Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
IMAGE INSPECTION DEVICE, IMAGE INSPECTION METHOD, AND TRAINED MODEL GENERATION DEVICE
Document Type and Number:
WIPO Patent Application WO/2022/172468
Kind Code:
A1
Abstract:
This invention causes a feature amount of a good quality image including a special pattern to be plotted close to a feature amount of a good quality image, and causes a feature amount of an inspection image of a defective product to be plotted far from the feature amount of said good quality image. The invention comprises: an extraction unit 234 that extracts a feature amount of a synthesis inspection division image by inputting a synthesis inspection division image generated by synthesizing an inspection division image and label information corresponding to the inspection division image to a trained model which has been trained so as to take as input a synthesis good quality division image generated by synthesizing a good quality division image and label information corresponding to the good quality division image and output a feature amount of the synthesis good quality division image; an acquisition unit 235 that acquires a defect degree of the inspection division image corresponding to the synthesis inspection division image on the basis of a feature space formed by the extracted feature amount of the synthesis inspection division image and the feature amount of the synthesis good quality division image output at the time of training; and an inspection unit 236 that inspects an inspection subject on the basis of the acquired defect degree.

Inventors:
IKEDA YASUYUKI (JP)
Application Number:
PCT/JP2021/009412
Publication Date:
August 18, 2022
Filing Date:
March 10, 2021
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
OMRON TATEISI ELECTRONICS CO (JP)
International Classes:
G06T1/00; G01N21/88
Foreign References:
JP2018005773A2018-01-11
JP2020136368A2020-08-31
JP2006220648A2006-08-24
CN110570393A2019-12-13
Attorney, Agent or Firm:
INABA, Yoshiyuki et al. (JP)
Download PDF: