TY - JOUR
T1 - Estimating the Orientation of 3D-Structured Switches Using the RGB-D Camera Attached to a Disaster Response Robot
AU - Kanda, Takuya
AU - Qi, Li
AU - Ohya, Jun
AU - Takanishi, Atsuo
AU - Ogata, Hiroyuki
AU - Hshimoto, Kenji
N1 - Funding Information:
The authors of this paper acknowledge to the support of Kazuya Miyakawa, Jeonghwang Hayashi, Kota Umebayashi, Reina Yoshizaki, Xiao Sun, Takashi Matsuzawa, Hiroshi Naito of Waseda University and Keishi Nishikawa of Mitsubishi Electric corporation.
Publisher Copyright:
© 2020. Int. J. Mech. Eng. Rob. Res.
PY - 2020/10
Y1 - 2020/10
N2 - Towards the actualization of autonomous switch operation method for the disaster response robot, it is necessary to estimate the position and orientation of the switch in the site. This paper proposes a method for estimating switch orientation information based on the RGB-D sensor information attached to the disaster response robot. The flow of the proposed method in this paper consists of the following three stages. (1) Acquisition of RGB-D information of the switch area using the detection result of YOLOv3. (2) Position estimation of switch operation part by CNN from RGB-D information of switch area. (3) Estimation of switch orientation information using switch operation part position information and information on wall plane on which the switch exists. In the experiment, the proposed method was applied to three types of switch boxes, and the calculation results of the orientation information were evaluated. It was found from the experimental results that the calculation with sufficient accuracy is possible.2020. Int. J. Mech. Eng. Rob. Res
AB - Towards the actualization of autonomous switch operation method for the disaster response robot, it is necessary to estimate the position and orientation of the switch in the site. This paper proposes a method for estimating switch orientation information based on the RGB-D sensor information attached to the disaster response robot. The flow of the proposed method in this paper consists of the following three stages. (1) Acquisition of RGB-D information of the switch area using the detection result of YOLOv3. (2) Position estimation of switch operation part by CNN from RGB-D information of switch area. (3) Estimation of switch orientation information using switch operation part position information and information on wall plane on which the switch exists. In the experiment, the proposed method was applied to three types of switch boxes, and the calculation results of the orientation information were evaluated. It was found from the experimental results that the calculation with sufficient accuracy is possible.2020. Int. J. Mech. Eng. Rob. Res
KW - disaster response robot
KW - image processing
KW - intelligence robotics
KW - object detection
KW - point cloud data processing
UR - http://www.scopus.com/inward/record.url?scp=85101735219&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85101735219&partnerID=8YFLogxK
U2 - 10.18178/ijmerr.9.10.1432-1438
DO - 10.18178/ijmerr.9.10.1432-1438
M3 - Article
AN - SCOPUS:85101735219
SN - 2278-0149
VL - 9
SP - 1432
EP - 1438
JO - International Journal of Mechanical Engineering and Robotics Research
JF - International Journal of Mechanical Engineering and Robotics Research
IS - 10
ER -