| Presentation | 2017-01-26 Object Identification by Ground Penetrating Radar using Deep Learning with Radar Images using FDTD Method on GPU Cluster Jun Sonoda, Tomoyuki Kimoto, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Recently, deteriorations of social infrastructures such as tunnels and bridges become a serious social problem. It is required to properly and rapidly detect for abnormal parts of the social infrastructures. The ground penetrating radar (GPR) is efficient for the social infrastructure inspection. However, it is difficult to identify material and size of the underground object from the radar image obtained the GPR. In this study, to objectively and quantitatively inspect from the GPR images by the deep learning, %for a social infrastructure using the GPR, we have automatically and massively generated the GPR images by a fast finite-difference time-domain (FDTD) simulation on graphics processing units (GPUs), and have learned the underground object using the generated GPR images by a deep convolutional neural network (CNN). It is shown that we have obtained five layers CNN can identify five materials and size with roughly 80 % accuracy in in-homogeneous underground media. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Deep learning / convolutional neural network / ground penetrating radar / FDTD method / GPU / object identification |
| Paper # | SANE2016-103 |
| Date of Issue | 2017-01-19 (SANE) |
| Conference Information | |
| Committee | SANE |
|---|---|
| Conference Date | 2017/1/26(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Nagasaki Prefectural Art Museum |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | Positioning, navigation, Radar and general |
| Chair | Hirokazu Kobayashi(Osaka Inst. of Tech.) |
| Vice Chair | Takahide Mizuno(JAXA) / Toshifumi Moriyama(Nagasaki Univ.) |
| Secretary | Takahide Mizuno(JAXA) / Toshifumi Moriyama(Mitsubishi Electric) |
| Assistant | Atsushi Kezuka(ENRI) / Manabu Akita(Univ. of Electro-Comm.) |
| Paper Information | |
| Registration To | Technical Committee on Space, Aeronautical and Navigational Electronics |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Object Identification by Ground Penetrating Radar using Deep Learning with Radar Images using FDTD Method on GPU Cluster |
| Sub Title (in English) | |
| Keyword(1) | Deep learning |
| Keyword(2) | convolutional neural network |
| Keyword(3) | ground penetrating radar |
| Keyword(4) | FDTD method |
| Keyword(5) | GPU |
| Keyword(6) | object identification |
| 1st Author's Name | Jun Sonoda |
| 1st Author's Affiliation | National Institute of Technology, Sendai College(NIT, Sendai College) |
| 2nd Author's Name | Tomoyuki Kimoto |
| 2nd Author's Affiliation | National Institute of Technology, Oita College(NIT, Oita College) |
| Date | 2017-01-26 |
| Paper # | SANE2016-103 |
| Volume (vol) | vol.116 |
| Number (no) | SANE-427 |
| Page | pp.pp.41-46(SANE), |
| #Pages | 6 |
| Date of Issue | 2017-01-19 (SANE) |