Three-dimensional face recognition

• US 7,623,687 B2
• Filed: 09/06/2005
• Issued: 11/24/2009
• Est. Priority Date: 10/07/2002
• Status: Active Grant

• Current Assignee: Intel Benelux B. V.
• Original Assignee: TECHNION RESEARCH AND DEVELOPMENT FOUNDATION, LTD.
• Inventors: Bronstein, Michael, Bronstein, Alexander, Kimmel, Ron
• Primary Examiner(s): Desire; Gregory M
• Application Number:
  US11/218,597
• Publication Number:
  US 20060251298A1
• Time in Patent Office: 1,540 Days
• Field of Search: 345/419, 345/420, 345/428, 345/441, 382/118, 382/225, 382/274, 382/299, 706/15, 706/20
• US Class Current: 382/118
• CPC Class Codes: G01S 17/89 for mapping or imaging
• CPC Class Codes: G06T 17/00 Three dimensional [3D] modelling, e.g. data description of 3D objects
• CPC Class Codes: G06V 20/653 by matching three-dimensional models, e.g. conformal mapping of Riemann surfaces
• CPC Class Codes: G06V 40/172 Classification, e.g. identification

Abstract

Apparatus for obtaining 3-Dimensional data of a geometric body for matching, and particularly for use in facial matching, comprises a three dimensional scanner for obtaining three-dimensional topographical data of the body, a triangulator for receiving or forming said data into a triangulated manifold, a geodesic converter, for converting the triangulated manifold into a series of geodesic distances between pairs of points of the manifold, and a multi-dimensional scaler, for forming a low dimensional Euclidean representation of the series of geodesic distances, to give a bending invariant representation of the geometric body. In one variant, matching is carried out by taking the principle eigenvalues from the representation and plotting as co-ordinates in a feature space. Tilted or different expression versions of the same face tend to form clusters in the feature space allowing for matching. The apparatus preferably uses the fast marching method for the triangulated domain to obtain the geodesic distances.

Claims

1. Apparatus for processing 3-dimensional data of a geometric body for matching, said geometric body comprising an outer surface having a topography, said apparatus comprising:

a geodesic converter, configured for receiving 3-dimensional topographical data of said geometric body, and for converting said 3-dimensional topographical data into a series of geodesic distances between pairs of points of said topographical data on a manifold, said geodesic distances being measured over said manifold of said geometric body, and

a bending invariant multi-dimensional scaler, connected subsequently to said geodesic converter, for forming a bending-invariant low dimensional Euclidean representation of said series of geodesic distances over said manifold, said bending-invariant low dimensional Euclidean representation providing a bending invariant representation of said geometric body suitable for matching with other geometric shapes.

2. Apparatus according to claim 1, further comprising a subsampler located prior to said geodesic converter, configured to subsample said 3-dimensional topographical data, and to provide to said geodesic converter a subsampled data set.

3. Apparatus according to claim 2, wherein said three-dimensional topographical data comprises a plurality of vertices and wherein said subsampler is operable to select a first vertex and to iteratively select a next vertex having a largest geodesic distance from vertices already selected, until a predetermined number of vertices has been selected.

4. Apparatus according to claim 3, wherein said subsampler is operable to use the fast marching method for triangulated domains to calculate geodesic distances between vertices for said iterative selection.

5. Apparatus according to claim 2, wherein said geometric body is a face, having soft geometric regions, being regions of said face susceptible to short term geometric changes, and hard geometric regions, being regions substantially insusceptible to said short term geometric changes, said apparatus comprising a preprocessor, located prior to said subsampler, for removing said soft geometric regions from said face.

6. Apparatus according to claim 5, wherein said preprocessor is operable to identify said soft regions by identification of an orientation point on said face.

7. Apparatus according to claim 6, wherein said orientation point is at least one of a nose tip, a pair of eyeball centers and a mouth center.

