Analyzing edge closures of metal surface particles based on graph structures

Patent metadata

Field	Value
Publication number	US-12014483-B2
Application number	US-202217671321-A
Country	US
Kind code	B2
Filing date	Feb 14, 2022
Priority date	Feb 14, 2022
Publication date	Jun 18, 2024
Grant date	Jun 18, 2024

Abstract

Analysis of edge closures of metal surface particles based on a graph structure.

First claim

What is claimed is:

1. A computer-implemented method, comprising:
using an image segmentation technique to detect points and edges in an image;
classifying at least some of the edges into candidate edges and strong edges based at least in part on a confidence level assigned to the respective edge;
for the points detected in the image, using a convolution operation to extract visual information around the respective point;
generating a graph structure including nodes based on the detected points and edges, wherein generating the graph structure includes using the visual information to assign initial information values to the nodes;
updating nodes in the graph structure with information about neighboring nodes; and
based at least in part on relationships between the nodes, determining whether each of the candidate edges is a closed edge.

2. The computer-implemented method of claim 1 , wherein the points are classified into intersection points, end points and free points based at least in part on the detected edges.

3. The computer-implemented method of claim 1 , wherein the strong edges have a confidence level greater than or equal to about 0.7.

4. The computer-implemented method of claim 1 , wherein the candidate edges have a confidence level between about 0.2 and about 0.7.

5. The computer-implemented method of claim 1 , wherein the image is a metallurgical image.

6. The computer-implemented method of claim 5 , comprising detecting imperfections in a product shown in the metallurgical image.

7. A computer program product, the computer program product comprising:
one or more computer readable storage media, and program instructions collectively stored on the one or more computer readable storage media, the program instructions comprising:
program instructions to use an image segmentation technique to detect points and edges in an image;
program instructions to classify at least some of the edges into candidate edges and strong edges based at least in part on a confidence level assigned to the respective edge;
program instructions to, for the points detected in the image, use a convolution operation to extract visual information around the respective point;
program instructions to generate a graph structure including nodes based on the detected points and edges, wherein generating the graph structure includes using the visual information to assign initial information values to the nodes;
program instructions to update nodes in the graph structure with information about neighboring nodes; and
program instructions to determine whether each of the candidate edges is a closed edge, based at least in part on relationships between the nodes.

8. The computer program product of claim 7 , wherein the points are classified into intersection points, end points and free points based at least in part on the detected edges.

9. The computer program product of claim 7 , wherein the strong edges have a confidence level greater than or equal to about 0.7.

10. The computer program product of claim 7 , wherein the candidate edges have a confidence level between about 0.2 and about 0.7.

11. The computer program product of claim 7 , wherein the image is a metallurgical image.

12. The computer program product of claim 11 , comprising program instructions to detect imperfections in a product shown in the metallurgical image.

13. A system, comprising:
a processor; and
logic integrated with the processor, executable by the processor, or integrated with and executable by the processor, the logic being configured to:
use an image segmentation technique to detect points and edges in an image;
classify at least some of the edges into candidate edges and strong edges based at least in part on a confidence level assigned to the respective edge;
for the points detected in the image, use a convolution operation to extract visual information around the respective point;
to generate a graph structure including nodes based on the detected points and edges, wherein generating the graph structure includes using the visual information to assign initial information values to the nodes;
update nodes in the graph structure with information about neighboring nodes; and
determine whether each of the candidate edges is a closed edge, based at least in part on relationships between the nodes.

14. The system of claim 13 , wherein the points are classified into intersection points, end points and free points based at least in part on the detected edges.

15. The system of claim 13 , wherein the strong edges have a confidence level greater than or equal to about 0.7.

16. The system of claim 13 , wherein the candidate edges have a confidence level between about 0.2 and about 0.7.

17. The system of claim 13 , wherein the image is a metallurgical image.

18. The system of claim 17 , comprising logic configured to detect imperfections in a product shown in the metallurgical image.

19. A computer-implemented method, comprising:
detecting points in an image of surface particles of a material;
detecting edges in the image;
classifying at least some of the edges into candidate edges and strong edges based at least in part on a confidence level assigned to the respective edge;
for the points detected in the image, using a convolution operation to extract visual information around the respective point;
generating a graph structure including nodes based on the detected points and edges, wherein generating the graph structure includes determining whether a strong edge exists between adjacent pairs of the points;
updating nodes in the graph structure with information about neighboring nodes; and
selecting candidate edges as closed edges based on relationships of the updated nodes in the graph structure.

20. The computer-implemented method of claim 19 , wherein each point is taken as the center in the convolution operation.

21. The computer-implemented method of claim 19 , wherein the edges and points are detected in the image using image segmentation technology.

22. The computer-implemented method of claim 19 , wherein the nodes in the graph structure are updated with embedding, wherein the embedding for each node is obtained by the convolution operation, wherein the embedding provides at least some of the information about the neighboring nodes.

23. A computer program product, the computer program product comprising:
one or more computer readable storage media, and program instructions collectively stored on the one or more computer readable storage media, the program instructions comprising:
program instructions to detect points in an image of surface particles of a material;
program instructions to detect edges in the image;
program instructions to classify at least some of the edges into candidate edges and strong edges based at least in part on a confidence level assigned to the respective edge;
program instructions to, for the points detected in the image, use a convolution operation to extract visual information around the respective point;
program instructions to generate a graph structure including nodes based on the detected points and edges, wherein generating the graph structure includes determining whether a strong edge exists between adjacent pairs of the points;
program instructions to update nodes in the graph structure with information about neighboring nodes; and
program instructions to select candidate edges as closed edges based on relationships of the updat

Assignees

• IBM

Inventors

• Yuan Zhong Fang
• Zhang Hong Bing
• Liu Tong
• Zhang Dan
• Zhong yi chen
• Min Xu

Classifications

• G06T7/13

  Edge detection · CPC title

• G06T2207/20164

  Salient point detection; Corner detection · CPC title

• G06T2207/30136

  Metal · CPC title

• G06T7/62

  of area, perimeter, diameter or volume · CPC title

• G06T7/0004 Primary

  Industrial image inspection · CPC title