Method and apparatus for classifying an object

• US 7,103,213 B2
• Filed: 03/02/2005
• Issued: 09/05/2006
• Est. Priority Date: 03/02/2004
• Status: Active Grant

• Current Assignee: ZAMA Innovations LLC
• Original Assignee: Sarnoff Corporation (SRI International, Inc.)
• Inventors: Hirvonen, David, Camus, Theodore Armand
• Primary Examiner(s): SHERALI, ISHRAT I
• Application Number:
  US11/070,566
• Publication Number:
  US 20050270286A1
• Time in Patent Office: 552 Days
• Field of Search: 382153-154, 382103-104, 382/224, 382/285, 382/209, 382262-264, 348143-149, 348/42, 348/47, 348150-154, 348/160, 348/161, 348169-172, 340435-436, 340/903, 340/961, 340/937, 340/925, 340/555, 340/565, 340/425.5, 345418-419
• US Class Current: 382/154
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
• CPC Class Codes: G06V 20/653 by matching three-dimensional models, e.g. conformal mapping of Riemann surfaces

Abstract

A method and apparatus for classifying an object in an image is disclosed. A depth image is provided. At least one area of the depth image unsatisfactory for object identification are identified. A plurality of two-dimensional projections of surface normals in the depth image is determined without considering the unsatisfactory at least one area. One or more objects are classified based on the plurality of two-dimensional projections of surface normals.

Claims

1. A method of classifying an object in an image, comprising:

providing a depth image;

identifying at least one area of said depth image unsatisfactory for object identification;

determining a plurality of two-dimensional projections of surface normals in said depth image without considering said unsatisfactory at least one area; and

classifying one or more objects based on the plurality of two-dimensional projections of surface normals.

2. The method of claim 1, wherein identifying areas of said depth image comprises determining Zdy where Zdy defines how three-dimensional depth coordinate (Z) changes across a ⅓

meter vertical interval projected into image rows.

3. The method of claim 2, wherein areas of said depth image are eliminated from consideration due to having a high Zdy.

4. The method of claim 1, wherein identifying areas of said depth image comprises determining pitch psi, where psi defines a pitch of a dominant road plane.

5. The method of claim 4, wherein areas of said depth image are eliminated from consideration due to having a low pitch psi.

6. The method of claim 1, wherein determining said plurality of two-dimensional projections of surface normals comprises:

determining Xdx where Xdx defines how three-dimensional lateral coordinate (X) changes across a ⅓

meter horizontal interval projected into image columns;

determining Zdx where Zdx defines how three-dimensional depth coordinate (Z) changes across a ⅓

meter horizontal interval projected into image columns; and

determining theta using an A tan 2 function with Xdx and Zdx as parameters in areas of the depth image that are satisfactory for object identification.

7. The method of claim 6, wherein a subset of said plurality of two-dimensional projections of surface normals comprises a collection of theta within a certain threshold.

8. The method of claim 1, wherein classifying said one or more objects comprises classifying said one or more objects as a side of a vehicle when a subset of said plurality of two-dimensional projections of surface normals is approximately orthogonal to an optical axis of a host.

9. The method of claim 1, wherein classifying said one or more objects comprises classifying using object width when over fifty percent of a subset of said plurality of two-dimensional projections of surface normals are within a range of +/−

20 degrees of being orthogonal to an optical axis of a host.

10. An apparatus for classifying an object in an image, comprising:

means for providing a depth image;

means for identifying at least one area of said depth image unsatisfactory for object identification;

means for determining a plurality of two-dimensional projections of surface normals in said depth image without considering said unsatisfactory at least one area; and

means for classifying one or more objects based on the plurality of two-dimensional projections of surface normals.

11. The apparatus of claim 10, wherein identifying areas of said depth image comprises determining Zdy where Zdy defines how three-dimensional depth coordinate (Z) changes across a ⅓

meter vertical interval projected into image rows.

12. The apparatus of claim 11, wherein areas of said depth image are eliminated from consideration due to having a high Zdy.

13. The apparatus of claim 10, wherein identifying areas of said depth image comprises determining pitch psi, where psi defines a pitch of a dominant road plane.

14. The apparatus of claim 13, wherein areas of said depth image are eliminated from consideration due to having a low pitch psi.

15. The apparatus of claim 10, wherein determining said plurality of two-dimensional projections of surface normals comprises:

determining Xdx where Xdx defines how three-dimensional lateral coordinate (X) changes across a ⅓

meter horizontal interval projected into image columns;

determining Zdx where Zdx defines how three-dimensional depth coordinate (Z) changes across a ⅓

meter horizontal interval projected into image columns; and

determining theta using an A tan 2 function with Xdx and Zdx as parameters in areas of the depth image that are satisfactory for object identification.

16. The apparatus of claim 15, wherein a subset of said plurality of two-dimensional projections of surface normals comprises a collection of theta within a certain threshold.

17. The apparatus of claim 10, wherein classifying said one or more objects comprises classifying said one or more objects as a side of a vehicle when a subset of said plurality of two-dimensional projections of surface normals is approximately orthogonal to an optical axis of a host.

18. The apparatus of claim 10, wherein classifying said one or more objects comprises classifying using object width when over fifty percent of a subset of said plurality of two-dimensional projections of surface normals are within a range of +/−

20 degrees of being orthogonal to an optical axis of a host.

19. A computer-readable medium having stored thereon a plurality of instructions, the plurality of instructions including instructions which, when executed by a processor, cause the processor to perform the steps of a method of classifying an object in an image, comprising:

providing a depth image;

identifying at least one area of said depth image unsatisfactory for object identification;

determining a plurality of two-dimensional projections of surface normals in said depth image without considering said unsatisfactory at least one area; and

classifying one or more objects based on the plurality of two-dimensional projections of surface normals.

20. The computer-readable medium of claim 19, wherein determining said plurality of two-dimensional projections of surface normals comprises:

determining Xdx where Xdx defines how three-dimensional lateral coordinate (X) changes as a function of depth image column (x);

determining Zdx where Zdx defines how three-dimensional depth coordinate (Z) changes across a ⅓

meter horizontal interval (projected into image columns (x)); and

determining theta by using an A tan 2 function with Xdx and Zdx as parameters in areas of the depth image that are satisfactory for object identification.