|
| | Abstract | | These days, multi-channel GPR (Ground-Penetrating Radar) systems are available to detect cavities underneath the road surface. A system on a vehicle runs faster than about 40 km/h and keeps acquiring a large amount of data. A certain level of the experience is necessary to analyze the acquired GPR data for identifying the types of reflectors, i.e., metal pipes, cavities and the others. The experts need to work long hours to process and analyze the GPR data. At the same time, the information technology tools of machine learnings, such as GPU computing and deep learning libraries have become easier to implement on a PC. In this study, the deep learning network was applied for the GPR data to distinguish the types of reflectors with the very limited training data. The result shows that the deep learning technique can distinguish the types of GPR reflectors to identify cavities, metal pipes and others. |
|
|
|
| |
|