Fast high-accuracy multi-dimensional pattern inspection

• US 7,251,366 B1
• Filed: 11/10/2003
• Issued: 07/31/2007
• Est. Priority Date: 11/26/1997
• Status: Expired due to Fees

• Current Assignee: Cognex Corporation
• Original Assignee: Cognex Corporation
• Inventors: Silver, William, Wagman, Adam, Wallack, Araon
• Primary Examiner(s): MARIAM, DANIEL G
• Application Number:
  US10/705,294
• Time in Patent Office: 1,359 Days
• Field of Search: 382/103, 382/149, 382/151, 382/159, 382/199, 382/201, 382/209, 382/218, 382/224, 382/291, 382/21
• US Class Current: 382/209
• CPC Class Codes: G06T 7/75 involving models
• CPC Class Codes: G06V 10/24 Aligning, centring, orientation detection or correction of the image
• CPC Class Codes: G06V 10/7515 Shifting the patterns to accommodate for positional errors

Abstract

A method and apparatus are provided for identifying differences between a stored pattern and a matching image subset, where variations in pattern position, orientation, and size do not give rise to false differences. The invention is also a system for analyzing an object image with respect to a model pattern so as to detect flaws in the object image. The system includes extracting pattern features from the model pattern; generating a vector-valued function using the pattern features to provide a pattern field; extracting image features from the object image; evaluating each image feature, using the pattern field and an n-dimensional transformation that associates image features with pattern features, so as to determine at least one associated feature characteristic; and using at least one feature characteristic to identify at least one flaw in the object image. The invention can find at least two distinct kinds of flaws: missing features, and extra features. The invention provides pattern inspection that is faster and more accurate than any known prior art method by using a stored pattern that represents an ideal example of the object to be found and inspected, and that can be translated, rotated, and scaled to arbitrary precision much faster than digital image re-sampling, and without pixel grid quantization errors. Furthermore, since the invention does not use digital image re-sampling, there are no pixel quantization errors to cause false differences between the pattern and image that can limit inspection performance.

Claims

1. A method of training a model pattern for use in geometric pattern matching, the method comprising:

obtaining a digital image of an object;

detecting boundary points in the digital image; and

generating information about the boundary points for inclusion in said model pattern, the information being stored as a vector-valued function of position within a region of the image that includes the boundary points, resulting in a model pattern that includes information that is a vector-valued function of position within the region of the image that includes the boundary points.

2. The method of claim 1, wherein the information is stored as a function of real-valued position within the region of the image that includes the boundary points.

3. The method of claim 1, wherein the information is stored at discrete points on a grid using a two-dimensional array.

4. The method of claim 1, wherein the vector-valued function relates a plurality of at least two-dimensional positions to a plurality of respective displacement vectors.

5. The method of claim 4, wherein each displacement vector indicates a distance and direction from a two-dimensional position to a nearest point along a pattern boundary formed by the boundary points.

6. The method of claim 1, wherein the vector-valued function relates a plurality of two-dimensional positions and associated directions to a plurality of respective displacement vectors.

7. The method of claim 6, wherein each displacement vector indicates distance and direction from a two-dimensional position to a nearest point along a pattern boundary that is substantially in the associated direction.

8. The method of claim 1, wherein generating the information about the boundary points includes a sequence of propagation events, each propagation event providing additional information at respective additional positions within the region of the image that includes the boundary points.

9. The method of claim 8, wherein positions within the region of the image that includes the boundary points are initialized to a pre-defined value before the sequence of propagation events.

10. The method of claim 9, wherein the pre-defined value is one of a value independent of any boundary point, and a value dependent on at least one boundary point.

11. The method of claim 8, wherein the sequence of propagating events occurs using a plurality of phases.

12. The method of claim 1, wherein generating information about the boundary points includes:

determining neighboring boundary points for each boundary point.

13. The method of claim 12, wherein generating information about the boundary points includes:

connecting neighboring boundary points to provide at least one chain of boundary points.

14. The method of claim 13, wherein generating information about the boundary points includes:

identifying and cataloging each chain of boundary points.

15. The method of claim 14, wherein identifying and cataloging includes:

for each chain, determining a starting boundary point, an ending boundary point, a total gradient magnitude, and number of boundary points in the chain.

16. The method of claim 13, wherein generating information about the boundary points includes:

identifying and ignoring weak chains.

17. The method of claim 16, wherein identifying and ignoring weak chains includes:

using gradient magnitude.

18. The method of claim 13, wherein generating information about the boundary points includes:

dividing chains into segments.

19. The method of claim 18, wherein dividing chains into segments includes:

dividing chains into segments based on curvature.

20. The method of claim 13, wherein generating information about the boundary points includes:

using at least one chain of boundary points to carry out at least one propagation event, each propagation event providing additional information at respective additional positions within the region of the image that includes the boundary points.

21. The method of claim 20, wherein there are four propagation events.

22. The method of claim 21, wherein each propagation event provides a plurality of vectors, each vector indicating a distance and direction toward a pattern boundary that includes a plurality of pattern boundary points.

23. The method of claim 22, wherein each propagation event provides a plurality of vectors that are disposed at a greater distance from the pattern boundary that includes a plurality of pattern boundary points than the plurality of vectors provided by a previous propagation event.

24. The method of claim 20, wherein the number of propagation events is determined by a parameter.

25. A method of training a model pattern for use in geometric pattern matching, the method comprising:

obtaining a digital image of an object;

detecting boundary points in the digital image;

connecting the boundary points to provide at least one chain of boundary points; and

using at least one chain of boundary points to carry out at least one propagation event, each propagation event providing a plurality of vectors, each vector indicating a distance and direction towards the at least one chain of pattern boundary points at respective additional positions within the region of the image that includes the boundary points.

26. The method of claim 25, wherein connecting the boundary points includes:

identifying neighboring boundary points for each boundary point.

27. The method of claim 25, before using at least one chain of boundary points, further including:

for each chain, determining a starting boundary point, an ending boundary point, a total gradient magnitude, and a number of boundary points in the chain.

28. The method of claim 25, before using at least one chain of boundary points, further including:

identifying and ignoring weak chains.

29. The method of claim 28, wherein identifying and ignoring weak chains includes:

identifying and ignoring weak chains using gradient magnitude.

30. The method of claim 25, before using at least one chain of boundary points, further including:

dividing chains into segments.

31. The method of claim 30, wherein dividing chains into segments includes:

dividing chains into segments based on curvature.

32. The method of claim 25, wherein using at least one chain of boundary points to carry out at least one propagation event includes:

performing four propagation events.

33. The method of claim 25, wherein each propagation event provides a plurality of vectors that are disposed at a greater distance from the pattern boundary that includes a plurality of pattern boundary points than the plurality of vectors provided by a previous propagation event.

34. The method of claim 25, wherein the number of propagation events is determined by a user-settable parameter.