| Presentation | 2019-10-18 Study of Recognizing Road Surface Conditions using Deep Learning Applied for RGBD images Obtained from an Environmental Monitoring Robot Takuya Hayashi, Takeo Kaneko, Junya Morimoto, Junji Yamato, Hiroyuki Ishii, Jun Ohya, Atsuo Takanishi, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | An environmental monitoring robot that moves safely and autonomously needs a function to recognize the state of the ground area. This paper presents a novel two-stage approach for recognizing road surface condition from RGBD image sensed using Kinect v2. The first stage, based on convolutional neural network (CNN) classifier, lists the candidates for the obstruction class and road surface class label (attribute) information for each pixel from the RGB information. The second stage, a determinator using 3-D point clouds information (height or surface curvature features), generates an estimated label image by determining whether it is an obstacle or a road surface class. In this paper, we evaluated SegNet-Basic and ENet as a comparison of the CNN models used in the approach. In the evaluation, we used data collected for the actual natural environment. Experimental results show the effectiveness of the action determination for autonomous search based on road surface recognition using ENet network model and height information obtained from point cloud information. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Deep Neural Network / SegNet-basic / ENet / 3-D Point Clouds / Recognizing Road Surface Conditions |
| Paper # | PRMU2019-39 |
| Date of Issue | 2019-10-11 (PRMU) |
| Conference Information | |
| Committee | PRMU |
|---|---|
| Conference Date | 2019/10/18(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Yoichi Sato(Univ. of Tokyo) |
| Vice Chair | Toru Tamaki(Hiroshima Univ.) / Akisato Kimura(NTT) |
| Secretary | Toru Tamaki(NTT) / Akisato Kimura(OMRON SINICX) |
| Assistant | Yusuke Uchida(DeNA) / Takayoshi Yamashita(Chubu Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Pattern Recognition and Media Understanding |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Study of Recognizing Road Surface Conditions using Deep Learning Applied for RGBD images Obtained from an Environmental Monitoring Robot |
| Sub Title (in English) | Comparative Studies of SegNet-Basic and ENet as well as Height and Surface Curvature Features |
| Keyword(1) | Deep Neural Network |
| Keyword(2) | SegNet-basic |
| Keyword(3) | ENet |
| Keyword(4) | 3-D Point Clouds |
| Keyword(5) | Recognizing Road Surface Conditions |
| 1st Author's Name | Takuya Hayashi |
| 1st Author's Affiliation | Waseda University(Waseda Univ.) |
| 2nd Author's Name | Takeo Kaneko |
| 2nd Author's Affiliation | Waseda University(Waseda Univ.) |
| 3rd Author's Name | Junya Morimoto |
| 3rd Author's Affiliation | Waseda University(Waseda Univ.) |
| 4th Author's Name | Junji Yamato |
| 4th Author's Affiliation | Kogakuin University(Kogakuin Univ.) |
| 5th Author's Name | Hiroyuki Ishii |
| 5th Author's Affiliation | Waseda University(Waseda Univ.) |
| 6th Author's Name | Jun Ohya |
| 6th Author's Affiliation | Waseda University(Waseda Univ.) |
| 7th Author's Name | Atsuo Takanishi |
| 7th Author's Affiliation | Waseda University(Waseda Univ.) |
| Date | 2019-10-18 |
| Paper # | PRMU2019-39 |
| Volume (vol) | vol.119 |
| Number (no) | PRMU-235 |
| Page | pp.pp.41-46(PRMU), |
| #Pages | 6 |
| Date of Issue | 2019-10-11 (PRMU) |