Face identification device

Abstract

A face feature value vector extracting section extracts a feature value from a registered image of a registrant to generate a face feature value vector. A virtual face feature value vector generating section converts the face feature value vector using a feature value converter which performs a conversion in a feature value space. Then, a registration information generating section generates face definition information of the registrant based on a face feature value vector before conversion and a virtual face feature value vector after conversion and registers this information in a registration information storing section.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 shows a block diagram of a functional configuration of a face identification device;

FIG. 2 shows a flow chart of a process flow of an identification function;

FIG. 3 shows a flow chart of a process flow of a registration function; and

FIG. 4A shows a view of an erroneous determination when face definition information is generated only from an original vector, and FIG. 4B shows a view of a determination when the face definition information is generated from the original vector and a virtual vector.

Claims

1. A face identification device comprising: a storing unit capable of storing face definition information which defines a face of a registrant;an identifying unit which performs face identification of a target to be identified by comparing a feature value extracted from a face image of the target to be identified with the face definition information stored in the storing unit; anda registering unit which generates the face definition information of the registrant and registers the face definition information of the registrant in the storing unit,wherein the registering unit includes:a feature value extracting unit which extracts the feature value from a face image of the registrant;a feature value converting unit which converts a vector of the extracted feature value using a converter which performs coordinate conversion in a feature value space; anda face definition information generating unit which generates the face definition information of the registrant based on the extracted feature value and a feature value including the vector after conversion.

2. The face identification device of claim 1, wherein the feature value converting unit is for converting a vector of a feature value obtained from a face image of the registrant captured under a first condition to a vector of a feature value obtained from a face image of the registrant captured under a second condition that is different from the first condition.

3. The face identification device of claim 2, wherein when the vector before conversion is set to V1 and the vector after conversion is set to V2, the feature value converting unit calculates V2 according to the equation: V2=A·V1+B where A and B are coefficient matrices.

4. The face identification device of claim 3, wherein the coefficient matrices are calculated from face images of a plurality of persons using a statistical method.

5. The face identification device of claim 3, wherein the registrant is the i-th person, the vector of the feature value obtained from the face image of the registrant captured under the first condition is set to Vai, and the vector of the feature value obtained from the face image of the registrant captured under the second condition is set to Vbi, and the coefficient matrices A and B are calculated from face images of N number of persons according to the equation:

6. The face identification device of claim 3, wherein the registrant is the i-th person, the vector of the feature value obtained from the face image of the registrant captured under the first condition is set to Vai and the vector of the feature value obtained from the face image imaged under the second condition is set to Vbi, and the coefficient matrices A and B are calculated from face images of N number of persons according to the equation:

7. The face identification device of claim 1, wherein the feature value converting unit converts the vector of the extracted feature value to a plurality of vectors, and the face definition information generating unit generates the face definition information of the registrant based on the extracted feature value and a feature value including the plurality of vectors after conversion.

8. A method for registering a face of a registrant with a face identification device for performing face identification of a target by comparing a feature value extracted from a face image of the target to be identified with stored face definition information, the face registration method comprising the steps of: extracting a feature value from a face image of the registrant;converting a vector of the extracted feature value using a converter which performs coordinate conversion in a feature value space;generating face definition information of the registrant based on the extracted feature value and a feature value including the vector after conversion; andregistering the generated face definition information in a storing unit capable of storing face definition information which defines a face of a registrant.

9. The face registration method of claim 8, wherein the vector of the feature value is obtained from a face image of the registrant captured under a first condition, and the vector of the feature value is converted to a vector of a feature value obtained from a face image of the registrant captured under a second condition that is different from the first condition.

10. A computer-readable medium encoded with a computer program for registering a face of a registrant on a face identification device for performing face identification of a target by comparing a feature value extracted from a face image of the target to be identified with stored face definition information, which, when executed by an information processing device, is capable of performing the steps of: extracting a feature value from a face image of the registrant;converting a vector of the extracted feature value using a converter which performs coordinate conversion in a feature value space;generating face definition information of the registrant based on the extracted feature value and a feature value including the vector after conversion; andregistering the generated face definition information in a storing unit capable of storing face definition information which defines a face of a registrant.

11. The computer-readable medium encoded with the computer program of claim 10, wherein the vector of the feature value is obtained from a face image of the registrant captured under a first condition, and the vector of the feature value is converted to a vector of a feature value obtained from a face image of the registrant captured under a second condition that is different from the first condition.