US 12,437,284 B2
Detection system, detection method, and storage medium
Hiroki Sugegaya, Tokyo (JP); Shunsuke Tsuda, Tokyo (JP); Ning Li, Tokyo (JP); Yuki Tsuruoka, Tokyo (JP); and Shuta Okada, Tokyo (JP)
Assigned to NEC CORPORATION, Tokyo (JP)
Appl. No. 18/267,634
Filed by NEC Corporation, Tokyo (JP)
PCT Filed Dec. 20, 2021, PCT No. PCT/JP2021/046952
§ 371(c)(1), (2) Date Jun. 15, 2023,
PCT Pub. No. WO2022/209055, PCT Pub. Date Oct. 6, 2022.
Claims priority of application No. 2021-056682 (JP), filed on Mar. 30, 2021.
Prior Publication US 2024/0054470 A1, Feb. 15, 2024
Int. Cl. G06Q 20/20 (2012.01); G06T 7/20 (2017.01); G06V 40/20 (2022.01)
CPC G06Q 20/208 (2013.01) [G06T 7/20 (2013.01); G06V 40/28 (2022.01); G06T 2207/30242 (2013.01)] 6 Claims
OG exemplary drawing
 
1. A detection system comprising:
at least one memory configured to store instructions; and
at least one processor configured to execute the instructions to:
acquire a number of one or more products to be registered based on an image in which a customer reads one or more product symbols with a scanner;
acquire a number of one or more products registered by the customer;
detect discrepancy between the number of the one or more products to be registered and the number of the one or more registered products; and
output a detection result comprising a message that encourages the customer to start over a product registration or a message that encourages the customer to confirm a content of the product registration,
wherein the number of the one or more products to be registered is a number of times of a product scanning motion that the customer reads the one or more product symbols with the scanner,
wherein the at least one processor is further configured to execute the instructions to:
detect, by analyzing the image using a posture estimation model, the product scanning motion that the customer reads the one or more product symbols with the scanner, wherein analyzing the image comprises:
recognize human joint points from the image using a model trained to recognize human joint points from images;
detect a right arm frame and a left arm frame of the customer from the human joint points;
acquire a number of times of the product scanning motion by the customer based on movements of the right arm frame and the left arm frame as the number of one or more products to be registered;
detect the number of one or more products before product registration based on an image of the one or more products before the product registration;
correct the number of times of the product scanning motion by the customer based on the detected number of one or more products detected from the images before the product registration; and
detect a discrepancy between the corrected number and the number of registered products.