Automatic recognition and tracking of highway layer-interface using Faster R-CNN

• Published in: Journal of applied geophysics Vol. 196; p. 104477
• Main Authors: Cui, Fan, Ning, Muwei, Shen, Jiawei, Shu, Xincheng
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.01.2022
• Subjects: Deep learning; Faster R-CNN; Ground penetrating radar; Highway layer-interface detection; Object detection; Deep learning; Highway layer-interface detection; Ground penetrating radar; Faster R-CNN; Object detection
• ISSN: 0926-9851; 1879-1859
• DOI: 10.1016/j.jappgeo.2021.104477

•A dataset of the highway layer-interface that was manually detected was created.•The highway layer-interface detection model based on Faster R-CNN framework.•The proposed model has 98.3% precision in detecting layer-interface in GPR profiles.•The proposed model can automatically identify and track highway layer-interface. Ground Penetrating Radar (GPR) is a non-destructive and efficient underground survey equipment, which has been widely used in the field of highway safety.Nowadays, it is common to manually interpret the GPR data detected on the highway to find the layer-interface. But, its efficiency and accuracy can no longer meet the needs of various engineering projects. In this manuscript, we propose a geological layer-interface detection framework based on Faster R-CNN, which is used to ensure the safety of highway traffic. The framework can realize the needs of end-to-end and real-time detection. We conduct a lot of experiments on the dataset of highway GPR profile, and the results show that the framework based on Faster R-CNN can be highly effectively used for highway underground layer-interface detection, and the precision of the model can reach 98.30%.