SYSTEM AND METHOD FOR PERFORMING RAPID FACIAL RECOGNITION

• US 20100239130A1
• Filed: 09/21/2009
• Published: 09/23/2010
• Est. Priority Date: 03/18/2009
• Status: Active Grant

• Current Assignee: Industrial Technology Research Institute
• Original Assignee: Industrial Technology Research Institute
• Inventors: Chang, Leii H., Luo, Yi-Fei, Chen, Jian-Ren
• Granted Patent: US 8,244,002 B2
• Time in Patent Office: Days
• US Class Current: 382/118
• CPC Class Codes: G06V 10/95 structured as a network, e.g. client-server architectures
• CPC Class Codes: G06V 40/172 Classification, e.g. identification

Abstract

A system and a method for performing rapid facial recognition are provided. The rapid facial recognition system includes an image capture device, a broadcasting feature computing unit and several response recognition computing units scattered on the network. Each of the broadcasting feature computing unit and the response recognition computing units includes a feature recognition module, wherein the broadcasting feature computing unit further includes a feature extraction module and an identification module. The image capture device captures a facial image, and the feature extraction module extracts features of the facial image to generate a set of feature data that is broadcasted to the response recognition computing units. The feature recognition modules, in accordance with the set of the feature data and their classes allocated, perform distributed facial recognition for generating recognition results as a response. The identification module identifies the recognition results to accomplish the recognition of an individual'"'"'s identity.

Claims

1. A rapid facial recognition system, for performing recognition according to a plurality of classes, comprising:

at least an image capture device used for capturing a facial image;

a plurality of response recognition computing units, each of which further comprising;

a second feature recognition module having a part of the plurality of classes performs rapid facial recognition in accordance with a set of feature data and the part of the plurality of classes allocated to the second feature recognition module for generating a second recognition result as a response; and

at least a broadcasting feature computing unit connected to the plurality of the response recognition computing units via the network, further comprising;

a first feature recognition module having the other part of the plurality of classes, performs rapid facial recognition in accordance with the set of feature data and the other part of the plurality of classes allocated to the first feature recognition module for generating a first recognition result as a response;

a first feature extraction module used for extracting features of the facial image to generate the set of feature data and further broadcast the set of feature data to the first feature recognition module and the plurality of second feature recognition modules; and

a first identification module used for identifying the first recognition result and the plurality of second recognition results to accomplish the recognition of an individual'"'"'s identity.

2. The rapid facial recognition system according to claim 1, wherein the fact that when the second feature recognition module categorizes the set of feature data to a specific class of the plurality of classes, implies that the second feature recognition module is successful in face recognition, and the fact that when the first feature recognition module categorizes the set of feature data to a specific class of the plurality of classes, implies that the first feature recognition module is successful in face recognition.

3. The rapid facial recognition system according to claim 2, wherein when the second feature recognition module is successful in face recognition, the second feature recognition module not only generates the second recognition result, but also sends a set of previous space-time recognition data of the specific class as a response, and when the first feature recognition module is successful in face recognition, the first feature recognition module not only generates the first recognition result, but also sends a set of previous space-time recognition data of the specific class as a response.

4. The rapid facial recognition system according to claim 3, wherein the set of previous space-time recognition data is the coordinate data of time and space related to a previous recognition result of the specific class.

5. The rapid facial recognition system according to claim 3, wherein the first identification module further performs recognition in accordance with the set of previous space-time recognition data of the specific class, the current first recognition result and the plurality of the second recognition results.

6. The rapid facial recognition system according to claim 2, wherein the first identification module is further connected to a set of attendance data, which comprises an attendance record of the specific class, and the first identification module further recognizes the first recognition result and the plurality of second recognition results in accordance with the attendance record related to the specific class.

7. The rapid facial recognition system according to claim 1, wherein the plurality of classes are distributed by the roughly average method to the first feature recognition module and the plurality of second feature recognition modules.

8. The rapid facial recognition system according to claim 2, wherein when the second feature recognition module is successful in face recognition and when there is a need to perform the action of class exchange, the second feature recognition module performs exchange between the specific class and a class of the non-specific classes of the first feature recognition module.

9. The rapid facial recognition system according to claim 8, wherein the class of the non-specific classes of the first feature recognition module that is perfomed exchang with the specific class of the second feature recognition module, is the class whose previous camera position is the farthest away from the current camera position among all non-specific classes of the first feature recognition module.

10. The rapid facial recognition system according to claim 1, wherein when the plurality of classes need to be reallocated, the rapid facial recognition system adopts the least total distance method to provide the first feature recognition module of the broadcasting feature computing unit and the plurality of second feature recognition modules of the response recognition computing units, either first feature recognition module and one of the second feature recognition modules, or any two of the second feature recognition modules for performing class reallocation.

11. The rapid facial recognition system according to claim 1, further comprising another first feature recognition module, wherein when the plurality of classes need to be reallocated, the rapid facial recognition system adopts the least total distance method to provide the first feature recognition module and the another first feature recognition module for performing class reallocation.

12. The rapid facial recognition system according to claim 1, wherein the plurality of response recognition computing units are divided into a plurality of groups in accordance with the distance between the plurality of response recognition computing units and the broadcasting feature computing unit;

and the broadcasting feature computing unit broadcasts the set of feature data to the nearest group first, and if the recognition fails, then the broadcasting feature computing unit broadcasts the set of feature data to the next nearest group until the second feature recognition module of one of the plurality of response recognition computing units is successful in face recognition, or the first feature extraction module has finished broadcasting to all the plurality of groups.

13. The rapid facial recognition system according to claim 1, further comprising another image capture device for capturing another facial image, wherein each of the plurality of response recognition computing units further comprises a second feature extraction module for extracting features of the another facial image to generate and broadcast another set of feature data;

the first feature recognition module performs rapid facial recognition in accordance with the another set of feature data and the other part of the plurality of classes allocated to the first feature recognition module for generating a third recognition result as a response; and

each of the second feature recognition module performs rapid facial recognition in accordance with the another set of feature data and the part of the plurality of classes allocated to each second feature recognition module for generating a fourth recognition result as a response.

14. The rapid facial recognition system according to claim 1, wherein each of the plurality of response recognition computing units comprises:

a second identification module used for identifying the third recognition result and the plurality of fourth recognition results to accomplish the recognition of an individual'"'"'s identity.

15. A rapid facial recognition method for performing recognition in accordance with a plurality of classes, comprising:

dividing a plurality of computing units connected via the network into at least a broadcasting feature computing unit and a plurality of response recognition computing units;

capturing a facial image by the broadcasting feature computing unit;

extracting features of the facial image to generate a set of feature data by the broadcasting feature computing unit;

broadcasting the set of feature data, by the broadcasting feature computing unit, to the feature recognition modules of the broadcasting feature computing unit, and to the plurality of feature recognition modules of the plurality of response recognition computing units;

performing rapid facial recognition by the plurality of feature recognition modules in accordance with the set of feature data and the plurality of classes allocated to the plurality of feature recognition modules for generating a plurality of recognition results as a response; and

further sending a set of previous space-time recognition data related to a specific class when face recognition is successfully accomplished; and

accomplishing the recognition of an individual'"'"'s identity by the broadcasting feature computing unit in accordance with the recognition results and the set of previous space-time recognition data of the specific class.

16. The rapid facial recognition method according to claim 15, wherein the steps of performing rapid facial recognition in accordance with the set of feature data and the plurality of classes allocated to the plurality of feature recognition modules further comprises:

determining whether the set of feature data is categorized in the specific class of the plurality of classes; and

implying that the action succeeds in facial recognition when one of the feature recognition modules categorizes the set of feature data as the specific class.

17. The rapid facial recognition method according to claim 15, wherein the step of accomplishing the recognition of an individual'"'"'s identity in accordance with the recognition results and the set of previous space-time recognition data of the specific class further comprises:

connecting a set of attendance data comprising an attendance record of the specific class, and recognizing the plurality of recognition results in accordance with the attendance record of the specific class.

18. The rapid facial recognition method according to claim 15, further comprising:

allocating the plurality of classes to the feature recognition modules by the roughly average method.

19. The rapid facial recognition method according to claim 16, wherein when the action succeeds in facial recognition, the method further comprises:

determining if it is necessary to exchange the plurality of classes allocated to the feature recognition modules; and

if it is, performing exchange action between the specific class and a class of the non-specific classes of the feature recognition module of the broadcasting feature computing unit, by the feature recognition modules of the response recognition computing units.

20. The rapid facial recognition method according to claim 19, wherein the exchanged class of the non-specific classes is the class whose previous camera position is the farthest away from the current camera position among all the non-specific classes of the feature recognition module of the broadcasting feature computing unit.

21. The rapid facial recognition method according to claim 15, further comprising:

determining if it is necessary to reallocate the plurality of classes allocated to the feature recognition modules; and

if it is, providing any two of the feature recognition modules by adopting the least total distance method for performing class reallocation.

22. The rapid facial recognition method according to claim 15, wherein the step of broadcasting the set of feature data to the broadcasting feature computing unit and a plurality of feature recognition modules of the plurality of response recognition computing units further comprises:

dividing the plurality of response recognition computing units into a plurality of groups in accordance with the distance between the plurality of response recognition computing units and the broadcasting feature computing unit;

broadcasting the set of feature data to the nearest group by the broadcasting feature computing unit; and

if recognition fails, broadcasting the set of feature data to the next nearest group until the second feature recognition module of one of the response recognition computing units is successful in face recognition or the first feature extraction module has finished broadcasting to all the plurality of groups.