Machine vision system for analyzing features based on multiple object images

• US 6,175,652 B1
• Filed: 12/31/1997
• Issued: 01/16/2001
• Est. Priority Date: 12/31/1997
• Status: Expired due to Fees

• Current Assignee: Cognex Corporation
• Original Assignee: Cognex Corporation
• Inventors: Ruzhitsky, Vladimir N., Jacobson, Lowell D.
• Primary Examiner(s): Tran, Phuoc
• Assistant Examiner(s): Mariam, Daniel G.
• Application Number:
  US09/002,190
• Time in Patent Office: 1,112 Days
• Field of Search: 382/181, 382/209, 382/216, 382/219, 382/294, 382/203, 382/278, 348/43, 348/50, 348/159, 117/14, 117/201
• US Class Current: 382/216
• CPC Class Codes: C30B 15/26 using television detectors; using photo or X-ray detectors
• CPC Class Codes: G01B 11/24 for measuring contours or curvatures

Abstract

Provided is the ability to produce an orthogonal-view representation of a selected feature plane of a three-dimensional object. A plurality of images of the object are acquired, each corresponding to a distinct orientation of the object about a selected object axis. In at least one acquired image, feature points are identified in the selected feature plane as-projected into that acquired image. Feature points are associated with an orthogonal-view representation of the selected feature plane, and feature points from at least one acquired image are correlated with physical orientations on the selected feature plane based on the object orientation corresponding to that acquired image. An orthogonal-view representation of the selected object feature plane can be analyzed for a specified feature configuration even when the acquired object images are perspective-view images. This can be accomplished even through only a subset of feature points may be available in any one given image of the object. Such may be the case, e.g., where portions of a complicated three-dimensional object obscure other portions of the object in a selected view of the object. Also provided is an ability to analyze an orthogonal-view representation of a circumferential object contour, e.g., a meniscus defined on a semiconductor melt surface from which a semiconductor crystal ingot is horizontally pulled to grow the crystal ingot. The invention enables monitoring of the crystal growth for faults occurring during the growth process.

Claims

1. A machine vision method for producing a representation of a selected feature plane of an object in a physical coordinate system of the object, the method comprising:

acquiring a plurality of images of the object that each correspond to a distinct orientation of the object about a selected object axis;

identifying, in at least one acquired image, feature points in the selected feature plane as-projected into that acquired image;

associating feature points with the selected feature plane in the physical coordinate system of the object; and

correlating feature points from at least one acquired image with physical orientations on the selected feature plane in the physical coordinate system of the object, based on the object orientation corresponding to that acquired image and wherein acquiring a plurality of object images;

comprises acquiring a sequence of object images, and wherein correlating feature points comprises correlating feature points to a physical orientation on the selected feature plane based on order position of that image in the sequence of images and based on difference in object orientation between adjacent images in the sequence.

3. The method of claim 1 wherein acquiring a plurality of object images comprises acquiring a sequence of object images from a fixed viewing location as the object rotates about the selected object axis.

4. The method of claim 1 wherein acquiring a plurality of object images comprises acquiring a sequence of object images by rotating an image acquisition camera about the selected object axis as the camera produces object images.

10. The method of claim 1 wherein acquiring a plurality of object images comprises acquiring a sequence of object images, a portion of each image in the sequence being common with a portion of an adjacent image in the sequence.

11. The method of claim 1 wherein acquiring a plurality of object images comprises acquiring a video sequence of object images.

12. The method of claim 3 wherein acquiring a plurality of object images comprises acquiring a sequence of at least a minimum number, M, of object images, where M is selected as M=T₀/T_a, where T₀ is a period of revolution of the object and T_a is a period required for acquiring one image.

13. The method of claim 1 wherein acquiring a sequence of object images comprises acquiring the image sequence from a fixed viewing location as the object rotates about the selected object axis, and wherein correlating feature points with physical orientations is determined by order position of images in the sequence and by angular object orientation offset between adjacent images in the sequence.

14. The method of claim 1 wherein identifying feature points in an image comprises searching for feature points in an area of the image where feature points are expected.

15. The method of claim 14 wherein searching for feature points in an area of an image where feature points are expected comprises:

applying a plurality of edge detection regions in that image at image positions corresponding to an expected image location of the feature plane projection;

searching each edge detection region for an edge of a feature point; and

correlating positions of detected edges to feature point positions in that image.

16. The method of claim 15 wherein searching for feature points in an area of an image where feature points are expected further comprises adjusting the expected location of the feature plane projection if a feature point edge is not detected in at least a specified minimum number of edge detection regions.

17. The method of claim 1 further comprising analyzing the representation of the selected feature plane for a specified feature configuration.

2. A machine vision method for producing a representation of a selected feature plane of an object in a physical coordinate system of the object, the method comprising:

acquiring a plurality of perspective-view images of the object that each correspond to a perspective view of a distinct orientation of the object about a selected object axis;

identifying, in at least one acquired perspective-view image, feature points in the selected feature plane as-projected into that acquired image;

mapping feature points from the perspective-view image projection to the selected feature plane in the physical coordinate system of the object; and

correlating feature points from at least one acquired image with physical orientations on the selected feature plane in the physical coordinate system of the object, based on the object orientation corresponding to that acquired image; and

wherein acquiring a plurality of object images comprises acquiring a sequence of object images, and wherein correlating feature points comprises correlating feature points to a physical orientation on the selected feature plane based on order position of that image in the sequence of images and based on difference in object orientation between adjacent images in the sequence.

5. The method of claim 2 wherein mapping feature points comprises mapping feature points from a two-dimensional image coordinate system that is parallel with a selected perspective view to a three-dimensional coordinate system that is aligned with the selected object axis.

6. The method of claim 5 wherein the selected feature plane of the object is a plane perpendicular to the selected object axis.

7. The method of claim 6 wherein acquiring a plurality of perspective-view object images comprises acquiring a sequence of perspective-view object images from a fixed viewing location as the object rotates about the selected object axis.

8. The method of claim 7 wherein the selected object axis is a vertical axis of revolution and wherein the selected feature plane is a horizontal plane located at a position along the axis of revolution.

9. The method of claim 8 wherein identifying feature points in the selected horizontal feature plane comprises detecting circumferential feature points of the object.

18. A machine vision method for producing a representation of a selected feature plane of an object, in a physical coordinate system of the object, as the object rotates about an axis of rotation that is perpendicular to the selected feature plane, the method comprising:

acquiring a sequence of perspective-view object images from a fixed viewing location as the object rotates about the axis of rotation, each perspective-view image corresponding to a perspective view of a distinct angular orientation of the object about the axis of object rotation;

identifying, in at least one acquired perspective-view image, feature points in the selected feature plane as-projected into that acquired image;

mapping feature points from the perspective-view image projection to the selected feature plane in the physical coordinate system of the object; and

correlating feature points from at least one acquired image with physical orientations on the selected feature plane in the physical coordinate system of the object, based on the distinct angular object orientation between adjacent images in the sequence and order position of that acquired image in the image sequence.

19. A machine vision method for analyzing a representation of a semiconductor melt surface from which a semiconductor crystal ingot is vertically pulled, in a physical coordinate system of the ingot, as the ingot is rotated in a first direction about a vertical axis of rotation and the melt is rotated in a second direction about the vertical axis of rotation, the method comprising:

acquiring a sequence of perspective-view images of the melt surface from a fixed viewing location, each perspective-view image corresponding to a perspective view of a distinct angular orientation of the ingot and the melt surface about the axis of rotation;

identifying, in at least one acquired perspective-view image, feature points in the melt surface as-projected into that image;

mapping identified feature points from the perspective-view image projection to the melt surface in the physical coordinate system of the ingot;

correlating mapped melt feature points from at least one acquired image with physical orientations on the melt surface in the physical coordinate system of the ingot, based on the distinct angular orientation corresponding to that acquired image and order position of that acquired image in the image sequence; and

analyzing the the melt surface features for a prespecified feature configuration.

20. A machine vision method for analyzing a representation of a selected cross-sectional circumferential contour of an object, in a physical coordinate system of the object, as the object rotates about an axis of rotation, the method comprising:

acquiring a sequence of perspective-view object images from a fixed viewing location as the object rotates about the axis of rotation, each perspective-view image corresponding to a perspective view of a distinct angular orientation of the object about the axis of object rotation;

identifying, in at least one acquired perspective-view image, circumferential contour points in a feature plane corresponding to the selected cross-sectional circumferential contour and as-projected into that acquired image;

mapping circumferential contour points from the perspective-view image projection to the selected circumferential contour in the physical coordinate system of the object;

correlating circumferential contour points from at least one acquired image with physical locations around the circumferential contour in the physical coordinate system of the object, based on the distinct angular object orientation corresponding to that acquired image and order position of that acquired image in the image sequence; and

analyzing the correlated circumferential contour points based on an expected configuration of features along the contour.

21. The method of claim 20 wherein the feature plane corresponding to the selected cross-sectional circumferential contour is a plane that is perpendicular to the axis of object rotation.

22. The method of claim 20 wherein analyzing the correlated circumferential contour points comprises:

detecting contour features along the circumferential contour; and

analyzing location of the detected contour features based on an expectation for feature location.

