IP Library › Granted Patent US 11,429,806
Granted Patent B2
US 11,429,806 · App. 16/678,880 · Granted Aug 30, 2022

Devices, systems, and methods for anomaly detection

Inventors: Xiwu Cao (Arcadia, CA); Bradley Scott Denney (Irvine, CA)
Assignee: Canon Virginia, Inc.
G06K9/623G06K9/6256G06V10/50
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 11,429,806
App. No.
16/678,880
Granted
Aug 30, 2022
Kind
B2
Abstract

Devices, systems, and methods obtain one or more training images; obtain a test image; select one or more associated pixels in the training images for a target pixel in the training images; calculate respective value relationships between a value of the target pixel and respective values of the associated pixels in the training images; select one or more associated pixels in the test image for a target pixel in the test image; and detect an anomaly in the target pixel in the test image based on the respective value relationships between the value of the target pixel and the respective values of the associated pixels in the training images and on respective value relationships between a value of the target pixel and respective values of the associated pixels in the test image.

Claims (87)

1. A device comprising:

one or more computer-readable storage media; and

one or more processors in communication with the one or more computer-readable storage media to cause the device to perform operations comprising:

obtaining training images;

obtaining a test image;

selecting associated pixels in the training images for a first target pixel in the training images;

calculating one or more value relationships between a value of the first target pixel and respective values of the associated pixels in the training images, wherein the one or more value relationships between the value of the first target pixel and the respective values of the associated pixels in the training images include:

a centrality of respective value differences between the value of the first target pixel and the respective values of the associated pixels in the training images, and

a variation of the respective value differences between the value of the first target pixel and the respective values of the associated pixels in the training images;

selecting one or more associated pixels in the test image for a first target pixel in the test image; and

detecting an anomaly in the first target pixel in the test image based on the one or more value relationships between the value of the first target pixel and the respective values of the associated pixels in the training images and on one or more value relationships between a value of the first target pixel and respective values of the one or more associated pixels in the test image.

2. The device of claim 1 , wherein the associated pixels in the training images are separated from each other by pixels that are not the associated pixels.

3. The device of claim 1 , wherein selecting the associated pixels in the training images for the first target pixel in the training images includes:

selecting one or more associated pixels in the training images that are located a first distance range away from the first target pixel, and

selecting one or more associated pixels in the training images that are located a second distance range away from the first target pixel, wherein the first distance range is different from the second distance range.

4. The device of claim 1 , wherein the training images and the test image both depict instances of an object,

wherein a position on the object that is depicted by the first target pixel in the training images is similar to a position on the object that is depicted by the first target pixel in the test image.

5. The device of claim 1 , wherein the operations further comprise:

selecting associated pixels in the training images for a second target pixel in the training images;

calculating one or more value relationships between a value of the second target pixel and respective values of the associated pixels in the training images;

selecting one or more associated pixels in the test image for a second target pixel in the test image; and

detecting an anomaly in the second target pixel in the test image based on the one or more value relationships between the value of the second target pixel and the respective values of the associated pixels in the training images and on one or more value relationships between a value of the second target pixel and respective values of the one or more associated pixels in the test image.

6. A device comprising:

one or more computer-readable storage media; and

one or more processors in communication with the one or more computer-readable storage media to cause the device to perform operations comprising:

obtaining one or more training images;

obtaining a test image;

selecting one or more associated pixels in the one or more training images for a first target pixel in the one or more training images;

selecting one or more associated pixels in the test image for a first target pixel in the test image;

generating a predicted target-pixel value for the first target pixel in the test image based on respective values of the one or more associated pixels in the test image, on a value of the first target pixel in the one or more training images, and on respective values of the one or more associated pixels in the one or more training images;

generating a target-pixel-value-to-predicted-target-pixel-value difference based on the predicted target-pixel value and on a value of the first target pixel in the test image; and

detecting an anomaly in the first target pixel in the test image based at least in part on the target-pixel-value-to-predicted-target-pixel-value difference.

7. A method comprising:

obtaining training images;

obtaining a test image;

selecting associated pixels in the training images for a first target pixel in the training images;

calculating one or more value relationships between a value of the first target pixel and respective values of the associated pixels in the training images, wherein the one or more value relationships between the value of the first target pixel and the respective values of the associated pixels in the training images include:

a centrality of respective value differences between the value of the first target pixel and the respective values of the associated pixels in the training images, and

a variation of the respective value differences between the value of the first target pixel and the respective values of the associated pixels in the training images;

selecting one or more associated pixels in the test image for a first target pixel in the test image; and

detecting an anomaly in the first target pixel in the test image based on the one or more value relationships between the value of the first target pixel and the respective values of the associated pixels in the training images and on one or more value relationships between a value of the first target pixel and respective values of the associated pixels in the test image.

8. The method of claim 7 , wherein the associated pixels in the training images are separated from each other by pixels that are not the associated pixels.

