Learning-type classifying apparatus and learning-type classifying method

• US 20070009152A1
• Filed: 09/13/2006
• Published: 01/11/2007
• Est. Priority Date: 03/31/2004
• Status: Active Grant

• Current Assignee: Olympus Corporation
• Original Assignee: Olympus Corporation
• Inventors: Kanda, Yamato
• Granted Patent: US 7,773,801 B2
• Time in Patent Office: Days
• US Class Current: 382/159
• CPC Class Codes: G01N 2021/8854 Grading and classifying of flaws
• CPC Class Codes: G01N 21/9501 Semiconductor wafers manufacturing processes per se of semiconductor devices implementing a measuring step H01L22/10
• CPC Class Codes: G01N 21/956 Inspecting patterns on the surface of objects contactless testing of electronic circuits G01R31/308; testing currency G07D; manufacturing processes per se of semiconductor devices implementing a measuring step H01L22/10
• CPC Class Codes: G06F 16/5838 using colour
• CPC Class Codes: G06F 18/40 Software arrangements specially adapted for pattern recognition, e.g. user interfaces or toolboxes therefor
• CPC Class Codes: G06T 2207/30148 Semiconductor; IC; Wafer
• CPC Class Codes: G06T 7/0004 Industrial image inspection
• CPC Class Codes: G06V 10/945 User interactive design; Environments; Toolboxes

Abstract

A learning-type classifying apparatus comprises defective region extracting unit for extracting defective regions of classification targets from an image in which the plurality of regions of the classification targets are present, characteristic value calculating unit for calculating characteristic values for the extracted regions of the classification targets, classifying unit for classifying the extracted regions of the classification targets into predetermined classes on the basis of the calculated characteristic values, region integrating unit for integrating the regions which belong to the same class as a result of the classification, display unit for displaying images of the integrated regions and the classification results, input unit for correcting errors in the classification results, and teacher data creating unit for creating teacher data for each of the regions so that the classification results of the integrated regions are reflected in each region included in the integrated regions.

Claims

1. A learning-type classifying apparatus comprising:

region extracting unit for extracting regions of classification targets from an image in which the plurality of regions of the classification targets are present;

characteristic value calculating unit for calculating characteristic values for the extracted regions of the classification targets;

classifying unit for classifying the extracted regions of the classification targets into predetermined classes on the basis of the calculated characteristic values;

region integrating unit for integrating the regions which belong to the same class as a result of the classification;

display unit for displaying images of the integrated regions and the classification results;

input unit for correcting errors in the classification results; and

teacher data creating unit for creating teacher data for each of the regions so that the classification results of the integrated regions are reflected in each region included in the integrated regions.

2. The learning-type classifying apparatus according to claim 1, wherein when erroneously integrated regions are selected by the input unit, the region integrating unit divides a plurality of regions in the integrated regions into at least two or more clusters within a characteristic space, and reintegrates the regions belonging to each cluster one by one.

3. The learning-type classifying apparatus according to claim 2, wherein the clusters are created by use of a distance on a characteristic axis specified by the input unit.

4. The learning-type classifying apparatus according to claim 1, further comprising judging unit for judging the necessity of creating additional teacher data from changes in the percentage of classification correctness in each class along with an increase in the number of teacher data.

5. The learning-type classifying apparatus according to claim 2, further comprising judging unit for judging the necessity of creating additional teacher data from changes in the percentage of classification correctness in each class along with an increase in the number of teacher data.

6. The learning-type classifying apparatus according to claim 3, further comprising judging unit for judging the necessity of creating additional teacher data from changes in the percentage of classification correctness in each class along with an increase in the number of teacher data.

7. A learning-type classifying method comprising:

extracting regions of classification targets from an image in which the plurality of regions of the classification targets are present;

calculating characteristic values for the extracted region;

classifying the regions into predetermined classes on the basis of the calculated characteristic values;

integrating the regions which belong to the same class as a result of the classification;

displaying images of the integrated regions and the classification results;

acquiring instruction information to correct errors in the classification results; and

creating teacher data for each of the regions so that the classification results of the integrated regions are reflected in each region included in the integrated regions.

8. The learning-type classifying method according to claim 7, wherein when erroneously integrated regions are selected by the acquired instruction information, a plurality of regions in the integrated regions are divided into at least two or more clusters within a characteristic space, and the regions belonging to each cluster are reintegrated one by one.

9. The learning-type classifying method according to claim 7, wherein the clusters are created by use of a distance on a characteristic axis specified by the input unit.

10. The learning-type classifying method according to claim 7, wherein the necessity of creating additional teacher data is judged from changes in the percentage of classification correctness in each class along with an increase in the number of teacher data.