System and method for work piece inspection

Patent metadata

Field	Value
Publication number	US-10755401-B2
Application number	US-201816208668-A
Country	US
Kind code	B2
Filing date	Dec 4, 2018
Priority date	Dec 4, 2018
Publication date	Aug 25, 2020
Grant date	Aug 25, 2020

Abstract

An inspection system includes one or more imaging devices and one or more processors. The imaging devices generate a first set of images of a work piece at a first position relative to the work piece and a second set of images of the work piece at a second position relative to the work piece. At least some of the images in the first and second sets are acquired using different light settings. The processors analyze the first set of images to generate a first prediction image associated with the first position, and analyze the second set of images to generate a second prediction image associated with the second position. The first and second prediction images include respective candidate regions. The processors merge the first and second prediction images to detect at least one predicted defect in the work piece depicted in at least one of the candidate regions.

First claim

What is claimed is:

1. An inspection system comprising:
one or more imaging devices configured to generate a first set of images of a work piece at a first position of the one or more imaging devices relative to the work piece and to generate a second set of images of the work piece at a second position relative to the work piece, wherein at least some of the images in the first set are acquired using different light settings and at least some of the images in the second set are acquired using the different light settings; and
one or more processors configured to obtain the first set of images and analyze the first set to generate a first prediction image associated with the first position of the one or more imaging devices, the one or more processors also configured to obtain the second set of images and analyze the second set to generate a second prediction image associated with the second position of the one or more imaging devices, the first and second prediction images including respective candidate regions that depict a potential defect in a surface of the work piece,
wherein the one or more processors are configured to merge the first prediction image and the second prediction image to detect at least one predicted defect in the surface of the work piece, wherein the at least one predicted defect is depicted in at least one of the candidate regions of the first and second prediction images.

2. The inspection system of claim 1 , further comprising multiple light sources operably connected to the one or more processors and configured to generate illumination light towards the surface of the work piece, wherein the one or more processors are configured to control the light sources to provide the different light settings by one or more of changing a wavelength of the illumination light, changing an intensity of the illumination light, activating a first light source of the light sources to generate the illumination light, deactivating the first light source of the light sources to cease generating the illumination light, or changing a position of at least one of the light sources.

3. The inspection system of claim 1 , wherein the one or more imaging devices in the second position have at least one of a different location or a different angle relative to the work piece than the one or more imaging devices in the first position.

4. The inspection system of claim 1 , wherein the one or more processors are configured to analyze the first set of images by examining the images one at a time as inputs in a forward propagation direction through layers of artificial neurons in an artificial neural network to generate a set of defect masks, each defect mask in the set based on a different one of the images in the first set, wherein the one or more processors generate the first prediction image by merging the defect masks in the set.

5. The inspection system of claim 4 , wherein the one or more processors are configured to merge the defect masks to generate the first prediction image via a linear combination of pixels of the defect masks.

6. The inspection system of claim 1 , wherein the one or more processors are configured to analyze the first set of images by examining the images one at a time as inputs in a forward propagation direction through layers of artificial neurons in an artificial neural network, wherein the layers include at least one long term short term memory layer configured to weight features depicted in the images and to fuse information extracted from the images to generate the first prediction image.

7. The inspection system of claim 1 , wherein, responsive to detecting the at least one predicted defect in the surface of the work piece, the one or more processors are configured to generate a control signal configured to segregate the work piece from other work pieces that lack at least one predicted defect.

8. The inspection system of claim 1 , wherein the one or more processors are configured to merge the first prediction image and the second prediction image by mapping the first prediction image and the second prediction image to a computer design model that simulates the work piece, wherein the one or more processors map the first and second prediction images to the computer design model based on the first and second positions of the one or more imaging devices relative to the work piece.

9. The inspection system of claim 1 , wherein the one or more processors are configured to detect the at least one predicted defect in the surface of the work piece after merging the first and second prediction images by determining an amount of overlap between the respective candidate regions of the first prediction image and the respective candidate regions of the second prediction image.

10. The inspection system of claim 1 , wherein the one or more processors are configured to merge the first prediction image and the second prediction image by examining the first and second prediction images as inputs in a forward propagation direction through layers of artificial neurons in an artificial neural network that generates an output probability that at least one of the first and second prediction images depict a defect, wherein the one or more processors detect the at least one predicted defect in the surface of the work piece responsive to the output probability exceeding a designated probability threshold.

11. A method for inspecting a work piece for defects, the method comprising:
obtaining a first set of images of the work piece generated by one or more imaging devices at a first position of the one or more imaging devices relative to the work piece, wherein at least some of the images in the first set are acquired using different light settings;
obtaining a second set of images of the work piece generated by the one or more imaging devices at a second position of the one or more imaging devices relative to the work piece, wherein at least some of the images in the second set are acquired using the different light settings;
analyzing the first set of images via one or more processors to generate a first prediction image associated with the first position of the one or more imaging devices, the first prediction image including a candidate region that depicts a potential defect in a surface of the work piece;
analyzing the second set of images via the one or more processors to generate a second prediction image associated with the second position of the one or more imaging devices, the second prediction image including a candidate region that depicts a potential defect in the surface of the work piece; and
merging the first prediction image and the second prediction image via the one or more processors to detect at least one predicted defect in the surface of the work piece, wherein the at least one predicted defect is depicted in at least one of the candidate regions of the first and second prediction images.

12. The method of claim 11 , further comprising modifying the light settings during the acquisition of the images in the first set and during the acquisition of the images in the second set by one or more of changing a wavelength of illumination light directed towards the surface of the work piece, changing an intensity of illumination light directed towards the surface of the work piece, changing a type of light source that is activated, changing a position of a light source that is activated, or changing a number of light sources that are activated.

13. The method of claim 11 , wherein the first set of images is analyzed by examining the images one at a time as inputs in a forward propagation direction through layers of artificial neurons in an artificial neura

Assignees

• Gen Electric

Inventors

• Bian Xiao
• Jain Arpit
• Diwinsky David Scott
• Bewlay Bernard Patrick
• Bouchard Steeves
• Choiniere Jean-Philippe
• Marois Marc-Andre
• Harel Stephane
• Karigiannis John

Classifications

• H04N23/74

  by influencing the scene brightness using illuminating means · CPC title

• G01N2021/8887

  based on image processing techniques · CPC title

• G01N21/8851 Primary

  Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges (G01N21/8806 and G01N21/93 - G01N21/95692 take precedence; optical measurement of dimensions G01B11/00; optical scanning G02B26/10; image transformation G06T3/00; computerised image enhancement G06T5/00; image processing per se for flaw detection G06T7/0002) · CPC title

• G06T2207/10152

  Varying illumination · CPC title

• G06T2207/20076

  Probabilistic image processing · CPC title