US 12,354,364 B2
Image processing for distinguishing individuals in groups
Brent Vance Zucker, Roswell, GA (US); Adam Justin Lieberman, Suwanee, GA (US); and Nathan Somavarapu, Atlanta, GA (US)
Assigned to NCR Voyix Corporation, Atlanta, GA (US)
Filed by NCR Voyix Corporation, Atlanta, GA (US)
Filed on Apr. 23, 2021, as Appl. No. 17/239,065.
Application 17/239,065 is a continuation of application No. 16/144,563, filed on Sep. 27, 2018, granted, now 11,055,539.
Prior Publication US 2021/0240998 A1, Aug. 5, 2021
Int. Cl. G06V 20/52 (2022.01); G06N 20/00 (2019.01); G06V 40/10 (2022.01); H04W 4/33 (2018.01)
CPC G06V 20/53 (2022.01) [G06N 20/00 (2019.01); G06V 40/107 (2022.01); H04W 4/33 (2018.02)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
maintaining maximum and minimum coordinates for each of a plurality of individuals depicted in an image;
utilizing a deep machine-leaning algorithm specifically configured to process the image and identify limb attributes for each individual based on the maintained coordinates by employing convolutional neural networks that perform localization with a combination of classification and regression to determined four coordinates of each individual in a group from the image;
determining poses for each individual by analyzing the limb attributes and assigning pose classifications;
identifying individual attributes for the individuals of the group from the image;
applying metadata to track movements of the individuals within the image and to subsequent images capturing the group;
associating items detected within the image to specific individuals based on the metadata and proximity analysis; and
updating the metadata in real-time as the individuals interact with the items and each other within the group;
wherein the metadata comprises coordinates within the image for identifying each individual of the group and corresponding limbs of each individual.