. The overall results demonstrate an accuracy of over 95 % in identifying and directly estimating cavity depth in radargrams, indicating its potential impact in various civil engineering applications.