Title:
OBSTACLE DETECTION METHOD, APPARATUS AND DEVICE, AND MEDIUM
Document Type and Number:
WIPO Patent Application WO/2021/068573
Kind Code:
A1
Abstract:
Disclosed is an obstacle detection method. The method comprises: acquiring a first image, wherein the first image is an image encoded on the basis of a three-primary-color light mode; reconstructing the first image to obtain a second image, wherein the second image is a hyperspectral image; extracting a hyperspectral characteristic from the hyperspectral image, and classifying candidate objects in the hyperspectral image according to the hyperspectral characteristic to obtain an obstacle detection result. Since hyperspectral characteristics corresponding to different materials are different, the candidate objects in the hyperspectral image are classified on the basis of the hyperspectral characteristics, such that an object that is of a similar color but is made of a different material can be distinguished. On the basis of this, an obstacle which is the same color as an environmental color or which is a color similar to the environmental color can be detected, and a missed detection rate is reduced. In addition, in the method, a hyperspectral image can be obtained by means of reconstructing an image by using only an ordinary camera, and an imaging spectrometer does not need to be used, and therefore, the cost is relatively low, and the method has relatively high feasibility. Further disclosed are a corresponding apparatus and device, and a medium.
Inventors:
ZHOU WEI (CN)
Application Number:
PCT/CN2020/100051
Publication Date:
April 15, 2021
Filing Date:
July 03, 2020
Export Citation:
Assignee:
HUAWEI TECH CO LTD (CN)
International Classes:
G06V10/14; G06V10/50; G06V10/58; G06V10/60; G06V10/778; G06V10/82; G06V20/58
Foreign References:
| CN110188606A | 2019-08-30 | |||
| CN102609963A | 2012-07-25 | |||
| CN108108721A | 2018-06-01 | |||
| US20070003165A1 | 2007-01-04 | |||
| CN201910954529A | 2019-10-09 |
Other References:
YU LIFU: "Research on Hyperspectral Image Spectral–Spatial Classification Based on Deep Autoencoder Network", ELECTRONIC TECHNOLOGY & INFORMATION SCIENCE, CHINA MASTER’S THESES FULL-TEXT DATABASE, 1 June 2017 (2017-06-01), XP055798729
See also references of EP 4030338A4
See also references of EP 4030338A4
Attorney, Agent or Firm:
SHENPAT INTELLECTUAL PROPERTY AGENCY (CN)
Download PDF: