Title:
METHOD AND APPARATUS FOR DETECTING METAL SURFACE DEFECTS
Document Type and Number:
WIPO Patent Application WO/2022/160170
Kind Code:
A1
Abstract:
The present application relates to a method for detecting metal surface defects, the method comprising: acquiring surface defect images of metal so as to obtain a metal surface defect dataset; building a first detection model by using the metal surface defect dataset; training a ResNet by using an ImageNet dataset; transferring the trained ResNet to the first detection model so as to obtain a second detection model, wherein the second detection model is a metal surface defect detection model based on deep transfer learning; and carrying out detection on a metal surface by using the second detection model. The present application further relates to an apparatus for detecting metal surface defects. In the present application, a ResNet trained on the basis of ImageNet is transferred to a metal surface defect detection model built in the present invention. By means of transfer learning, the metal surface defect detection model has gained a more powerful feature expression ability.
More Like This:
Inventors:
XIONG CHANGWEI (CN)
SHU YUFENG (CN)
CAO HUIYUAN (CN)
YU XIAOQIONG (CN)
LIU YUAN (CN)
LV WEIHAN (CN)
MO TIANXU (CN)
SHU YUFENG (CN)
CAO HUIYUAN (CN)
YU XIAOQIONG (CN)
LIU YUAN (CN)
LV WEIHAN (CN)
MO TIANXU (CN)
Application Number:
PCT/CN2021/074114
Publication Date:
August 04, 2022
Filing Date:
January 28, 2021
Export Citation:
Assignee:
DONG GUAN POLYTECHNIC (CN)
International Classes:
G06K9/62; G06N3/02
Foreign References:
CN109447018A | 2019-03-08 | |||
CN112184619A | 2021-01-05 | |||
CN109767427A | 2019-05-17 | |||
CN111523540A | 2020-08-11 | |||
CN112150407A | 2020-12-29 | |||
US20200175352A1 | 2020-06-04 |
Attorney, Agent or Firm:
BEIJING XIANGSHI INTELLECTUAL PROPERTY AGENCY (GENERAL PARTNERSHIP) (CN)
Download PDF: