Determining three-dimensional orientation of objects

• US 6,594,623 B1
• Filed: 12/30/1999
• Issued: 07/15/2003
• Est. Priority Date: 11/12/1999
• Status: Expired due to Term

• Current Assignee: Cognex Corporation
• Original Assignee: Cognex Technology & Investment Corporation (Cognex Corporation)
• Inventors: Wang, Lei, Persky, Andrew, Jacobson, Lowell
• Primary Examiner(s): Broda, Samuel
• Assistant Examiner(s): Thangavelu, Kandasamy
• Application Number:
  US09/475,084
• Time in Patent Office: 1,293 Days
• Field of Search: 703/6, 703/7, 703/2, 703/1, 382/151-154, 382/141, 382/191, 382/130, 382/156, 382/195, 382/190, 382/209, 382/103, 700/95-98, 700/91, 700/262, 700/213, 700/110, 700/264, 701/200, 701/216, 702/150-154, 702/95, 702/99
• US Class Current: 703/1
• CPC Class Codes: G06T 2207/30164 Workpiece; Machine component
• CPC Class Codes: G06T 7/75 involving models

Abstract

The three-dimensional orientation of a given object is determined. The given object may comprise an object having surface characteristics which provide poor contrast in representative images of the given object. In a specific embodiment, the object may comprise a golf ball. A model representation is obtained of a reference object, and a run-time image is obtained of the given object. From the run-time image, shape information is obtained for certain identified shapes collectively forming a subset of the visible shapes. The shape information is matched to a number of incremental poses of the model representation varying along three degrees of freedom.

Claims

1. A method for determining the 3-dimensional orientation of a given object, the method comprising:

obtaining a model representation of a reference object, said model representation comprising information concerning a reference orientation of said reference object and information representing repeating patterns of visible shapes in relation to said reference orientation;

obtaining a run-time image of said given object;

from said run-time image, obtaining shape information for certain identified shapes collectively forming a subset of said visible shapes;

performing a matching of said shape information to a number of incremental poses of said model representation varying along three degrees of freedom so as to determine the 3-dimensional orientation of the given object.

2. The method according to claim 1, wherein said given object comprises a mass-produced object.

3. The method according to claim 2, wherein said mass-produced object is inspected before packaging.

4. The method according the claim 2, wherein said mass-produced object comprises an object on which a marker will be placed at a predetermined position with respect to its orientation after the 3-dimensional orientation of the object is determined.

5. The method according to claim 1, wherein said given object has surface characteristics which provide poor contrast in representative images of said given object.

6. The method according to claim 5, wherein said given object comprises a golf ball.

7. The method according to claim 6, wherein said reference orientation comprises one of a north and south pole of a golf ball.

8. The method according to claim 1, wherein said shape information comprises, for a given identified shape, a location in three dimensions of the given identified shape and the size of the given identified shape.

9. The method according to claim 1, comprises placing said shape information for said subset into vectors of a run-time model.

10. The method according to claim 1, wherein said matching comprises:

coarse matching of said shape information to a number of candidate matching poses of said model representation from among a larger number of incremental poses of said model representation varying along three degrees of freedom, the coarse matching comprising;

(i) comparing the shapes within said subset with nearby pose shapes from the poses of said model representation,

(ii) for the subset of shapes, generating scores indicative of respective distances between the shapes and the nearby pose shapes, and

(iii) comparing the scores among poses to determine a closest subset of the poses which comprise the candidate matching poses; and

fine matching of said shape information to determine a precise matching pose, the fine matching comprising;

(i) comparing the shapes within said closest subset with corresponding pose shapes of a given candidate matching pose to determine an error value, and

(ii) adjusting the candidate matching pose along the three degrees of freedom using the error value so as to provide the precise matching pose thereby determining the 3-dimensional orientation of the given object.

11. A system for determining the three-dimensional orientation of a given object, said system comprising:

a population mechanism for populating a run-time model of the given object and a reference model, said reference model comprising information concerning a reference orientation of a reference object and information representing repeating patterns of visible shapes in relation to said reference orientation, and said run-time model comprising shape information for certain identified shapes collectively forming a subset of said visible shapes; and

a matching mechanism for performing a matching of said shape information to a number of incremental poses of said reference model varying along three degrees of freedom so as to determine the 3-dimensional orientation of the given object.

12. Computer-readable media comprising certain information thereon, the certain information, when read by a machine, causing:

obtaining a model representation of a reference object, said model representation comprising information concerning a reference orientation of said reference object and information representing repeating patterns of visible shapes in relation to said reference orientation;

from a run-time image of said given object, obtaining shape information for certain identified shapes collectively forming a subset of said visible shapes; and

performing a matching of said shape information to a number of incremental poses of said model representation varying along three degrees of freedom so as to determine the 3-dimensional orientation of the given object.