Methods and apparatus for automatic image inspection of continuously moving objects

• US 5,365,596 A
• Filed: 12/17/1992
• Issued: 11/15/1994
• Est. Priority Date: 12/17/1992
• Status: Expired due to Term

• Current Assignee: Philip Morris USA Incorporated (Altria Group, Inc.)
• Original Assignee: Philip Morris USA Incorporated (Altria Group, Inc.)
• Inventors: Lowitz, David A., Dante, Henry M.
• Primary Examiner(s): Mancuso, Joseph
• Assistant Examiner(s): Fox, David
• Application Number:
  US07/992,139
• Time in Patent Office: 698 Days
• Field of Search: 382/8, 382/1, 382/14, 382/30, 382/33, 382/34, 358/106, 358/101, 356/237, 356/430, 250/572
• US Class Current: 382/141
• CPC Class Codes: G01N 21/95607 using a comparative method
• CPC Class Codes: G06T 2207/30164 Workpiece; Machine component
• CPC Class Codes: G06T 7/001 using an image reference approach
• CPC Class Codes: G06V 10/751 Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
• CPC Class Codes: H04N 7/18 Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast

Abstract

The acceptability of images is determined by comparing each image to a template image which may be an average of several images which are known to be acceptable. The amount by which the image deviates from the template image is used as a measure of the acceptability of the image. Portions of the image at or near edges in the images may be automatically excluded from this test. Another test which includes the edge portions excluded from the first test may be performed to ensure that all portions of the image are inspected. For example, this other test may include ensuring that the correct number of edge pixels are present in the image as a whole.

Claims

1. The method of determining whether an image is substantially like a predetermined image which is subject to some minor but acceptable variations, said method comprising the steps of:

(1) forming a template image which is representative of said predetermined image and said minor but acceptable variations by performing the steps of;

(a) identifying a plurality of training images, each of which is substantially like said image, and at least some of which includes at least some of said minor but acceptable variations from said image;

(b) representing each of a plurality of portions of each of said training images by a pixel value, each of said portions of each of said training images respectively corresponding to an associated one of said portions of said template image; and

(c) for each portion of said template image, using as the template image pixel value the largest of the pixel values for the associated portion in said training images;

(2) representing each of a plurality of portions of said template image by a pixel value;

(3) representing each of a plurality of portions of said image of a pixel value, each of said portions of said image respectively corresponding to an associated one of said portions of said template image;

(4) for each of said portions of said image, determining the difference between the pixel value of that portion of said image and the pixel value for the associated template image portion;

(5) cumulating said differences for at least some of said portions of said image to produce a cumulative difference value;

(6) comparing said cumulative difference value to a predetermined threshold value; and

(7) identifying said image as unlike said predetermined image if said cumulative differences value bears a predetermined relationship to said threshold value.

2. The method defined in claim 1 wherein each of said pixel values of each of said training images is a grey scale value.

3. The method defined in claim 1, further comprising the steps of

identifying the portions of said image which are at or near regions of abrupt change in pixel value in the image;

and

excluding said portions of said image which are at or near said regions of abrupt change in pixel value from the portions of said image for which said step of cumulating said differences is performed.

4. The method defined in claim 1, further comprising the steps of

identifying which of said portions of said image are associated with regions of abrupt change in pixel value in said image;

counting the number of said portions of said image which are associated with regions of abrupt change in pixel value; and

identifying said image as unlike said predetermined image if said number is not within a predetermined range.

5. The method defined in claim 3 wherein said step of identifying the portions of said image which are at or near regions of abrupt change in pixel value comprises the steps of:

identifying a plurality of training images, each of which is substantially like said image, and at least some of which include at least some of said minor but acceptable variations from said image;

representing each of a plurality of portions of each of said training images by a pixel value, each of said portions of each of said training images respectively corresponding to an associated one of said portions of said template image;

for each portion of each of said training images, determining the deviation between the pixel value for that portion of that training image and the pixel value for the associated template image portion; and

identifying as portions of said image that are at or near regions of abrupt change in pixel value any portions of any of said training images for which said deviation is greater than a predetermined amount.

