In this study, a novel method based on ground penetration radar (GPR) is proposed to categorize underground objects by using both B-scan and C-scan images. Three-dimensional GPR data obtained from a multichannel GPR system are reconstructed into a two-dimensional (2D) grid image which consists of several B-scan and C-scan images. Three-dimensional shape information of an underground object can be well represented in 2D grid image. The 2D grid images are then trained using deep convolutional neural networks (CNN) that is a state-of-the-art technique for image classification problem. The proposed method is validated through field applications on urban roads in Seoul, South Korea.

• Namgyu Kim(Department of Civil and Environmental Engineering, Sejong University)
• Ki-Deok Kim(Department of Civil and Environmental Engineering, Sejong University)
• Yun-Kyu An(Department of Architectural Engineering, Sejong University)
• Hyun-Jong Lee(Department of Civil and Environmental Engineering, Sejong University)
• Jong-Jae Lee(Department of Civil and Environmental Engineering, Sejong University)