Product defect detection

Patent metadata

Field	Value
Publication number	US-10964015-B2
Application number	US-201916248114-A
Country	US
Kind code	B2
Filing date	Jan 15, 2019
Priority date	Jan 15, 2019
Publication date	Mar 30, 2021
Grant date	Mar 30, 2021

Abstract

Embodiments of the present invention facilitate product defect detection. A computer-implemented method comprises: receiving, by a device operatively coupled to one or more processors, a template image of a normal product; generating, by the device, one or more geometric training parameters for transforming the template image; and transforming, by the device, the template image using the one or more geometric training parameters to generate a transformed image for training a data model, wherein the trained data model being used for aligning the template image and an image under inspection of a product.

First claim

What is claimed is:

1. A computer-implemented method, comprising:
receiving, by a device operatively coupled to one or more processors, a template image of a defined product;
iteratively employing, by the device, a random number generator to randomly generate respective random transformation parameters to be employed by a geometric transformation module for geometric transformation of the template image;
iteratively generating, by the device, respective geometric training parameters for transforming the template image, wherein the iteratively generated respective geometric training parameters are based on the respective random transformation parameters; and
iteratively transforming, by the device, the template image using the respective geometric training parameters to generate respective transformed template images that are randomly generated to train a data model, wherein the trained data model is employed to align the template image and an image of a product under inspection to determine whether the product has a defect.

2. The computer-implemented method of claim 1 , wherein types of the respective geometric training parameters are selected from a group consisting of a rotation angle parameter, a shearing parameter and a scale parameter.

3. The computer-implemented method of claim 1 , further comprising:
receiving, by the device, the image of the product under inspection;
aligning, by the device, the image under inspection with the template image, wherein the aligning is performed based on the trained data model generated from the respective transformed template images; and
comparing, by the device, the aligned image and the template image to identify the defect of the product under inspection.

4. The computer-implemented method of claim 1 , wherein the data model is built based on a neural network selected from a group consisting of a Convolution Neural Network, a Recurrent Neural Network and a Deep Neural Network.

5. The computer-implemented method of claim 4 , wherein the neural network on which the data model is based contains multiple convolution layers and multiple max pooling layers.

6. The computer-implemented method of claim 1 , wherein the data model is trained by machine learning of a correspondence between the transformed image and the template image through the respective geometric training parameters.

7. A system, comprising:
a memory that stores computer executable components;
a processor that executes the computer executable components stored in the memory, wherein the computer executable components comprise:
at least one computer-executable component that:
receives a template image of a defined product;
employs a random number generator to randomly generate one or more random transformation parameters to be employed by a geometric transformation module for geometric transformation of the template image;
generates one or more geometric training parameters for transforming the template image, wherein the one or more geometric training parameters are based on the one or more random transformation parameters; and
transforms the template image using the one or more geometric training parameters to generate a transformed template image that is randomly generated to train a data model, wherein the trained data model is employed to align the template image and an image of a product under inspection to determine whether the product has a defect.

8. The system of claim 7 , wherein types of the one or more geometric parameters are selected a group consisting of a rotation angle parameter, a shearing parameter and a scale parameter.

9. The system of claim 7 , wherein the at least one computer-executable component that:
receives an image of the product under inspection;
aligns the image under inspection with the template image; and
compares the aligned image and the template image to identify the defect of the product.

10. The system of claim 7 , wherein the data model is built based on a neural network, wherein the neural network is selected from a group consisting of a Convolution Neural Network, a Recurrent Neural Network and a Deep Neural Network.

11. The system of claim 10 , wherein the neural network on which the data model is based contains multiple convolution layers and multiple max pooling layers.

12. The system of claim 7 , wherein the data model is trained by machine learning of the correspondence between the transformed template image and the template image through the one or more geometric training parameters.

13. A computer program product facilitating detection of product defects by aligning product images under inspection with template product images, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by one or more processors to cause the one or more processors to:
receive, by the one or more processors, a template image of a defined product;
employ, by the one or more processors, a random number generator to randomly generate one or more random transformation parameters to be employed by a geometric transformation module for geometric transformation of the template image;
generate, by the one or more processors, one or more geometric training parameters for transforming the template image, wherein the one or more geometric training parameters are based on the one or more random transformation parameters; and
transform, by the one or more processors, the template image using the one or more geometric training parameters to generate a transformed template image that is randomly generated to train a data model, wherein the trained data model is employed to align the template image and an image of a product under inspection of a product to determine whether the product has a defect.

14. The computer program product of claim 13 , wherein types of the one or more geometric parameters are selected from group consisting of a rotation angle parameter, a shearing parameter and a scale parameter.

15. The computer program product of claim 13 , wherein the program instructions are further executable by the one or more processors to cause the one or more processors to:
receive an image of the product under inspection;
align the image under inspection with the template image; and
compare the aligned image and the template image to identify the defect of the product.

Assignees

• IBM

Inventors

• Hu Guo Qiang
• Yan jun chi
• Zhu Jun
• Huang Jing Chang
• Ji Peng

Classifications

• G06V10/774

  Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06V10/82

  using neural networks · CPC title

• G06V10/764

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

• G06T2207/20081

  Training; Learning · CPC title

• G06T2207/20076

  Probabilistic image processing · CPC title