False alarm reduction system for automatic manufacturing quality control

Patent metadata

Field	Value
Publication number	US-11087452-B2
Application number	US-201916248897-A
Country	US
Kind code	B2
Filing date	Jan 16, 2019
Priority date	Feb 5, 2018
Publication date	Aug 10, 2021
Grant date	Aug 10, 2021

Abstract

A false alarm reduction system and method are provided for reducing false alarms in an automatic defect detection system. The false alarm reduction system includes a defect detection system, generating a list of image boxes marking detected potential defects in an input image. The false alarm reduction system further includes a feature extractor, transforming each of the image boxes in the list into a respective set of numerical features. The false alarm reduction system also includes a classifier, computing as a classification outcome for the each of the image boxes whether the detected potential defect is a true defect or a false alarm responsive to the respective set of numerical features for each of the image boxes.

First claim

What is claimed is:

1. A false alarm reduction system for reducing false alarms in an automatic defect detection system, the false alarm reduction system comprising:
a defect detection system, generating a list of image boxes marking detected potential defects in an input image;
a feature extractor, transforming each of the image boxes in the list into a respective set of numerical features; and
a classifier, computing as a classification outcome for the each of the image boxes whether the detected potential defect is a true defect or a false alarm responsive to respective values of the respective set of numerical features for each of the image boxes.

2. The false alarm reduction system of claim 1 , wherein the defect detection system generates the list of boxes marking the potential defects using at least one object selected from the group consisting of an contextual autoencoder, a deep autoencoder, a one-class Support Vector Machine, a nearest neighbor classifier, a binary classifier, and a multi-class classifier.

3. The false alarm reduction system of claim 2 , wherein the numeral features comprise features selected from the group consisting of an autoencoder output, features from intermediate layers of one or more pre-trained deep neural networks, scale invariant feature transform features, and histogram of Gabor functions features.

4. The false alarm reduction system of claim 1 , wherein the classifier computes the classification outcome using a one-class classifier selected from the group consisting of a one-class Support Vector Machine classifier and a Nearest-Neighbor classifier.

5. The false alarm reduction system of claim 4 , wherein the classifier is trained by using as training examples defects generated by the defect detection system over a set of images known to be defect free.

6. The false alarm reduction system of claim 1 , wherein the classifier computes the classification outcome using a classifier selected from the group consisting of a binary classifier and a multi-class classifier.

7. The false alarm reduction system of claim 1 , wherein the classifier comprises an object selected from the group consisting of a Support Vector Machine and a deep neural network.

8. The false alarm reduction system of claim 7 , wherein the classifier is trained by using as training examples all defects generated by the defect detection system over a set of images known to be free of defects, combined with examples known to include true defects.

9. The false alarm reduction system of claim 8 , wherein the true defects are obtained by operating the false alarm reduction system without false alarm reduction for a period of time.

10. The false alarm reduction system of claim 1 , wherein the defect detection system further includes a correction element for performing a corrective action responsive to the classification outcome being the true defect.

11. A false alarm reduction method for reducing false alarms in an automatic defect detection system, the false alarm reduction method comprising:
generating, by a defect detection system, a list of image boxes marking detected potential defects in an input image;
transforming, by a feature extractor, each of the image boxes in the list into a respective set of numerical features; and
computing, by a classifier, as a classification outcome for the each of the image boxes whether the detected potential defect is a true defect or a false alarm responsive to respective values of the respective set of numerical features for each of the image boxes.

12. The false alarm reduction method of claim 11 , wherein the list of boxes marking the potential defects are generated using at least one object selected from the group consisting of an contextual autoencoder, a deep autoencoder, a one-class Support Vector Machine, a nearest neighbor classifier, a binary classifier, and a multi-class classifier.

13. The false alarm reduction method of claim 12 , wherein the numeral features comprise features selected from the group consisting of an autoencoder output, features from intermediate layers of one or more pre-trained deep neural networks, scale invariant feature transform features, and histogram of Gabor functions features.

14. The false alarm reduction method of claim 11 , wherein the classifier computes the classification outcome using a one-class classifier selected from the group consisting of a one-class Support Vector Machine classifier and a Nearest-Neighbor classifier.

15. The false alarm reduction method of claim 14 , further comprising training the classifier by using as training examples defects generated by the defect detection system over a set of images known to be defect free.

16. The false alarm reduction method of claim 11 , wherein the classifier computes the classification outcome using a classifier selected from the group consisting of a binary classifier and a multi-class classifier.

17. The false alarm reduction method of claim 11 , wherein the classifier comprises an object selected from the group consisting of a Support Vector Machine and a deep neural network.

18. The false alarm reduction method of claim 17 , further comprising training the classifier by using as training examples all defects generated by the defect detection system over a set of images known to be free of defects, combined with examples known to include true defects.

19. The false alarm reduction method of claim 18 , wherein the true defects are obtained by operating the false alarm reduction system without false alarm reduction for a period of time.

20. The false alarm reduction method of claim 11 , further comprising performing a corrective action responsive to the classification outcome being the true defect.

Assignees

• Nec Lab America Inc
• Nec Corp

Inventors

• Niculescu-Mizil Alexandru
• Min Renqiang
• Cosatto Eric
• Lai Farley
• Graf Hans Peter
• Fontaine Xavier

Classifications

• G06V10/82

  using neural networks · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title

• G06T7/001 Primary

  using an image reference approach · CPC title

• G06T7/0004 Primary

  Industrial image inspection · CPC title

• G06F18/2433

  Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection · CPC title