Method of pattern location using color image data

• US 20090141973A1
• Filed: 12/01/2005
• Published: 06/04/2009
• Est. Priority Date: 12/01/2005
• Status: Active Grant

• Current Assignee: Cognex Technology & Investment Corporation (Cognex Corporation)
• Original Assignee: Cognex Technology & Investment Corporation (Cognex Corporation)
• Inventors: Michael, David, Wallack, Aaron S.
• Granted Patent: US 7,965,887 B2
• Time in Patent Office: Days
• US Class Current: 382/165
• CPC Class Codes: G06V 10/28 Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
• CPC Class Codes: G06V 10/7515 Shifting the patterns to accommodate for positional errors

Abstract

Machine vision tools are applied to color images using methods that that utilize an optimized spectrum of the color information. Such methods include full color normalized correlation techniques and methods to convert color images to greyscale using weighting factors that maximize color contrast in a corresponding greyscale image.

Claims

1. A machine vision method for locating a pattern of an object in a color image, the method comprising:

providing a color model image of an expected pattern;

providing a color run-time image of the pattern; and

using color information from the color model image to determine a pose having at least one non-translational degree of freedom of the expected pattern in the color run-time image of the pattern.

2. The method as recited in claim 1 wherein the step of using color information to determine the pose comprises a color rotation-scale invariant search.

3. The method as recited in claim 2 wherein the color model image and the color run-time image each comprise a plurality of color channels, and the step of determining a pose comprises computing a vector representation of each of the color model image and the color nin-time image.

4. The method as recited in claim 2 further comprising;

computing a greyscale runtime image of the color runtime image; and

analyzing the greyscale runtime image using the pose.

5. The method as recited in claim 1 further comprising:

computing a greyscale run-time image of the color run-time image; and

analyzing the greyscale run-time image using the pose.

6. The method as recited in claim 1 further comprising:

computing a dynamic weight for each of a plurality of color channels in at least one of the color model image and the color run-time image;

computing a greyscale image of the color model image and the color run-time image using the dynamic weights; and

analyzing the greyscale image of the color run-time image using the pose.

7. The method as recited in claim 1 further comprising:

computing a dynamic weight for each of a plurality of color channels in at least one of the color model image and the color mm-time image;

computing a greyscale image of the color model image and the color run-time image using the dynamic weights; and

using the pose as a coarse alignment of the pattern to determine a fine alignment of the pattern using the greyscale image of the color model image and the greyscale image of the color run-time image.

8. The method as recited in claim 1 wherein the color model image and the color run-time image each comprise a plurality of color channels, and the step of determining a pose comprises computing a vector representation of each of the color model image and the color run-time image.

9. A machine vision method for locating a pattern of an object in a color image, the method comprising:

providing a color model image of an expected pattern;

providing a color run-time image of the pattern;

computing a dynamic weight for each of a plurality of color channels in the color model image;

computing a greyscale image of the color model image using the dynamic weight, and a greyscale image of the color run-time image using the dynamic weight; and

using the greyscale image of the color model image to determine a pose having at least one non-translational degree of freedom of the expected pattern in the greyscale run-time image of the pattern.

10. The method as recited in claim 9 wherein the step of computing a dynamic weight for each of the plurality of color channels in the color model image comprises searching for a weighting factor for each color channel and evaluating each respective weighting factor by computing a variance in contrast.

11. A machine vision method for locating a pattern of an object in a color image, the method comprising:

providing a color model image of an expected pattern;

providing a color run-time image of the pattern;

computing a greyscale image of the color model image and a greyscale image of the color run-time image;

using the greyscale image of the color model image to determine a pose having at least one non-translational degree of freedom of the expected pattern in the greyscale run-time image of the pattern; and

analyzing the color run-time image using the pose.

12. The method as recited in claim 11 wherein the step of analyzing the color run-time image using the pose further comprises:

performing an affine transformation of the color model image using the pose; and

using the color model image to locate the pattern of the object in the color run-time image.

13. The method as recited in claim 11 further comprising:

computing a dynamic weight for each of a plurality of color channels in the run-time image; and

wherein

the step of computing a greyscale image of the color model image and a greyscale image of the color run-time image uses the dynamic weight.