| Presentation | 2021-10-22 Differences in Classification Accuracy of Landslide Hazard using Fixed-point Observation Images due to Network and Image Processing in Deep Learning Keisuke Tokumoto, Makoto Kobayashi, Koichi Shin, Masahiro Nishi, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In recent years，several landslides included by heavy rains have caused a lot of human damage in Hiroshima. Early evacuation is necessary to prevent human damage caused by landslides. In order to encourage early evacuation，we have installed a camera system at the dangerous area of landslides to obtain real-time images there.The real-time image can be viewed on a web page.In addition，it would be good to display a numerical index of danger along with this image. In this paper，we attempted to classify the degree of risk using deep learning for fixed-point images observed at dangerous area.In our trial，we measured and evaluated the accuracy with a more accurate network and image processing，as well as representative values that are closer to the actual danger.The networks of 10 popular models were evaluated for image classification.However，these networks have not been constructed on the assumption that they will be applied to fixed-point observation images.Therefore，the purpose is to consider a network structure suitable for fixed-point images by measuring the accuracy.And we aims to extract the intensity of changes in the water surface by previously processing images used for deep learning using images with different times.The purpose of the representative value is to calculate a value that is closer to the actual danger than the conventional method.These methods were examined using images obtained by actual observation. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Deep Learning / Landslide Disaster / Hazard Classification / Image Processing |
| Paper # | ICTSSL2021-26 |
| Date of Issue | 2021-10-14 (ICTSSL) |
| Conference Information | |
| Committee | ICTSSL / IEE-SMF / IN |
|---|---|
| Conference Date | 2021/10/21(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Online |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Koichi Gyoda(Shibaura Inst. of Tech.) / / Kenji Ishida(Hiroshima City Univ.) |
| Vice Chair | Munenari Inoguchi(Toyama Univ.) / Tomotaka Wada(Kansai Univ.) / / Kunio Hato(Internet Multifeed) |
| Secretary | Munenari Inoguchi(Synspective) / Tomotaka Wada(Hiroshima City Univ.) / / Kunio Hato(NTT) |
| Assistant | Shunichi Yokoyama(Shinshu Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Information and Communication Technologies for Safe and Secure Life / Technical Committee on Smart Facilities / Technical Committee on Information Networks |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Differences in Classification Accuracy of Landslide Hazard using Fixed-point Observation Images due to Network and Image Processing in Deep Learning |
| Sub Title (in English) | |
| Keyword(1) | Deep Learning |
| Keyword(2) | Landslide Disaster |
| Keyword(3) | Hazard Classification |
| Keyword(4) | Image Processing |
| 1st Author's Name | Keisuke Tokumoto |
| 1st Author's Affiliation | Hiroshima City University(Hiroshima City Univ.) |
| 2nd Author's Name | Makoto Kobayashi |
| 2nd Author's Affiliation | Hiroshima City University(Hiroshima City Univ.) |
| 3rd Author's Name | Koichi Shin |
| 3rd Author's Affiliation | Hiroshima City University(Hiroshima City Univ.) |
| 4th Author's Name | Masahiro Nishi |
| 4th Author's Affiliation | Hiroshima City University(Hiroshima City Univ.) |
| Date | 2021-10-22 |
| Paper # | ICTSSL2021-26 |
| Volume (vol) | vol.121 |
| Number (no) | ICTSSL-208 |
| Page | pp.pp.48-53(ICTSSL), |
| #Pages | 6 |
| Date of Issue | 2021-10-14 (ICTSSL) |