US 9,811,537 B2
Product identification via image analysis
Amir Geva, Yoneam (IL); Mattias Marder, Haifa (IL); Yoav Nahshon, Tsorit (IL); and Amnon Ribak, Misgav (IL)
Assigned to International Business Machines Corporation, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on May 17, 2016, as Appl. No. 15/156,357.
Application 15/156,357 is a continuation of application No. 14/582,225, filed on Dec. 24, 2014, granted, now 9,442,958.
Prior Publication US 2016/0260057 A1, Sep. 8, 2016
This patent is subject to a terminal disclaimer.
Int. Cl. G06K 9/62 (2006.01); G06F 17/30 (2006.01); G06Q 10/08 (2012.01); G06K 9/46 (2006.01); H04N 5/225 (2006.01)
CPC G06F 17/30256 (2013.01) [G06F 17/30247 (2013.01); G06K 9/4604 (2013.01); G06K 9/6202 (2013.01); G06K 9/6215 (2013.01); G06Q 10/087 (2013.01); H04N 5/2257 (2013.01)] 15 Claims
OG exemplary drawing
 
1. A method, comprising:
capturing an image of a shelf and products on the shelf, in a merchandising facility;
identifying, in the captured image, a shelf, a price tag affixed to the shelf, a first group of one or more first products resting on the shelf and positioned on a first side of the price tag, and a second group of one or more second products resting on the shelf and positioned on a second side of the price tag;
identifying one or more first candidates for the one or more first products and one or more second candidates for the one or more second products;
receiving information indicating expected locations for the first and the second candidates;
determining, based on the expected locations and the one or more first candidates, an identity of the one or more first products;
determining, based on the expected locations and the one or more second candidates, an identity of the one or more second products; and
managing the merchandising facility responsive to the determined identities of the first and second products,
wherein identifying the one or more second candidates comprises extracting a sub-image comprising a given product from the second group, calculating a similarity value between the sub-image and a plurality of product images having respective product identifiers, and selecting the product identifiers whose respective similarity value is greater than or equal to a predefined threshold.