Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
ANOMALY DETECTION SYSTEM, LEARNING DEVICE, ANOMALY DETECTION PROGRAM, AND LEARNING PROGRAM
Document Type and Number:
WIPO Patent Application WO/2022/137841
Kind Code:
A1
Abstract:
The present invention ensures stable determination accuracy irrespective of image size during anomaly detection for detecting a defect in the appearance of an object. An anomaly detection system 100 comprises: an input unit 111 that inputs inspection images of a target object which have a plurality of types of image sizes greater than or equal to a prescribed size; a feature extraction unit 201 that has been trained in advance to extract a feature map 355 from training images including non-defective images of the target object; an image generation unit 202 that has been trained in advance to restore the training images from the feature map 355 extracted by the feature extraction unit 201; and a detection unit 116. The detection unit 116 detects an anomaly in the target object to be inspected on the basis of similarity calculated by comparing, with restoration images of the inspection images restored by the feature extraction unit 201 and the image generation unit 202, the inspection images of the target object, which have been input to the input unit 111 and which have the plurality of types of image sizes greater than or equal to the prescribed size.

Inventors:
OKUNO KIICHI (JP)
OKAZAKI TOMOYA (JP)
SASHIDA TAKEHIKO (JP)
TACHI KOKI (JP)
Application Number:
PCT/JP2021/040920
Publication Date:
June 30, 2022
Filing Date:
November 08, 2021
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
KONICA MINOLTA INC (JP)
International Classes:
G01N21/88; G06T7/00
Foreign References:
JP2019522897A2019-08-15
JP2017097718A2017-06-01
JP2020181532A2020-11-05
JP2019069145A2019-05-09
CN109615604A2019-04-12
Attorney, Agent or Firm:
HATTA & ASSOCIATES (JP)
Download PDF: