System for video based face recognition using an adaptive dictionary

• US 10,007,840 B2
• Filed: 12/31/2015
• Issued: 06/26/2018
• Est. Priority Date: 12/31/2014
• Status: Expired due to Fees

• Current Assignee: Alcohol Countermeasure Systems (International) Inc.
• Original Assignee: Alcohol Countermeasure Systems (International) Inc.
• Inventors: Quddus, Azhar, Nan, Xiaoming, Khan, Naimul, Guan, Ling
• Primary Examiner(s): Woldemariam, Aklilu
• Application Number:
  US14/986,344
• Publication Number:
  US 20160188964A1
• Time in Patent Office: 908 Days
• Field of Search: 382103, 382115, 382118, 382218, 382225, 382305, 382128, 382170, 382209, 382199
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06F 18/295 Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models
• CPC Class Codes: G06V 10/85 Markov-related models; Markov random fields
• CPC Class Codes: G06V 40/172 Classification, e.g. identification
• CPC Class Codes: G06V 40/40 Spoof detection, e.g. liveness detection

Abstract

The method includes a dictionary including a target collection defined by images that are known with a defined level of certainty to include a subject and an imposter collection defined by images of individuals other than the subject. In the method, images of an area are captured over a period of time. In respect of each image: a matching calculation is carried out, based upon a comparison of the image captured with the images in the dictionary to result in a measure of confidence that the subject is in the area; and an inference determination is made to replace one of the target collection images with a further image that is known with the defined level of certainty, the determination being a function of the measure of confidence resultant from the captured image, the measure resultant from one or more previously captured images and the associated capture times.

Claims

1. A method for assessing the presence of a subject in an area, the method comprising the steps of:

capturing a plurality of photographs of the area over a period of time;

maintaining, over the period of time, a dictionary, the dictionary including;

a training matrix defined by a plurality of photographs of the area that are known with a defined level of certainty to include the subject, wherein a portion of the photographs in the training matrix at the beginning of the period define a static portion of the training matrix;

and an imposter collection defined by a plurality of photographs of the area in respect of which it is known that the subject does not appear, wherein the imposter collection is defined by a plurality of photographs of individuals other than the subject;

in respect of each photograph captured, and as each photograph is captured, carrying out a matching calculation, based upon a comparison of the photograph captured with the photographs in the dictionary to result in a measure of confidence that the subject is in the area;

carrying out a calculation using the measure of confidence resultant from the captured, photograph, the measure of confidence resultant from one of the previously captured photographs and the times of capture thereof and making a determination responsive thereto based upon a statistical inference from posterior probabilities;

and selectively replacing one of the photographs in the training matrix with the one of the previously captured the photographs responsive to the determination.

2. A method according to claim 1, wherein if the measure of confidence resultant from the captured photograph is in excess of a defined threshold, replacement does not occur.

3. A method according to claim 1, wherein the one of the previously captured photographs is selected from the one or more previously captured photographs and the captured photograph based upon a calculation that compares the similarity of the captured photograph and previously captured photographs to the training matrix and the time between the respective photograph captures.

4. A method according to claim 1, wherein the static portion is a nil portion such that all photographs in the training matrix at the beginning of the period are susceptible to replacement.

5. A method according to claim 1, wherein no photograph in the training matrix that was not in the training matrix at the beginning of the period is replaced until such time as all of the photographs in the training matrix but for the static portion have been replaced.

6. A method for assessing the presence of a subject in an area over a plurality of time periods, the method comprising the steps of:

applying the method of claim 1 to each of the time periods such that each of the time periods is associated with a respective application of the method.

7. A method according to claim 6, wherein, the training matrix at commencement of each application of the method is one and the same.

8. A method according to claim 7, wherein the static portion of the training matrix is the same in each application of the method.

9. A method according to claim 1, wherein the photographs in the training matrix are initially generated in an enrolment process adapted to result in a plurality of photographs that are representative of photographs that are likely to be captured when the subject is in the area and are suitable for recognition purposes.

10. A method according to claim 9, wherein the enrolment process involves a photography session wherein the subject is photographed in a variety of poses.

11. A method according to claim 10, wherein the imposter collection is defined by a plurality of photographs of each of the individuals other than the subject.

12. A method according to claim 11, wherein the plurality of photographs of each individual other than the subject is a plurality of photographs of said each individual in the variety of poses.

13. A method according to claim 12, wherein the variety of poses is defined by head shots with angular variation up to 15°

from face-front.

14. A method according to claim 1, wherein the matching calculation is a sparse-representation based classification calculation using the sparse dictionary.

15. A method according to claim 1, wherein the calculation using the measure of confidence is an inference determination based upon a Bayesian inference scheme.