System and method for face recognition

• US 7,783,082 B2
• Filed: 06/30/2004
• Issued: 08/24/2010
• Est. Priority Date: 06/30/2003
• Status: Active Grant

• Current Assignee: Honda Motor Co., Ltd. (Honda Motor Company)
• Original Assignee: Honda Motor Co., Ltd. (Honda Motor Company)
• Inventors: Koshizen, Takamasa, Heisele, Bernd, Tsujino, Hiroshi
• Primary Examiner(s): Bella; Matthew C
• Assistant Examiner(s): Rush; Eric
• Application Number:
  US10/561,256
• Publication Number:
  US 20060280341A1
• Time in Patent Office: 2,246 Days
• Field of Search: 382/118, 382155-161
• US Class Current: 382/118
• CPC Class Codes: G06F 18/28 Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
• CPC Class Codes: G06V 10/772 Determining representative reference patterns, e.g. averaging or distorting patterns; Generating dictionaries
• CPC Class Codes: G06V 10/94 Hardware or software architectures specially adapted for image or video understanding
• CPC Class Codes: G06V 40/16 Human faces, e.g. facial parts, sketches or expressions
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships

Abstract

A face recognition system includes a component learning/extraction module, component classifier training module, knowledge base for component classification (KBCC), component extraction module (CEM), object identification training module (OITM), knowledge base for face identification (KBFI), and object identification module (OIM). The CEM receives image data of faces at various viewpoints and extracts outputs of classification of the component data, using the results of classifier training of the component data, stored in the KBCC. The OITM receives the outputs of classification of the component data and determines indicator component for each person by Bayesian estimation so that posterior probability of a predetermined attention class is maximized under the outputs of classification of the component data at various viewpoints. The KBFI stores indicator components for the individuals. The OIM receives the outputs of classification of the component data and identifies faces using the indicator components stored in the KBFI.

Claims

1. A method for recognizing faces of persons, comprising:

training a system to recognize a facial component;

populating a first knowledge base with facial components and, for each facial component, the facial component'"'"'s body part classification, wherein the facial components in the first knowledge base include a first set of facial components extracted from facial identification training image data of a face of a first person at a first set of viewpoints;

determining a first set of body part classifications associated with the first set of facial components;

for each body part classification in the first set of body part classifications;

determining, from the first set of facial components, a subset of facial components that are associated with the body part classification, wherein the facial components in the subset represent a same body part classification of a same person at multiple viewpoints; and

determining a probability that a person class of the subset of facial components is the first person; and

determining a first body part classification in the first set of body part classifications that maximizes the probability.

2. The method of claim 1, wherein the facial components in the first knowledge base further include a second set of facial components extracted from facial identification training image data of a face of a second person at a second set of viewpoints, and wherein the method further comprises:

determining a second set of body part classifications associated with the second set of facial components;

for each body part classification in the second set of body part classifications;

determining, from the second set of facial components, a subset of facial components that are associated with the body part classification; and

determining a probability that a person class of the subset of facial components is the second person; and

determining a second body part classification in the second set of body part classifications that maximizes the probability.

3. The method of claim 2, wherein the first body part classification and the second body part classification are different.

4. The method of claim 2, wherein the first set of viewpoints and the second set of viewpoints are different.

5. The method of claim 1, wherein determining the probability comprises determining the probability using Bayesian estimation.

6. The method of claim 1, wherein determining the probability that the person class of the subset of facial components is the first person comprises:

determining a first conditional probability, that a class is the first person, of a facial component of the subset of facial components at a first viewpoint;

determining a first posterior probability, that a class is the first person, by multiplying the conditional probability at the first viewpoint by a prior probability, that a class is the first person;

determining a second conditional probability, that a class is the first person, of a facial component of the subset of facial components at an additional viewpoint; and

determining a second posterior probability, that a class is the first person, by multiplying the second conditional probability by the first posterior probability.

7. The method of claim 6, wherein the prior probability, that the class is the first person, comprises one Nth where N is a number of person classes.

8. The method of claim 1, further comprising storing, in a second knowledge base, the facial components in the first set of facial components that are associated with the first body part classification.

9. The method of claim 8, further comprising:

receiving facial components at various viewpoints of a person to be identified; and

identifying the person using a facial component stored in the second knowledge base.

10. The method of claim 1, wherein training the system to recognize the facial component comprises:

receiving facial component training image data of faces of persons at various viewpoints;

extracting facial components at various viewpoints from the facial component training image data of faces of persons at various viewpoints; and

training a body part classifier module using the extracted facial components.

11. The method of claim 10, wherein the body part classifier module performs one-versus-all classification.

12. A system for recognizing faces of persons, comprising:

a training module configured to train a facial component recognition system to recognize a facial component;

a population module configured to populate a first knowledge base with facial components and, for each facial component, the facial component'"'"'s body part classification, wherein the facial components in the first knowledge base include a first set of facial components extracted from facial identification training image data of a face of a first person at a first set of viewpoints;

a computer readable storage medium storing an indicator component module configured to;

determine a first set of body part classifications associated with the first set of facial components;

for each body part classification in the first set of body part classifications;

determine, from the first set of facial components, a subset of facial components that are associated with the body part classification, wherein the facial components in the subset represent a same body part classification of a same person at multiple viewpoints; and

determine a probability that a person class of the subset of facial components is the first person; and

determine a first body part classification in the first set of body part classifications that maximizes the probability; and

a processor configured to execute the indicator component module stored by the computer readable storage medium.

13. The system of claim 12, wherein the facial components in the first knowledge base further include a second set of facial components extracted from facial identification training image data of a face of a second person at a second set of viewpoints, and wherein the indicator component module is further configured to:

determine a second set of body part classifications associated with the second set of facial components;

for each body part classification in the second set of body part classifications;

determine, from the second set of facial components, a subset of facial components that are associated with the body part classification; and

determine a probability that a person class of the subset of facial components is the second person; and

determine a second body part classification in the second set of body part classifications that maximizes the probability.

14. The system of claim 13, wherein the first body part classification and the second body part classification are different.

15. The system of claim 13, wherein the first set of viewpoints and the second set of viewpoints are different.

16. The system of claim 12, wherein the indicator component module is further configured to determine the probability using Bayesian estimation.

17. The system of claim 12, wherein the indicator component module is further configured to determine the probability that the person class of the subset of facial components is the first person by:

determining a first conditional probability, that a class is the first person, of a facial component of the subset of facial components at a first viewpoint;

determining a first posterior probability, that a class is the first person, by multiplying the conditional probability at the first viewpoint by a prior probability, that a class is the first person;

determining a second conditional probability, that a class is the first person, of a facial component of the subset of facial components at an additional viewpoint; and

determining a second posterior probability, that a class is the first person, by multiplying the second conditional probability by the first posterior probability.

18. The system of claim 17, wherein the prior probability, that the class is the first person, comprises one Nth where N is a number of person classes.

19. The system of claim 12, further comprising a storage module configured to store, in a second knowledge base, the facial components in the first set of facial components that are associated with the first body part classification.

20. The system of claim 19, further comprising:

a receiving module configured to receive facial components at various viewpoints of a person to be identified; and

an identification module configured to identify the person using a facial component stored in the second knowledge base.

21. The system of claim 12, wherein the training module is further configured to:

receive facial component training image data of faces of persons at various viewpoints;

extract facial components at various viewpoints from the facial component training image data of faces of persons at various viewpoints; and

train a body part classifier module using the extracted facial components.

22. The system of claim 21, wherein the body part classifier module performs one-versus-all classification.