9. A method comprising:

obtaining one or more training images;

obtaining a test image;

selecting one or more associated pixels in the one or more training images for a first target pixel in the one or more training images;

selecting one or more associated pixels in the test image for a first target pixel in the test image;

generating a predicted target-pixel value for the first target pixel in the test image based on respective values of the one or more associated pixels in the test image, on a value of the first target pixel in the one or more training images, and on respective values of the one or more associated pixels in the one or more training images;

generating a target-pixel-value-to-predicted-target-pixel-value difference based on the predicted target-pixel value and on a value of the first target pixel in the test image; and

detecting an anomaly in the first target pixel in the test image based at least in part on the target-pixel-value-to-predicted-target-pixel-value difference.

10. The method of claim 9 , wherein selecting the one or more associated pixels in the one or more training images for the first target pixel in the one or more training images includes:

selecting one or more associated pixels in the one or more training images that are located a first distance range away from the first target pixel, and

selecting one or more associated pixels in the one or more training images that are located a second distance range away from the first target pixel, wherein the first distance range is different from the second distance range.

11. The method of claim 10 , wherein the one or more training images and the test image both depict instances of an object,

wherein a position on the object that is depicted by the first target pixel in the one or more training images is similar to a position on the object that is depicted by the first target pixel in the test image.

12. The method of claim 9 , further comprising:

selecting one or more associated pixels in the training images for a second target pixel in the one or more training images;

selecting one or more associated pixels in the test image for a second target pixel in the test image;

generating a second predicted target-pixel value for the second target pixel in the test image based

on respective values of the one or more associated pixels, of the second target pixel, in the test image,

on a value of the second target pixel in the one or more training images, and

on respective values of the one or more associated pixels, of the second target pixel, in the one or more training images;

generating a second target-pixel-value-to-predicted-target-pixel-value difference based on the second predicted target-pixel value and on a value of the second target pixel in the test image; and

detecting a second anomaly in the second target pixel in the test image based at least in part on the second target-pixel-value-to-predicted-target-pixel-value difference.

13. One or more non-transitory computer-readable media storing instructions that, when executed by one or more computing devices, cause the one or more computing devices to perform operations comprising:

obtaining training images;

obtaining a test image;

selecting associated pixels in the training images for a first target pixel in the training images;

calculating one or more value relationships between a value of the first target pixel and respective values of the associated pixels in the training images, wherein the one or more value relationships between the value of the first target pixel and the respective values of the associated pixels in the training images include:

a centrality of respective value differences between the value of the first target pixel and the respective values of the associated pixels in the training images, and

a variation of the respective value differences between the value of the first target pixel and the respective values of the associated pixels in the training images;

selecting one or more associated pixels in the test image for a first target pixel in the test image; and

detecting an anomaly in the first target pixel in the test image based on the one or more value relationships between the value of the first target pixel and the respective values of the associated pixels in the training images and on one or more value relationships between a value of the first target pixel and respective values of the associated pixels in the test image.

14. The one or more non-transitory computer-readable media of claim 13 , wherein the associated pixels in the training images are separated from each other by pixels that are not the associated pixels.

15. The one or more non-transitory computer-readable media of claim 13 , wherein the training images and the test image both depict instances of an object,

wherein a position on the object that is depicted by the first target pixel in the training images is similar to a position on the object that is depicted by the first target pixel in the test image.

16. One or more non-transitory computer-readable media storing instructions that, when executed by one or more computing devices, cause the one or more computing devices to perform operations comprising:

obtaining one or more training images;

obtaining a test image;

selecting one or more associated pixels in the one or more training images for a first target pixel in the one or more training images;

selecting one or more associated pixels in the test image for a first target pixel in the test image;

generating a predicted target-pixel value for the first target pixel in the test image based on respective values of the one or more associated pixels in the test image, on a value of the first target pixel in the one or more training images, and on respective values of the one or more associated pixels in the one or more training images;

generating a target-pixel-value-to-predicted-target-pixel-value difference based on the predicted target-pixel value and on a value of the first target pixel in the test image; and

detecting an anomaly in the first target pixel in the test image based at least in part on the target-pixel-value-to-predicted-target-pixel-value difference.

17. The one or more non-transitory computer-readable media of claim 16 , wherein selecting the one or more associated pixels in the one or more training images for the first target pixel in the one or more training images includes:

selecting one or more associated pixels in the one or more training images that are located a first distance range away from the first target pixel, and

selecting one or more associated pixels in the one or more training images that are located a second distance range away from the first target pixel, wherein the first distance range is different from the second distance range.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Nov 8, 2019
From: CAO, XIWU; DENNEY, BRADLEY SCOTT
To: CANON VIRGINIA, INC.
Reel/Frame 050961/0833 →
Continuity (2)
Provisional Application 62758281 · Nov 9, 2018
Related Publication 20200151496A1 · May 14, 2020