23. The method of claim 20 wherein analyzing the correlated circumferential contour points comprises:

detecting contour features along the circumferential contour; and

analyzing number of the detected contour features based on an expectation for feature number.

24. The method of claim 20 wherein analyzing the correlated circumferential contour points comprises:

detecting contour features along the circumferential contour; and

analyzing shape of the detected contour features based on an expectation for feature shape.

25. The method of claim 20 further comprising producing an indication of a degree of compliance of the circumferential contour with an expected contour feature configuration.

26. The method of claim 20 wherein analyzing the correlated circumferential contour points comprises:

determining a center point of the circumferential contour;

determining a radius for each contour point referenced to the center point;

determining for the contour points a mean center point-to-contour point radius;

producing an offset radius for each contour point corresponding to a difference between that contour point'"'"'s radius and the mean radius;

correlating each radius offset with an arc along the contour, each arc specified with an arc radius offset value that is an average of all radius offset values correlated with that arc;

estimating a statistical fluctuation of radius offset values for a plurality of arcs known to not include a contour feature;

applying a radius offset threshold, based on the estimated radius offset value statistical fluctuation, to detect candidate contour feature locations along the contour arcs;

applying a prespecified feature width threshold to the candidate contour feature locations to confirm contour features of the circumferential contour; and

analyzing the detected contour features based on an expected contour feature configuration.

27. A machine vision method for analyzing representation of a cross-sectional meniscus contour produced by a semiconductor ingot on a semiconductor melt surface, in a physical coordinate system of the ingot, as the ingot is vertically pulled out of the melt surface and the ingot and melt are rotated in opposite directions about a common vertical axis of rotation, the method comprising:

acquiring a sequence of perspective-view images of the ingot and melt from a fixed viewing location, each perspective-view image corresponding to a perspective view of a distinct angular orientation of the ingot and melt surface about the axis of rotation;

identifying, in at least one acquired perspective-view image, meniscus contour points as-projected into that acquired image;

mapping identified meniscus contour points from the perspective-view image projection to the meniscus contour in the physical coordinate system of the ingot;

correlating mapped meniscus contour points from at least one acquired image with physical orientations around the meniscus in the physical coordinate system of the ingot, based on the distinct angular orientation corresponding to that image and order position of that image in the image sequence; and

analyzing the correlated meniscus contour points to determine if a fault condition has occurred.

28. The method of claim 27 wherein analyzing the correlated meniscus contour points comprises:

detecting ingot crystalline facets along the meniscus contour; and

analyzing locations of detected crystalline facets based on an expectation for facet location.

29. The method of claim 27 wherein analyzing the correlated meniscus contour points comprises:

detecting ingot crystalline facets along the meniscus contour; and

analyzing number of the detected crystalline facets based on an expectation for facet number.

30. The method of claim 27 wherein analyzing the correlated meniscus contour points comprises:

detecting ingot crystalline facets along the meniscus contour; and

analyzing shape of detected crystalline facets based on an expectation for facet shape symmetry.

31. The method of claim 27 further comprising producing an indication of a fault condition for the semiconductor ingot if the meniscus contour is found to include crystalline facets that do not comply with an expected facet configuration.

32. The method of claim 27 wherein analyzing the correlated meniscus contour points comprises:

determining a center point of the meniscus contour;

determining a radius for each contour point referenced to the center point;

determining for the contour points a mean center point-to-contour point radius;

producing an offset radius for each contour point corresponding to a difference between that contour point'"'"'s radius and the mean radius;

correlating each radius offset to an arc along the contour, each arc specified with an arc radius offset value that is an average of all radius offset values correlated to that arc;

estimating a statistical fluctuation of radius offset values for a plurality of arcs known to not include an ingot crystalline facet;

applying a radius offset threshold, based on the estimated radius offset value statistical fluctuation, to detect candidate crystalline facet locations along the contour arcs;

applying a prespecified crystalline facet width threshold to the candidate crystalline facet locations to confirm detected crystalline facets; and

analyzing the detected crystalline facets based on an expected facet configuration.

33. A machine vision system for producing a representation of a selected feature plane of an object in a physical coordinate system of the object, the system comprising:

means for acquiring a plurality of images of the object that each correspond to a distinct orientation of the object about a selected object axis;

means for identifying, in at least one acquired image, feature points in the selected feature plane as-projected into that acquired image;

means for associating feature points with the selected feature plane in the physical coordinate system of the object; and

means for correlating feature points from at least one acquired image with physical orientations on the selected feature plane in a physical coordinate system of the object, based on the object orientation corresponding to that acquired image; and

wherein acquiring a plurality of object images comprises acquiring a sequence of object images, and wherein correlating feature points comprises correlating feature points to a physical orientation on the selected feature plane based on order position of that image in the sequence of images and based on difference in object orientation between adjacent images in the sequence.

