METHOD FOR DYNAMIC TARGET TRACKING BY LEGGED ROBOT

Title:

METHOD FOR DYNAMIC TARGET TRACKING BY LEGGED ROBOT

Document Type and Number:

WIPO Patent Application WO/2024/093209

Kind Code:

A1

Abstract:

Disclosed in the present invention is a method for dynamic target tracking by a legged robot. Firstly, images and videos in walking processes of pedestrians wearing masks, not wearing masks, and incorrectly wearing masks are collected to build a data set used for pedestrian mask wearing detection. Secondly, a target recognition module based on a recurrent neural network is used to determine whether a pedestrian wears a mask, and acquire face position coordinates of a pedestrian not wearing a mask in a current frame. Thirdly, a residual network is used as a reference network to extract a face semantic feature of the pedestrian not wearing a mask in the current image frame, and predict a face semantic feature of the pedestrian in a next image frame. Finally, a target tracking module based on a siamese network is designed to, by calculating a correlation coefficient between pedestrian face position semantic feature mapping in the current frame and the next frame, track the pedestrian not wearing a mask, so as to implement tracking of a dynamic target by the legged robot. The present invention can be applied to inspection processes of legged robots under epidemic prevention and control conditions.

More Like This:

WO/2023/004250	BIOMETRIC DETECTION USING PHOTODETECTOR ARRAY
WO/2024/012607	PERSONNEL DETECTION METHOD AND APPARATUS, DEVICE, AND STORAGE MEDIUM
JP7449408	Electronic devices for automatic identification of users

Inventors:

CAO ZHENGCAI (CN)
LI JUNNIAN (CN)
SHAO SHIBO (CN)
ZHANG DONG (CN)
JU RENJIE (CN)
DU GUOFENG (CN)

Application Number:

PCT/CN2023/096170

Publication Date:

May 10, 2024

Filing Date:

May 25, 2023

Export Citation:

Click for automatic bibliography generation Help

Assignee:

UNIV BEIJING CHEM TECH (CN)

International Classes:

G06V40/10; G06V10/44; G06V10/62; G06V10/82; G06V20/00

Foreign References:

CN115909396A	2023-04-04
CN113902991A	2022-01-07
CN114550269A	2022-05-27

Other References:

LUCA BERTINETTO; JACK VALMADRE; JO\~AO F. HENRIQUES; ANDREA VEDALDI; PHILIP H. S. TORR: "Fully-Convolutional Siamese Networks for Object Tracking", ARXIV.ORG, 1 December 2021 (2021-12-01), XP091107012

Attorney, Agent or Firm:

BEIJING SIHAI TIANDA INTELLECTUAL PROPERTY AGENT LTD. (CN)

Download PDF:

View/Download PDF PDF Help

Previous Patent: MOBILE TERMINAL POCKET MODE IDENTIFICATION METHOD, MOBILE TERMINAL AND STORAGE MEDIUM

Next Patent: COMPOSITE MAGNET ASSEMBLY BASED ON PERMANENT MAGNET AND ELECTROMAGNET, AND ASSEMBLY DESIGN METHOD TH...