Title:
THREE-DIMENSIONAL POINT CLOUD LABEL LEARNING DEVICE, THREE-DIMENSIONAL POINT CLOUD LABEL ESTIMATION DEVICE, THREE-DIMENSIONAL POINT CLOUD LABEL LEARNING METHOD, THREE-DIMENSIONAL POINT CLOUD LABEL ESTIMATION METHOD, AND PROGRAM
Document Type and Number:
Japanese Patent JP2019185665
Kind Code:
A
Abstract:
To label a point that makes up a target regardless of the type of target for a large point cloud with no limit on a range or score.SOLUTION: A three-dimensional point cloud label learning device 10A includes a ground height calculation unit for outputting a three-dimensional point cloud with a ground height as an inputted three-dimensional point cloud and ground height; an intensity-RGB transform unit for outputting the three-dimensional point cloud with the Intensity-RGB converted ground height using the ground height of the three-dimensional point cloud as the input; a super voxel clustering Unit for outputting a super voxel data with correct label with the Intensity-RGB converted ground height three-dimensional point cloud, learning point cloud label, and clustering hyperparameter as inputs; and a deep neural network learning unit for outputting learned deep neural network parameters using super voxel data and deep neural network hyper parameters with correct labels as inputs.SELECTED DRAWING: Figure 2
Inventors:
YAO YASUHIRO
NIIGAKI JIN
TSUTSUGUCHI KEN
KINEBUCHI TETSUYA
NIIGAKI JIN
TSUTSUGUCHI KEN
KINEBUCHI TETSUYA
Application Number:
JP2018079243A
Publication Date:
October 24, 2019
Filing Date:
April 17, 2018
Export Citation:
Assignee:
NIPPON TELEGRAPH & TELEPHONE
International Classes:
G06T7/00; G06V10/56; G06V10/764
Domestic Patent References:
| JP2017532092A | 2017-11-02 |
Foreign References:
| US20170109611A1 | 2017-04-20 |
Other References:
PAPON JEREMIE, ET AL.: "Voxel Cloud Connectivity Segmentation - Supervoxels for Point Clouds", 2013 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, JPN6021035643, 2013, pages 2027 - 2034, XP032493181, ISSN: 0004592890, DOI: 10.1109/CVPR.2013.264
LI BO: "3D Fully Convolutional Network for Vehicle Detection in Point Cloud", 2017 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELIGENT ROBOTS AND SYSTEMS, JPN6021035641, 2017, pages 1513 - 1518, XP033266106, ISSN: 0004592891, DOI: 10.1109/IROS.2017.8205955
LI BO: "3D Fully Convolutional Network for Vehicle Detection in Point Cloud", 2017 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELIGENT ROBOTS AND SYSTEMS, JPN6021035641, 2017, pages 1513 - 1518, XP033266106, ISSN: 0004592891, DOI: 10.1109/IROS.2017.8205955
Attorney, Agent or Firm:
Patent Service Corporation Taiyo International Patent Office