8. Apparatus according to claim 5, wherein said preprocessor is further operable to center said face.

9. Apparatus according to claim 5, wherein said preprocessor is further operable to crop said face.

10. Apparatus according to claim 5, wherein said preprocessor is operable to carry out removal of said soft regions by application of a geodesic mask.

11. Apparatus according to claim 1, wherein said geodesic converter is operable to use the fast marching method for triangulated domains to calculate said geodesic distances.

12. Apparatus according to claim 1, further comprising a triangulator for forming said topographical data from scan data of a geometric body.

13. Apparatus according to claim 12, further operable to embed said topographical data into a space of higher than two dimensions, thereby to include additional information with said topographical information.

14. Apparatus according to claim 13, wherein said additional information is any one of a group comprising texture information, albedo information, grayscale information, and color information.

15. Apparatus according to claim 1, wherein said subsampler comprises an optimizer for allowing a user to select an optimum subsampling level by defining a trade-off between calculation complexity and representation accuracy.

16. Apparatus according to claim 1, wherein said multi-dimensional scalar is configured such that said Euclidean representation comprises a predetermined number of eigenvalues extractable to be used as co-ordinates in a feature space.

17. Apparatus according to claim 16, wherein said predetermined number of eigenvalues is at least three, and said feature space has a number of dimensions corresponding to said predetermined number.

18. A method of image preprocessing of three-dimensional topographical data of a manifold surface having a topography for subsequent classification, the method comprising:

providing image preprocessing hardware; and

within said image preprocessing hardware;

placing said three-dimensional topographical data over a two-dimensional manifold to form said manifold surface having a topography,

converting said manifold into a series of geodesic distances between pairs of points over said manifold, said geodesic distances being measured over said topography of said manifold,

generating a matrix of geodesic distances over said manifold to selected vertices of said manifold, and

using multi-dimensional scaling to reduce said matrix to a bending-invariant canonical representation in a low-dimensional Euclidean space, thereby to provide a representation suitable for subsequent classification.

19. The method of claim 18, further comprising selecting said vertices for generating said matrix of geodesic distances by a procedure comprising iteratively selecting a next vertex having a largest geodesic distance from vertices already selected, until a predetermined number of vertices has been selected.

20. The method of claim 19, comprising determining geodesic distances for said procedure using the fast marching method for triangulated domains.

21. The method of claim 18, wherein said three-dimensional topographical data is a representation of a face, and further comprising cropping said face to exclude parts of said face being susceptible to short term changes, thereby to render said subsequent classification substantially invariant to said short term changes.

22. A method of classifying images of three-dimensional bodies having surface topographies, the method comprising:

providing image processing hardware; and

within said image processing hardware;

obtaining bending invariant representations of said three dimensional bodies as canonical form representations derived from geodesic distances between selected sample points taken from surfaces of said bodies, said geodesic distances being measured over said respective surface topographies of said three-dimensional bodies;

from each representation deriving co-ordinates on a feature space, and

classifying said bodies according to clustering on said feature space.

23. The method of claim 22, wherein said deriving co-ordinates comprises deriving first m eigenvalues from respective canonical form representations, where n is a predetermined number, and using said m eigenvalues to provide said co-ordinates, said feature space being constructed to have n dimensions.

24. The method of claim 23, wherein n is at least three.

25. The method of claim 23, wherein n is three.

26. Apparatus for processing 3-dimensional data of a geometric body having a three-dimensional surface topography, said processing being for matching with other bodies, said apparatus comprising:

a geodesic converter, configured for receiving 3-dimensional topographical data of said geometric body as a 2-dimensional manifold representing said surface, and for converting said manifold into a series of geodesic distances between pairs of points over said manifold, said geodesic distances being measured over a topography of said manifold, thereby to provide a Euclidian bending invariant representation of said geometric body suitable for matching with other geometric bodies, and wherein said geometric body for matching and said other geometric bodies comprises a face.

27. Apparatus according to claim 26, wherein said bending invariant representation permits matching of faces independently of facial expressions.