Face recognition method

• US 9,898,648 B2
• Filed: 02/16/2016
• Issued: 02/20/2018
• Est. Priority Date: 02/15/2016
• Status: Expired due to Fees

• Current Assignee: Electronics and Telecommunications Research Institute
• Original Assignee: Electronics and Telecommunications Research Institute
• Inventors: Oh, Sang Yoon
• Primary Examiner(s): AHMED, SAMIR ANWAR
• Application Number:
  US15/045,023
• Publication Number:
  US 20170236002A1
• Time in Patent Office: 735 Days
• CPC Class Codes: G06F 18/22 Matching criteria, e.g. proximity measures
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships
• CPC Class Codes: G06V 40/172 Classification, e.g. identification

Abstract

A face recognition method of the present disclosure includes configuring aggregations of feature data that include a plurality of feature data of faces and match to a plurality of personnel data; extracting from an input image a plurality of input feature data that correspond to the feature data and that is equal to or more than a critical value; comparing an aggregation of input feature data that includes the input feature data with each of the pre-stored aggregations of feature data, and selecting from the aggregations of feature data an aggregation of feature data having the greatest similarity with the aggregation of the input feature data; and identifying a person on the image based on personnel data that matches the aggregation of feature data having the greatest similarity.

Claims

1. A face recognition method comprising:

storing a plurality of aggregations of sample feature data matched with corresponding personnel data, each of the plurality of aggregations of sample feature data including a ratio of a plurality of widths of feature areas of a sample face;

receiving, from an imager, an input image of a person;

extracting, from the input image, input feature data that are each equal to or more than a critical value;

comparing an aggregation of the input feature data with each of the plurality of stored aggregations;

selecting an aggregation among the plurality of stored aggregations having the greatest similarity with the aggregation of the input feature data; and

identifying the person based on the personnel data matched with the selected aggregation.

2. The method of claim 1,

wherein extracting the plurality of input feature data includes setting a value of the input feature data that corresponds to an unrecognizable feature or that is less than the critical value to zero (0).

3. The method of claim 2,

wherein selecting an aggregation of feature data having the greatest similarity comprises:

computing a plurality of values of comparison between the input feature data and the sample feature data for each of the plurality of stored aggregations;

computing a sum of the computed plurality of values of comparison for each of the plurality of stored aggregations; and

selecting an aggregation corresponding to the smallest sum of values of comparison as the aggregation having the greatest similarity.

4. The method of claim 3,

wherein computing the values of comparison includes setting each of the values of comparison corresponding to an unrecognizable feature, and the input feature data corresponding to the unrecognizable feature, to zero (0).

5. The method of claim 3, further comprising determining whether or not the sum of the values of comparison for each of the plurality of stored aggregations is below a reference value.

6. The method of claim 5,

wherein identifying the person includes, when the sum of the values of comparison of the selected aggregation is equal to or above the reference value, determining that there is no personnel data corresponding to the person on the image.

7. The method of claim 1,

wherein the plurality of aggregations of sample feature data are stored in a feature data database (DB), the face of each of the plurality of aggregations being of a pre-identified person.

8. The method of claim 1,

wherein the plurality of types of sample feature data includes a distance between feature points.

9. The method of claim 8,

wherein the distance between the feature points or each of the plurality of the widths of feature areas is calculated in pixel units.

10. The method of claim 8,

wherein extracting the input feature data involves recognizing a face outline or a feature area of a face from the input image.

11. The method of claim 10,

wherein the plurality of types of input feature data comprises:

first feature data that is a ratio of a width of a right eye and a first face width;

second feature data that is a ratio of a width of a nose and a second face width;

third feature data that is a ratio of a width of a left eye and the width of the nose;

fourth feature that is a ratio of a width of a mouth and a third face width;

fifth feature that is a ratio of the width of the right eye and the width of the mouth; and

sixth feature data that is a ratio of the width of the nose and the width of the mouth.

12. The method of claim 11,

wherein the width of the right eye is defined as a maximum width of a right eye area in a width direction.

13. The method of claim 11,

wherein the width of the left eye is defined as a maximum width of a left eye area in a width direction.

14. The method of claim 12,

wherein the first face width is defined as a distance between the points at which a horizontal line from a central point of the right eye area and a central point of the left eye area intersects with the face outline.

15. The method of claim 11,

wherein the width of the nose is defined as a maximum width of a nose area in a width direction.

16. The method of claim 15,

wherein the second face width is defined as a distance between the points at which a horizontal line extended from the maximum width of the nose area in the width direction intersects with the face outline.

17. The method of claim 11,

wherein the width of the mouth is defined as a maximum width of a mouth area in a width direction.

18. The method of claim 17,

wherein the third face width is defined as a distance between the points at which a horizontal line extended from the maximum width of the mouth area in the width direction intersects with the face outline.

19. A face recognition method comprising:

pre-storing aggregations of feature data that include a plurality of feature data of faces and match to a plurality of personnel data;

receiving, from an imager, an input image of a person;

extracting, from the input image, a plurality of input feature data that correspond to the feature data and that is equal to or more than a critical value;

comparing an aggregation of input feature data that includes the input feature data with each of the pre-stored aggregations of feature data, and selecting from the aggregations of feature data an aggregation of feature data having the greatest similarity with the aggregation of the input feature data; and

identifying the person based on personnel data that matches the aggregation of feature data having the greatest similarity,

wherein the feature data comprises;

first feature data that is a ratio of a width of a right eye and a first face width;

second feature data that is a ratio of a width of a nose and a second face width;

third feature data that is a ratio of a width of a left eye and the width of the nose;

fourth feature that is a ratio of a width of a mouth and a third face width;

fifth feature that is a ratio of the width of the right eye and the width of the mouth; and

sixth feature data that is a ratio of the width of the nose and the width of the mouth.

20. The method of claim 19,

wherein the width of the right eye is defined as a maximum width of a right eye area in a width direction.