Frictionless access control system embodying satellite cameras for facial recognition
A frictionless access control system and method embodying satellite cameras for facial recognition are disclosed. The cameras capture image data of individuals at an access point such as a door. Preferably, two or more cameras are placed on opposite sides of the access point to increase the likelihood that the individuals are captured in the image data. A facial cropper module extracts facial patches from the image data, and a facial signature module computes facial signatures from the facial patches. A facial recognition module receives the computed facial signatures from the facial signature module, matches the computed facial signatures to stored facial signatures, and sends user identity information of individuals corresponding to the stored facial signatures to the facial signature module when the computed facial signatures match the stored facial signatures.
1. An access control system, comprising:
one or more surveillance cameras at an access point that capture image data;
a first device comprising:
a first memory storing instructions thereon; and
at least one first processor coupled with the first memory and configured by the first instructions to:
extract facial patches from the image data;
rank the facial patches above a predefined threshold by comparing a plurality of machine learning trained image quality factors to a plurality of detected features of a candidate facial patch of the facial patches to determine acceptable facial patches for individuals; and
compute facial signatures from the acceptable facial patches; and
a second device comprising:
a second memory storing second instructions thereon; and
at least one second processor coupled with the second memory and configured by the second instructions to:
receive the facial signatures from the first device;
match the facial signatures to stored facial signatures; and
send user identity information of individuals corresponding to the stored facial signatures to the first device when the facial signatures match the stored facial signatures.
2. The access control system of claim 1 , wherein the at least one first processor is further configured by the first instructions to determine a highest ranked acceptable facial patch for each of the individuals, and compute a facial signature for each of the individuals from the highest ranked acceptable facial patch for each of the individuals.
3. The access control system of claim 1 , wherein the at least one first processor is further configured by the first instructions to compare each of the facial patches against one another to determine whether the facial patches are associated with same individuals or different individuals.
4. The access control system of claim 1 , wherein the one or more surveillance cameras include the first device.
5. The access control system of claim 1 , wherein the first device is located at the access point.
6. The access control system of claim 1 , wherein the second device is located at the access point.
7. The access control system of claim 6 , wherein the second device comprises a cache of the stored facial signatures, and the at least one second processor is further configured by the second instructions to match the facial signatures to respective ones of the cache of the stored facial signatures.
8. The access control system of claim 1 , wherein the second device is remote to the access point.
9. The access control system of claim 1 , wherein the second device includes a facial recognition database that includes the stored facial signatures and the user identity information of individuals corresponding to the stored facial signatures.
10. The access control system of claim 1 , wherein the at least one second processor is further configured by the second instructions to determine whether the individuals are authorized based on the user identity information.
11. The access control system of claim 10 , further comprising:
a door lock system at the access point, wherein the at least one first processor is further configured by the first instructions to send a signal to unlock the door lock system when the second device determines that the individuals are authorized users.
12. A method for controlling access to an access control system, comprising:
capturing image data by one or more surveillance cameras at an access point;
extracting facial patches from the image data;
ranking the facial patches above a predefined threshold by comparing a plurality of machine learning trained image quality factors to a plurality of detected features of a candidate facial patch of the facial patches to determine acceptable facial patches for individuals;
computing facial signatures from the acceptable facial patches;
matching the facial signatures to stored facial signatures; and
using user identity information of individuals corresponding to the stored facial signatures when the facial signatures match the stored facial signatures.
13. The method of claim 12 , further comprising determining a highest ranked acceptable facial patch for each of the individuals, and computing a facial signature for each of the individuals from the highest ranked acceptable facial patch for each of the individuals.
14. The method of claim 12 , further comprising comparing each of the facial patches against one another to determine whether the facial patches are associated with same individuals or different individuals.
15. The method of claim 12 , wherein extracting the facial patches from the image data comprises extracting, by the one or more surveillance cameras, the facial patches from the image data, and wherein computing facial signatures from the acceptable facial patches comprises computing, by the one or more surveillance cameras, the facial signatures from the acceptable facial patches.
16. The method of claim 12 , wherein extracting the facial patches from the image data comprises extracting, by a device at the access point, the facial patches from the image data, and wherein matching the facial signatures to the stored facial signatures comprises matching by the device at the access point, the facial signatures to the stored facial signatures.
17. The method of claim 12 , wherein matching the facial signatures to the stored facial signatures comprises matching, by a device at the access point, the facial signatures to the stored facial signatures.
18. The method of claim 12 , wherein matching the facial signatures to the stored facial signatures comprises matching, by a device remote to the access point, the facial signatures to the stored facial signatures.
19. The method of claim 12 , wherein a connected services system includes a facial recognition database that includes the stored facial signatures and user identity information of individuals corresponding to the stored facial signatures.
20. The method of claim 19 , further comprising:
determining whether the individuals are authorized based on the user identity information;
and transmitting, to a door lock system, a signal to unlock the door lock system when a facial recognition module determines that the individuals are authorized users.
21. A non-transitory computer-readable medium having instructions thereon that, when executed by at least one computing device, causes the at least one computing device to perform operations comprising:
capturing image data by one or more surveillance cameras at an access point;
extracting facial patches from the image data;
ranking the facial patches above a predefined threshold by comparing a plurality of machine learning trained image quality factors to a plurality of detected features of a candidate facial patch of the facial patches to determine acceptable facial patches for individuals;
computing facial signatures from the acceptable facial patches;
matching the facial signatures to stored facial signatures; and
using user identity information of individuals corresponding to the stored facial signatures when the facial signatures match the stored facial signatures.