US 12,333,856 B1
System and method for access control using a plurality of images
Gennady Livitz, Belmont, MA (US); Patrick L. Quinlan, Brooklyn, NY (US); Yann Henon, Somerville, MA (US); Robert Banks, North Wales, PA (US); Kelly A. Bucey, Doylestown, PA (US); Robert R. Seaner, Jr., Doylestown, PA (US); Sanjay Manandhar, Cambridge, MA (US); and Samson Timoner, Cambridge, MA (US)
Assigned to WICKET, LLC, Las Vegas, NV (US)
Filed by Wicket, LLC, Cambridge, MA (US)
Filed on Nov. 28, 2023, as Appl. No. 18/521,403.
Application 18/521,403 is a continuation of application No. 18/146,420, filed on Dec. 26, 2022, granted, now 11,837,018.
Application 18/146,420 is a continuation of application No. 17/241,223, filed on Apr. 27, 2021, granted, now 11,544,965, issued on Jan. 3, 2023.
Application 17/241,223 is a continuation in part of application No. 16/679,021, filed on Nov. 8, 2019, granted, now 11,132,532, issued on Sep. 28, 2021.
Application 16/679,021 is a continuation in part of application No. 16/409,422, filed on May 10, 2019, granted, now 11,010,597, issued on May 18, 2021.
Claims priority of provisional application 63/016,957, filed on Apr. 28, 2020.
Claims priority of provisional application 62/758,268, filed on Nov. 9, 2018.
Claims priority of provisional application 62/669,698, filed on May 10, 2018.
This patent is subject to a terminal disclaimer.
Int. Cl. G06V 40/16 (2022.01); G06F 18/22 (2023.01); G06V 10/98 (2022.01); G06V 40/60 (2022.01); G07C 9/00 (2020.01)
CPC G06V 40/172 (2022.01) [G06F 18/22 (2023.01); G06V 10/993 (2022.01); G06V 40/166 (2022.01); G06V 40/67 (2022.01); G07C 9/00563 (2013.01)] 22 Claims
OG exemplary drawing
 
1. A method of face recognition, comprising
A. receiving a plurality of frames of a video stream imaging a candidate individual in a field of view of a camera,
B. generating for each of the plurality of frames one or more scores of the frame and/or of the candidate individual imaged therein,
C. selecting fewer than all of the plurality of frames for matching by a face recognizer, where such selection is based on one or more of the scores of each respective frame, where the selecting step includes selecting frames until each of M frames have a composite score of at least k1, where the composite score of a frame is a composite of the one or more scores generated for that frame in Step (B),
D. invoking the face recognizer on the selected frames to match the candidate individual imaged therein against a set of one or more designated individuals and to generate a match score indicative of a closeness of match,
E. generating an output indicative of approval or denial of the candidate individual based on comparison of said match score generated by the face recognizer with a low watermark value, and
F. adjusting
(i) k1 in real-time based on feedback rejecting candidates for which output indicating approval has been generated for access in Step (E), and
(ii) the low water mark value to improve the correctness of outputs indicative of approval generated in step (E).
 
22. A system for face recognition, comprising
A. a camera,
B. a plurality of scorers coupled to the camera and receiving therefrom a plurality of frames of a video stream imaging a candidate individual in a field of view of the camera,
C. each scorer generating for each of the plurality of frames one or more scores of the respective frame and/or of the candidate individual imaged therein,
D. a frame selector coupled to the scorers, the frame selector selecting fewer than all of the plurality of frames for matching by a face recognizer, where such selection is based on one or more scores of each respective frame, where the frame selector selects frames until each of M frames have a composite score of at least k1, where the composite score of a frame is a composite of the one or more scores generated for that frame by the scorer,
E. the face recognizer coupled to the frame selector and matching the candidate individual, as imaged in a said respective frame, against a set of one or more images of designated individuals and to generate a match score indicative of a closeness of match,
F. a controller that is coupled to the face recognizer and to the frame selector and that generates an output indicative of approval or denial of the candidate individual based on comparison of said match score generated by the face recognizer with a low watermark value, and
G. wherein the controller adjusts
(i) k1 in real-time based on feedback rejecting candidates for which output indicating approval has been generated for access in Step (E), and
(ii) the low water mark value to improve the correctness of outputs indicative of approval generated in step (E).