Morphological based segmenter

• US 7,899,245 B2
• Filed: 12/30/2002
• Issued: 03/01/2011
• Est. Priority Date: 12/30/2002
• Status: Active Grant

• Current Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Original Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Inventors: Olson, Teresa
• Primary Examiner(s): Ge; Yuzhen
• Application Number:
  US10/330,088
• Publication Number:
  US 20040126013A1
• Time in Patent Office: 2,983 Days
• Field of Search: 382/173
• US Class Current: 382/173
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/10048 Infrared image
• CPC Class Codes: G06T 2207/20104 Interactive definition of region of interest [ROI]
• CPC Class Codes: G06T 2207/20132 Image cropping
• CPC Class Codes: G06T 2207/30212 Military
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/155 involving morphological operators
• CPC Class Codes: G06T 7/246 using feature-based methods, e.g. the tracking of corners or segments

Abstract

A method for segmenting an object within an image including extracting an edge image containing a first set of pixels from the image, generating a second set of pixels from the first set of pixels using mathematical morphology, and identifying a segment corresponding to the object.

Claims

1. A method for segmenting an object within an image in a tracking device, comprising:

placing a processing window within the image that is centered about a centroid of the object, wherein the processing window comprises a subarea of the image, and dimensions and position of the processing window are based on priori information regarding size of a pre-screener box for the object and location of a centroid of the object based on input from an operator;

extracting, in the tracking device, from the subarea of the image within the processing window, an edge image containing a first set of pixels from the object;

generating, in the tracking device, a second set of pixels from the first set of pixels using mathematical morphology to fill in pixels of the edge image;

identifying, in the tracking device, a segment corresponding to the object from the second set of pixels; and

autonomously tracking the object with a pre-screener in the tracking device and adjusting the size of the processing window to prevent a signature of the object from outgrowing the processing window to permit subsequent morphological processing of the object, wherein a size of the signature changes in accordance with a distance between the pre-screener and the object.

2. The method for segmenting an object within an image according to claim 1, wherein the first set of pixels have at least a predetermined gradient magnitude.

3. The method for segmenting an object within an image according to claim 1, wherein dimensions and position of the processing window are based on priori information regarding size of the pre-screener box for the object and location of the centroid for the object from the pre-screener in the subsequent morphological processing of the object.

4. The method for segmenting an object within an image according to claim 1, wherein dimensions of the processing window are at least two times greater than dimensions of the signature for the object within the image.

5. The method for segmenting an object within an image according to claim 1, wherein dimensions of the processing window are at least two times greater than dimensions of the pre-screener box.

6. The method for segmenting an object within an image according to claim 1, wherein the mathematical morphology dilates the first set of pixels N times and then erodes N times to generate the second set of pixels.

7. The method for segmenting an object within an image according to claim 5, wherein the mathematical morphology dilates the first set of pixels N number of times and then erodes N number of times to generate the second set of pixels and N is equal to a lesser dimension of the pre-screener box in pixels divided by four and then rounded up to a whole integer.

8. The method for segmenting an object within an image according to claim 6, wherein N is equal to a lesser dimension of the object in pixels based on priori information regarding size of the object divided by four and then rounded up to a whole integer.

9. The method for segmenting an object within an image according to claim 1, wherein identifying a segment includes determining a third set of pixels representing the object from the second set of pixels.

10. The method for segmenting an object within an image according to claim 5, wherein identifying a segment includes determining a third set of pixels representing the object from the second set of pixels.

11. The method for segmenting an object within an image according to claim 9, wherein the third set of pixels are determined based on a predetermined centroid shape.

12. The method for segmenting an object within an image according to claim 10, wherein the third set of pixels are determined based on a predetermined centroid shape and dimensions of the centroid shape are no more than half of the dimensions of the pre-screener box.

13. The method for segmenting an object within an image according to claim 9, wherein identifying a segment includes marking pixels within the third set of pixels that represent a perimeter of the object within the image.

14. The method for segmenting an object within an image according to claim 12, wherein identifying a segment includes marking pixels within the third set of pixels that represent a perimeter of the object.

15. The method for segmenting an object within an image according to claim 11, wherein the predetermined centroid shape is from the pre-screener.

16. The method for segmenting an object within an image according to claim 1, wherein extracting an edge image from the subarea of the image within the window containing a first set of pixels from the object includes extracting a gradient edge image by applying an edge detection operator to the subarea of the image within the window.

17. The method for segmenting an object within an image according to claim 16, wherein the edge detection operator comprises a Sobel edge operator.

18. A method for segmenting an object within an image in a tracking device, comprising:

placing a processing window within the image that is centered about a centroid of the object, wherein the processing window comprises a subarea of the image, and dimensions and position of the processing window are based on priori information regarding size of a pre-screener box for the object and location of a centroid of the object based on input from a pre-screener in the tracking device autonomously tracking the object;

extracting, in the tracking device, from the subarea of the image within the processing window, an edge image containing a first set of pixels from the object;

generating, in the tracking device, a second set of pixels from the first set of pixels using mathematical morphology to fill in pixels of the edge image;

identifying, in the tracking device, a segment corresponding to the object from the second set of pixels; and

adjusting, in the tracking device, the size of the processing window to prevent a signature of the object from outgrowing the processing window to permit subsequent morphological processing of the object, wherein size of the signature changes in accordance with a distance between the pre-screener and the object.

19. The method for segmenting an object within an image according to claim 18, wherein the first set of pixels have at least a predetermined gradient magnitude.

20. The method for segmenting an object within an image according to claim 18, wherein dimensions of the processing window are at least two times greater than dimensions of the signature for the object within the image.

21. The method for segmenting an object within an image according to claim 18, wherein dimensions of the processing window are at least two times greater than dimensions of the pre-screener box.

22. The method for segmenting an object within an image according to claim 18, wherein the mathematical morphology dilates the first set of pixels N times and then erodes N times to generate the second set of pixels.

23. The method for segmenting an object within an image according to claim 21, wherein the mathematical morphology dilates the first set of pixels N number of times and then erodes N number of times to generate the second set of pixels and N is equal to a lesser dimension of the pre-screener box in pixels divided by four and then rounded up to a whole integer.

24. The method for segmenting an object within an image according to claim 22, wherein N is equal to a lesser dimension of the object in pixels based on priori information regarding size of the object divided by four and then rounded up to a whole integer.

25. The method for segmenting an object within an image according to claim 18, wherein identifying a segment includes determining a third set of pixels representing the object from the second set of pixels based on a predetermined centroid shape from the pre-screener.

26. The method for segmenting an object within an image according to claim 21, wherein identifying a segment includes determining a third set of pixels representing the object from the second set of pixels based on a predetermined centroid shape and wherein dimensions of the predetermined centroid shape are no more than half of the dimensions of the pre-screener box.

27. The method for segmenting an object within an image according to claim 18, wherein identifying a segment includes determining a third set of pixels representing the object from the second set of pixels and marking pixels within the third set of pixels that represent a perimeter of the object within the image.

28. The method for segmenting an object within an image according to claim 21, wherein identifying a segment includes determining a third set of pixels representing the object from the second set of pixels and marking pixels within the third set of pixels that represent a perimeter of the object.

29. The method for segmenting an object within an image according to claim 18, wherein extracting an edge image from the subarea of the image within the window containing a first set of pixels from the object includes extracting a gradient edge image by applying an edge detection operator to the subarea of the image within the window.

30. The method for segmenting an object within an image according to claim 29, wherein the edge detection operator comprises a Sobel edge operator.