Back to results
US 12,322,079 B2Granted
Inspection device, inspection method, and recording medium for inspecting an inspection target with machine learning
AssigneeMitsubishi Electric Corporation
InventorsYuichi Ikeda, Kazuyuki Miyazawa, Akira Minezawa, Yukihiro Toku
Life of this patent
Estimating the term…
Abstract
A position detector applies an image to be inspected to a position detection model generated by machine learning using image information of a training image to calculate a score for a portion of an image to be inspected. The position detector specifies a portion of the image where the calculated score is a score that satisfies a predetermined relationship to detect the position of a metal part in the image to be inspected. A position outputter outputs the position of the metal part detected by the position detector.
Claims19
An inspection device comprising:
- a position detector to detect a position of an inspection target in an inspection image by (i) extracting a plurality of portions of the inspection image on the basis of a rough detection result of the position of the inspection target calculated by applying the inspection image including the inspection target to a position rough detection model generated by machine learning using image information of a training image, (ii) applying the plurality of extracted portions of the inspection image to a position detection model generated by machine learning using the image information of the training image to calculate a position detection score for the plurality of extracted portions of the inspection image, (iii) specifying an extracted portion from the plurality of extracted portions of the inspection image where the position detection score satisfies a predetermined relationship, and then (iv) specifying a position of a pixel in the inspection image where a brightness value of a pixel among a plurality of pixels is greatest in a predetermined range centered on a position of the specified extracted portion of the inspection image and (v) detecting, based on the specified position of the pixel in the inspection image where the brightness value is greatest, the position of the inspection target in the inspection image;
- and
- a position outputter to output the position of the inspection target detected by the position detector.
An inspection method comprising:
- detecting a position of an inspection target in an inspection image by (i) extracting a plurality of portions of the inspection image on the basis of a rough detection result of the position of the inspection target calculated by applying the inspection image including the inspection target to a position rough detection model generated by machine learning using image information of a training image, (ii) applying the plurality of extracted portions of the inspection image to a position detection model generated by machine learning using the image information of the training image to calculate a position detection score for the plurality of extracted portions of the inspection image, (iii) specifying an extracted portion from the plurality of extracted portions of the inspection image where the position detection score satisfies a predetermined relationship, and then (iv) specifying a position of a pixel in the inspection image where a brightness value of a pixel among a plurality of pixels is greatest in a predetermined range centered on a position of the specified extracted portion of the inspection image and (v) detecting, based on the specified position of the pixel in the inspection image where the brightness value is greatest, the position of the inspection target in the inspection image;
- and
- outputting the position of the inspection target detected in the detecting.
A non-transitory computer-readable recording medium storing a program, the program causing a computer to function as:
- a position detector to detect a position of an inspection target in an inspection image by (i) extracting a plurality of extracted portions of the inspection image on the basis of a rough detection result of the position of the inspection target calculated by applying the inspection image including the inspection target to a position rough detection model generated by machine learning using image information of a training image, (ii) applying the plurality of extracted portions of the inspection image to a position detection model generated by machine learning using the image information of the training image to calculate a position detection score for the plurality of extracted portions of the inspection image, (iii) specifying an extracted portion from the plurality of extracted portions of the inspection image where the position detection score satisfies a predetermined relationship, and then (iv) specifying a position of a pixel in the inspection image where a brightness value of a pixel among a plurality of pixels is greatest in a predetermined range centered on a position of the specified extracted portion of the inspection image and (v) detecting, based on the specified position of the pixel in the inspection image where the brightness value is greatest, the position of the inspection target in the inspection image;
- and
- a position outputter to output the position of the inspection target detected by the position detector.
Dependencies are inferred from the claim text.