US 12,002,336 B2
Methods and systems for detecting and tracking objects
Chithrai Selvakumar Mani, Richardson, TX (US); and Venkat Shiva Pandiri, Irving, TX (US)
Assigned to DIGIT7 INDIA PRIVATE LIMITED, Richardson, TX (US)
Filed by DIGIT7 INDIA PRIVATE LIMITED, Richardson, TX (US)
Filed on Jun. 10, 2021, as Appl. No. 17/303,937.
Prior Publication US 2022/0398903 A1, Dec. 15, 2022
Int. Cl. G07G 1/00 (2006.01); A47F 9/04 (2006.01); G06K 17/00 (2006.01); G06Q 20/20 (2012.01)
CPC G07G 1/0036 (2013.01) [A47F 9/048 (2013.01); G06K 17/00 (2013.01); G06Q 20/20 (2013.01)] 12 Claims
OG exemplary drawing
 
1. A method for detecting and tracking objects in a physical store, the method comprising:
detecting, by a detection and tracking device, at least one object in a physical store on receiving media from a plurality of media acquisition devices positioned in the physical store, each media acquisition device covering at least three points of a plurality of points of the physical store, wherein a point depicts a portion of an entire area of the physical store, with adjacent media acquisition devices overlapping each other by at least one point; and
tracking, by the detection and tracking device, the at least one object in the physical store by projecting input data points of each of the plurality of media acquisition devices onto a groundplot by applying a calibration process and a matrix multiplication process, and clustering the input data points into a single cluster, wherein clustering of the input data points comprises:
generating a groundplot visualization by visualizing a continuous stream of output data points of each media acquisition device on the groundplot, wherein the groundplot visualization indicates data points of each object on the groundplot that are strongly connected with each other;
forming a cluster for each object by grouping the data points of each object that are strongly connected with each other using a k-means clustering method;
identifying a centroid of the cluster of each object; and
assigning a tracking identifier (ID) for the centroid of the cluster of each object for tracking each object in the physical store, wherein the centroid of the cluster corresponds to the associated object.