Fast high-accuracy multi-dimensional pattern inspection

• US 7,088,862 B1
• Filed: 11/10/2003
• Issued: 08/08/2006
• Est. Priority Date: 11/26/1997
• Status: Expired due to Fees

• Current Assignee: Cognex Corporation
• Original Assignee: Cognex Corporation
• Inventors: Wallack, Aaron, Silver, William, Wagman, Adam
• Primary Examiner(s): MARIAM, DANIEL G
• Application Number:
  US10/705,297
• Time in Patent Office: 1,002 Days
• Field of Search: 382/103, 382/149, 382/199, 382/261, 382/264, 382/266, 382/300
• US Class Current: 382/199
• 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 tunable boundary detector for detecting features in a source image, the boundary detector comprising:

a tunable low-pass filter for filtering the source image to provide a filtered image;

an adjustable image sub-sampler for sub-sampling the filtered image to provide a sub-sampled image; and

an edge detector for detecting edges in the sub-sampled image to provide edges, each edge having an edge position, a gradient magnitude, and a gradient direction wherein the edge detector includes a peak detector configured to use the gradient magnitude and the gradient direction to identify points where the gradient magnitude is a local maximum.

2. The tunable boundary detector of claim 1, wherein the position is expressed as real-valued coordinates.

3. The tunable boundary detector of claim 1, wherein the image sub-sampler is controlled by a plurality of parameters.

4. The tunable boundary detector of claim 1, wherein the sub-sampling amount of the image sub-sampler is controlled by at least one parameter.

5. The tunable boundary detector of claim 1, wherein the constant-time low-pass filter is controlled by a plurality of parameters.

6. The tunable boundary detector of claim 1, wherein the source image is one of a training image and a run-time image.

7. The tunable boundary detector of claim 1, wherein the low-pass filter is a constant-time low-pass filter using substantially similar computational time over a range of adjustments of the filter.

8. The tunable boundary detector of claim 1, wherein the low-pass filter is a second-order low-pass filter.

9. The tunable boundary detector of claim 1, wherein the low-pass filter is substantially an approximation to a Gaussian low-pass filter.

10. The tunable boundary detector of claim 1, wherein the low-pass filter is substantially an approximation to a parabolic low-pass filter.

11. The tunable boundary detector of claim 1, wherein the edge detector comprises:

a gradient estimator configured to provide an estimate of horizontal and vertical components of image gradient at each pixel position;

a Cartesian-to-polar converter configured to convert each estimate of horizontal and vertical components of image gradient into an estimate of gradient magnitude and gradient direction;

a peak detector configured to use each estimate of gradient magnitude and gradient direction to identify points where the gradient magnitude is a local maximum, and to provide a column number, a row number, a gradient magnitude and a gradient direction for each such point; and

a sub-pixel interpolator configured to use each column number, row number, gradient magnitude and gradient direction to provide a vertical component of real valued edge position, a horizontal component of real valued edge position, a gradient magnitude, and a gradient direction for each such point.

12. The tunable boundary detector of claim 11, wherein the peak detector can be adjusted using a plurality of parameters.

13. The tunable boundary detector of claim 11, wherein a noise threshold of the peak detector is adjusted using a parameter.

14. The tunable boundary detector of claim 11, wherein the sub-pixel interpolator can be adjusted using a plurality of parameters.

15. The tunable boundary detector of claim 11, wherein the Cartesian-to-polar converter uses a CORDIC method to compute gradient magnitude and direction.

16. The tunable boundary detector of claim 11, wherein the Cartesian-to-polar converter computes both gradient magnitude and gradient direction to at least six bits.

17. The tunable boundary detector of claim 11, wherein the gradient estimator computes the x and y components of gradient to 16 bits.

18. The tunable boundary detector of claim 11, wherein the gradient estimator uses a Sobel kernel.

19. The tunable boundary detector of claim 11, wherein the gradient estimator receives a 16-bit filtered image.

20. The tunable boundary detector of claim 11, wherein the gradient estimator receives a 8 bit unfiltered image.

21. The tunable boundary detector of claim 11, wherein the peak detector identifies a plurality of points where the gradient magnitude exceeds a noise threshold and is a local maximum along a one-dimensional profile that lies in approximately the gradient direction, and provides grid coordinates, gradient magnitude, and gradient direction for each such point.

22. The tunable boundary detector of claim 21, wherein the sub-pixel interpolator interpolates position of maximum gradient magnitude along the one-dimensional profile to determine real-valued coordinates of each point so as to provide a plurality of points that lie along boundaries in the source image, including the grid coordinates, gradient direction, and gradient magnitude of each point.

23. The tunable boundary detector of claim 1, wherein a parameter of the low-pass filter is set to pass fine detail so as to provide a high-resolution pattern.

24. The tunable boundary detector of claim 1, wherein a parameter of the low-pass filter is set to attenuate fine detail so as to provide a low-resolution pattern.

25. The tunable boundary detector of claim 1, wherein a parameter of the low-pass filter is set to disable the low-pass filter.

26. The tunable boundary detector of claim 1, wherein the source image has eight bits of gray-scale per pixel, and the low-pass filter provides a filtered image having 16 bits of gray-scale per pixel.

27. A tunable boundary detector for detecting features in a source image, the boundary detector comprising:

a tunable low-pass filter for filtering the source image to provide a filtered image;

a gradient estimator configured to provide an estimate of horizontal and vertical components of image gradient at each pixel position;

a Cartesian-to-polar converter configured to convert each estimate of horizontal and vertical components of image gradient into an estimate of gradient magnitude and gradient direction;

a peak detector configured to use each estimate of gradient magnitude and gradient direction to identify points where the gradient magnitude is a local maximum, and to provide for each such point a column number, a row number, a gradient magnitude and a gradient direction; and

a sub-pixel interpolator configured to use each column number, row number, gradient magnitude and gradient direction to provide for each such point a vertical component of real valued edge position, a horizontal component of real valued edge position, a gradient magnitude, and a gradient direction.

28. The tunable boundary detector of claim 27, further comprising:

an adjustable image sub-sampler, cooperative with the low-pass filter, for sub-sampling the filtered image to provide a sub-sampled image.

29. A tunable boundary detection method for detecting features in a source image, the method comprising:

filtering the source image to provide a filtered image;

providing an estimate of horizontal and vertical components of image gradient at each pixel position;

converting each estimate of horizontal and vertical components of image gradient into an estimate of gradient magnitude and gradient direction;

using each estimate of gradient magnitude and gradient direction to identify points where the gradient magnitude is a local maximum, and to provide for each such point a column number, a row number, a gradient magnitude and a gradient direction; and

using each column number, row number, gradient magnitude, and gradient direction to provide for each such point a vertical component of real valued sub-pixel edge position, a horizontal component of real valued sub-pixel edge position, a gradient magnitude, and a gradient direction.

30. The method of claim 29, further comprising:

after filtering, sub-sampling the filtered image to provide a sub-sampled image.