US 11,657,400 B2
Loss prevention using video analytics
Sivaram Annadurai, Bangalore (IN); Satyajit Nair, Karnataka (IN); and Rajkiran Kumar Gottumukkal, Bangaluru (IN)
Assigned to Sensormatic Electronics, LLC, Boca Raton, FL (US)
Filed by SENSORMATIC ELECTRONICS, LLC, Boca Raton, FL (US)
Filed on Feb. 28, 2020, as Appl. No. 16/805,298.
Claims priority of provisional application 62/907,122, filed on Sep. 27, 2019.
Prior Publication US 2021/0097544 A1, Apr. 1, 2021
Int. Cl. G06Q 20/00 (2012.01); G06Q 20/40 (2012.01); G06K 7/10 (2006.01)
CPC G06Q 20/4016 (2013.01) [G06K 7/10297 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method of loss prevention in a store, comprising:
receiving a visual code identifier associated with a transaction of one or more transaction merchandises, wherein the visual code identifier indicates at least one of the one or more transaction merchandises, a number of the one or more transaction merchandises, or a price of the one or more transaction merchandises;
receiving a radio frequency identification (RFID) identifier associated with the transaction, wherein the RFID identifier indicates at least one of one or more scanned merchandises, a number of the one or more scanned merchandises, or a price of the one or more scanned merchandises;
receiving an image associated with the transaction;
analyzing the image to determine at least one of one or more imaged merchandises, a number of the one or more imaged merchandises, or a price of the one or more imaged merchandises, wherein analyzing the image comprises analyzing the image using a neural network having:
a plurality of feature layers configured to extract features of the image;
at least one classification layer configured to output a merchandise identifier label based on the extracted features;
a classification error component configured to compare the merchandise identifier label with a ground truth merchandise identifier to output a merchandise error; and
a feedback component configured to output one or more updated parameters to the plurality of feature layers based on the merchandise error;
determining at least one of one or more detected merchandises, a number of the one or more detected merchandises, and a price of the one or more detected merchandises based on information in the RFID identifier and the analyzed image;
determining at least one of a merchandise discrepancy between the one or more transaction merchandises and the one or more detected merchandises, a number discrepancy between the number of the one or more transaction merchandises and the number of the one or more detected merchandises, or a price discrepancy between the price of the one or more transaction merchandises and the price of the one or more detected merchandises; and
transmitting, in response to determining the at least one of the merchandise discrepancy, the number discrepancy, or the price discrepancy, an alert.