6. The method defined in claim 5 wherein said step of identifying as a region of abrupt change in pixel value comprises the steps of:

binarizing said deviations by assigning a first binary value to all portions of each training image for which said deviation is less than a predetermined amount, and by assigning a second binary value to all portions of each training image for which said difference is greater than said predetermined amount; and

identifying as portions of said image that are at or near regions of abrupt change in pixel value any portions of any of said training images which have said second binary value.

7. The method defined in claim 4 wherein said step of identifying which of said portions of said image are associated with edges in said image comprises the step of edge detecting said image.

8. Apparatus for determining whether an image is substantially like a predetermined image which is subject to some minor but acceptable variation, said apparatus comprising:

(1) means for forming a template image which is representative of said predetermined image and said minor but acceptable variation wherein said means for forming a template image comprises;

(a) means for identifying a plurality of training images, each of which is substantially like said image, and at least some of which include at least some of said minor but acceptable variations from said image;

(b) means for representing each of a plurality of portions of each of said training images by a pixel value, each of said portions of each of said training images respectively corresponding to an associated one of said portions of said template image; and

(c) means for, for each portion of said template image, using as the template image pixel value the largest of the pixel values for the associated portion in said training images;

(2) means for representing each of a plurality of portions of said template image by a pixel value;

(3) means for representing each of a plurality of portions of said image by a pixel value, each of said portions of said image respectively corresponding to an associated one of said portions of said template image;

(4) means for, for each of said portions of said image, determining the difference between the pixel value for that portion of said image and the pixel value for the associated template image portion;

(5) means for cumulating said differences for at least some of said portions of said image to produce a cumulative difference value;

(6) means for comparing said cumulative difference value to a predetermined threshold value; and

(7) means for identifying said image as unlike said predetermined image of said cumulative difference value bears a predetermined relationship to said threshold value.

9. The apparatus defined in claim 8 wherein each of said pixel values of each of said training images is a grey scale value.

10. The apparatus defined in claim 8, further comprising

means for identifying which of said portions of said image are associated with regions of abrupt change in pixel value in said image;

means for counting the number of said portions of said image which are associated with regions of abrupt change in pixel value; and

means for identifying said image as unlike said predetermined image if said number is not within a predetermined range.

11. The apparatus defined in claim 8, further comprising

means for identifying the portions of said image which are at or near regions of abrupt change in pixel value in the image;

and

means for excluding said portions of said image which are at or near regions of abrupt change in pixel value in the image from the portions of said image on which said means for cumulating said differences operates.

12. The apparatus defined in claim 10 wherein said means for identifying which of said portions of said image are associated with edges in said image comprises means for edge detecting said image.

13. The apparatus defined in claim 11 wherein said means for identifying the portions of said image which are at or near regions of abrupt change in pixel value comprises:

means for identifying a plurality of training images, each of which is substantially like said image, and at least some of which include at least some of said minor but acceptable variations from said image;

means for representing each of a plurality of portions of each of said training images by a pixel value, each of said portions of each of said training images respectively corresponding to an associated one of said portions of said template image;

means for, for each portion of each of said training images, determining the deviation between the pixel value for that portion of that training image and the pixel value for the associated template image portion; and

means for identifying as portions of said image that are at or near regions of abrupt change in pixel value any portions of any of said training images for which said deviation is greater than a predetermined amount.

14. The apparatus defined in claim 13 wherein said means for identifying as a region of abrupt change in pixel value comprises:

means for binarizing said deviations by assigning a first binary value to all portions of each training image for which said deviation is less than a predetermined amount, and by assigning a second binary value to all portions of each training image for which said difference is greater than said predetermined amount; and

means for identifying as portions of said image that are at or near regions of abrupt change in pixel value any portions of any of said training images which have said second binary value.