Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
SURFACE DEFECT DETECTION METHOD, APPARATUS, SYSTEM, STORAGE MEDIUM, AND PROGRAM PRODUCT
Document Type and Number:
WIPO Patent Application WO/2022/127919
Kind Code:
A1
Abstract:
Disclosed by embodiments of the present application are a surface defect detection method, apparatus, system, storage medium, and program product, belonging to the technical field of deep learning. In embodiments of the present application, if it is detected and determined by means of a neural network defect segmentation model that no surface defects of known defect types are present in an image to be detected, then extracting, by means of a neural network defect feature extraction model, an image feature of the image to be detected to obtain a feature to be compared; if the similarity between the feature to be compared and a normal data representation feature is relatively small, then determining the presence of a surface defect of an unknown defect type in the image to be detected, that is, the present solution is capable of detecting surface defects of unknown defect types, improving the accuracy of detection.

Inventors:
ZHANG YINGYING (CN)
ZHONG QIAOYONG (CN)
XIE DI (CN)
PU SHILIANG (CN)
Application Number:
PCT/CN2021/139333
Publication Date:
June 23, 2022
Filing Date:
December 17, 2021
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
HANGZHOU HIKVISION DIGITAL TEC (CN)
International Classes:
G06T7/00
Foreign References:
CN110533058A2019-12-03
CN110619618A2019-12-27
CN111652319A2020-09-11
CN104751198A2015-07-01
CN105719291A2016-06-29
US20210279858A12021-09-09
US20200111217A12020-04-09
CN202011495050A2020-12-17
Other References:
LI, CHAO ET AL.: "Effective Method of Weld Defect Detection and Classification based on Machine Vision", COMPUTER ENGINEERING AND APPLICATIONS, vol. 54, no. 6, 31 December 2018 (2018-12-31), pages 264 - 270, XP055944608
Attorney, Agent or Firm:
BEIJING SAN GAO YONG XIN INTELLECTUAL PROPERTY AGENCY CO., LTD. (CN)
Download PDF: