Title:
DEPTH CAMERA-BASED PEDESTRIAN FLOW ANALYSIS METHOD AND APPARATUS
Document Type and Number:
WIPO Patent Application WO/2022/193516
Kind Code:
A1
Abstract:
A depth camera-based pedestrian flow analysis method and apparatus, applicable to the field of machine vision. The method comprises: acquiring a pedestrian image set concerning a target detection region by means of a depth camera (S101), the pedestrian image set comprising multiple frames of pedestrian images; importing the pedestrian images into a human detection network, and outputting a human detection image corresponding to each of the multiple frames of pedestrian images (S102); determining, on the basis of the human detection images corresponding to multiple consecutive frames of pedestrian images, motion tracking data of a plurality of pedestrian objects present within the target detection region (S103); and generating pedestrian flow analysis information about the target detection region on the basis of the motion tracking data of the plurality of pedestrian objects (S104). The human detection network is trained by means of deep learning, a pedestrian can be accurately tracked, and the pedestrian image is a depth map, such that the problem that a color image may violate privacy of the pedestrian is avoided and information except for that in plane dimension is also displayed.
Inventors:
WANG WEIFANG (CN)
GONG GUOJI (CN)
ZHU YIBO (CN)
HU ZHENG (CN)
GONG GUOJI (CN)
ZHU YIBO (CN)
HU ZHENG (CN)
Application Number:
PCT/CN2021/107938
Publication Date:
September 22, 2022
Filing Date:
July 22, 2021
Export Citation:
Assignee:
ORBBEC INC (CN)
International Classes:
G06K9/00; G06T7/246; G06K9/62; G06N3/04; G06N3/08
Foreign References:
| CN113034544A | 2021-06-25 | |||
| CN110717408A | 2020-01-21 | |||
| CN111192293A | 2020-05-22 | |||
| CN111709974A | 2020-09-25 | |||
| CN108986064A | 2018-12-11 | |||
| US20190378287A1 | 2019-12-12 |
Attorney, Agent or Firm:
CHINA TRUER IP (CN)
Download PDF: