Method and apparatus for foreground geometry and topology based face anti-spoofing
View Patent ↗A method and system to detect visual spoofing of a process of authenticating a person's identity employs computer vision techniques to define characteristics of different kinds of spoofing. Embodiments identify a foreground object within an image and by examining positions and/or orientations of that foreground object within the image, determine whether the presentation of the foreground object is an attempt to spoof the authentication process.
1. A method comprising:
obtaining an image to be analyzed, the image comprising one of a still image or a video image;
identifying a boundary of the obtained image;
identifying a foreground object from the obtained image using one or more computer vision techniques;
performing one or more analyses on the foreground object using a computing system, the analyses selected from the group consisting of foreground contour analysis, facial landmark orientation analysis, and skeletal analysis; and
responsive to the performing, determining whether the foreground object represents an attempt to spoof a process to authenticate an individual as an authorized individual.
2. A method according to claim 1 , further comprising identifying a facial region of interest within the foreground object using one or more computer vision techniques.
3. A method according to claim 1 , wherein the identifying the foreground object comprises comparing a predetermined number of video frames to distinguish stationary portions within the frames from moving portions within the frames.
4. A method according to claim 1 , wherein the identifying the facial region of interest comprises identifying facial landmarks within the foreground object.
5. A method according to claim 4 , further comprising performing a homography regression on results of identifying the facial landmarks.
6. A method according to claim 1 , wherein the performing comprises performing the foreground contour analysis followed by the facial landmark orientation analysis.
7. A method according to claim 1 , wherein the performing comprises performing the facial landmark orientation analysis, the performing the facial landmark orientation analysis comprising computing a transformation matrix on the foreground object.
8. A method according to claim 3 , wherein the comparing comprises performing an orientation analysis on the foreground object after the performing the one or more analyses on the foreground object.
9. A method according to claim 8 , wherein performing the orientation analysis comprises:
identifying a first plurality of points in the foreground object when the foreground object is in a first position within the boundary, and a second plurality of points in the foreground object when the foreground object is at a second position within the boundary; and
responsive to a determination that positions of the first plurality of points and the second plurality points correspond to a rigid transformation, determining the foreground object to be a spoof.
10. A method according to claim 1 , wherein the performing comprises performing foreground contour analysis, the performing foreground contour analysis comprising determining a contour of the foreground object using one or more computer vision techniques and measuring a distance of the facial region of interest from the contour,
the method further comprising, responsive to determining that the distance of the facial region of interest from the contour exceeds a predetermined amount, determining the foreground object to be a spoof.
11. A method according to claim 1 , wherein the performing comprises performing skeletal analysis, the performing skeletal analysis comprising analyzing, using computer vision techniques, the foreground object and, responsive to a determination that the foreground object represents a person, deriving a position of one or more skeletal features within the foreground object.
12. A method according to claim 11 , further comprising:
measuring a distance of the one or more skeletal features from the boundary of the obtained image, and
responsive to a determination that the measured distance is non-zero, determining the foreground object to be a spoof.
13. A computer-implemented system to perform a method comprising:
obtaining an image to be analyzed, the image comprising one of a still image or a video image;
identifying a boundary of the obtained image;
identifying a foreground object from the obtained image using one or more computer vision techniques;
performing one or more analyses on the foreground object using a computing system, the analyses selected from the group consisting of foreground contour analysis, facial landmark orientation analysis, and skeletal analysis; and
responsive to the performing, determining whether the foreground object represents an attempt to spoof a process to authenticate an individual as an authorized individual.
14. A computer-implemented system according to claim 13 , wherein the identifying the foreground object comprises comparing a predetermined number of video frames to distinguish stationary portions within the frames from moving portions within the frames.
15. A computer-implemented system according to claim 13 , wherein the performing comprises performing the foreground contour analysis followed by the facial landmark orientation analysis.
16. A computer-implemented system according to claim 14 , wherein the comparing comprises performing an orientation analysis on the foreground object after the performing the one or more analyses on the foreground object.
17. A computer-implemented system according to claim 16 , wherein performing the orientation analysis comprises:
identifying a first plurality of points in the foreground object when the foreground object is in a first position within the boundary, and a second plurality of points in the foreground object when the foreground object is at a second position within the boundary; and
responsive to a determination that positions of the first plurality of points and the second plurality points correspond to a rigid transformation, determining the foreground object to be a spoof.
18. A computer-implemented system according to claim 13 , wherein performing foreground contour analysis comprises determining a contour of the foreground object using one or more computer vision techniques and measuring a distance of the facial region of interest from the contour,
the method further comprising, responsive to determining that the distance of the facial region of interest from the contour exceeds a predetermined amount, determining the foreground object to be a spoof.
19. A computer-implemented system according to claim 13 , wherein the performing comprises performing skeletal analysis, the performing skeletal analysis comprising analyzing, using computer vision techniques, the foreground object and, responsive to a determination that the foreground object represents a person, deriving a position of one or more skeletal features within the foreground object.
20. A computer-implemented system according to claim 19 , wherein the method further comprises:
measuring a distance of the one or more skeletal features from the boundary of the obtained image, and
responsive to a determination that the measured distance is non-zero, determining the foreground object to be a spoof.