34. A machine vision system for producing a representation of a selected feature plane of an object in a physical coordinate system of the object, the system comprising:

means for acquiring a plurality of perspective-view images of the object that each correspond to a perspective view of a distinct orientation of the object about a selected object axis;

means for identifying, in at least one acquired perspective-view image, feature points in the selected feature plane as-projected into that acquired image;

means for mapping feature points from the perspective-view image projection to the selected feature plane in the physical coordinate system of the object; and

means for correlating feature points from at least one acquired image with physical orientations on the selected feature plane in the physical coordinate system of the object, based on the object orientation corresponding to that acquired image; and

wherein acquiring a plurality of object images comprises acquiring a sequence of object images, and wherein correlating feature points comprises correlating feature points to a physical orientation on the selected feature plane based on order position of that image in the sequence of images and based on difference in object orientation between adjacent images in the sequence.

35. A machine vision system for producing a representation of a selected feature plane of an object, in a physical coordinate system of the object, as the object rotates about an axis of rotation that is perpendicular to the selected feature plane, the system comprising:

means for acquiring a sequence of perspective-view object images from a fixed viewing location as the object rotates about the axis of rotation, each perspective-view image corresponding to a perspective view of a distinct angular orientation of the object about the axis of object rotation;

means for identifying, in at least one acquired perspective-view image, feature points in the selected feature plane as-projected into that acquired image;

means for mapping feature points from the perspective-view image projection to the selected feature plane in the physical coordinate system of the object; and

means for correlating feature points from at least one acquired image with physical orientations on the selected feature plane in the physical coordinate system of the object, based on the distinct angular object orientation between adjacent imges in the sequene and order position of that acquired image in the image sequence.

36. A machine vision system for analyzing a representation of a semiconductor melt surface from which a semiconductor crystal ingot is vertically pulled, in a physical coordinate system of the ingot, as the ingot is rotated in a first direction about a vertical axis of rotation and the melt is rotated in a second direction about the vertical axis of rotation, the system comprising:

means for acquiring a sequence of perspective-view images of the melt surface from a fixed viewing location, each perspective-view image corresponding to a perspective view of a distinct angular orientation of the ingot and the melt surface about the axis of rotation;

means for identifying, in at least one acquired perspective-view image, feature points in the melt surface as-projected into that image;

means for mapping identified feature points from the perspective-view image projection to the melt surface in the physical coordinate system of the ingot;

means for correlating mapped melt feature points from at least one acquired image with physical orientations on the melt surface in the physical coordinate system of the ingot, based on the distinct angular orientation corresponding to that acquired image and order position of that acquired image in the image sequence; and

means for analyzing the the melt surface features for a prespecified feature configuration.

37. A machine vision system for analyzing a representation of a selected cross-sectional circumferential contour of an object, in a physical coordinate system of the object, as the object rotates about an axis of rotation, the system comprising:

means for acquiring a sequence of perspective-view object images from a fixed viewing location as the object rotates about the axis of rotation, each perspective-view image corresponding to a perspective view of a distinct angular orientation of the object about the axis of object rotation;

means for identifying, in at least one acquired perspective-view image, circumferential contour points in a feature plane corresponding to the selected cross-sectional circumferential contour and as-projected into that acquired image;

means for mapping circumferential contour points from the perspective-view image projection to the selected circumferential contour in the physical coordinate system of the object;

means for correlating circumferential contour points from at least one acquired image with physical locations around the circumferential contour in the physical coordinate system of the object, based on the distinct angular object orientation corresponding to that acquired image and order position of that acquired image in the image sequence; and

means for analyzing the correlated circumferential contour points based on an expected configuration of features along the contour.

38. A machine vision system for analyzing a representation of a cross-sectional meniscus contour produced by a semiconductor ingot on a semiconductor melt surface, in a physical coordinate system of the ingot, as the ingot is vertically pulled out of the melt surface and the ingot and melt are rotated in opposite directions about a common vertical axis of rotation, the system comprising:

means for acquiring a sequence of perspective-view images of the ingot and melt from a fixed viewing location, each perspective-view image corresponding to a perspective view of a distinct angular orientation of the ingot and melt surface about the axis of rotation;

means for identifying, in at least one acquired perspective-view image, meniscus contour points as-projected into that acquired image;

means for mapping identified meniscus contour points from the perspective-view image projection to the meniscus contour in the physical coordinate system of the ingot;

means for correlating mapped meniscus contour points from at least one acquired image with physical orientations around the meniscus in the physical coordinate system of the ingot, based on the distinct angular orientation corresponding to that image and order position of that image in the image sequence; and

means for analyzing the correlated meniscus contour points to determine if a fault condition has occurred.