US 12,444,072 B1
Dynamic object detection using posture data
Boshen Niu, San Mateo, CA (US); and Shreekant Gayaka, San Jose, CA (US)
Assigned to Amazon Technologies, Inc., Seattle, WA (US)
Filed by Amazon Technologies, Inc., Seattle, WA (US)
Filed on Nov. 9, 2021, as Appl. No. 17/522,557.
Int. Cl. G06T 7/70 (2017.01); G01S 17/931 (2020.01); G06T 7/20 (2017.01); G06V 20/10 (2022.01); G06V 40/10 (2022.01)
CPC G06T 7/70 (2017.01) [G06T 7/20 (2013.01); G06V 20/10 (2022.01); G06V 40/10 (2022.01); G01S 17/931 (2020.01); G06T 2207/10028 (2013.01); G06T 2207/30196 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method, the method comprising:
generating, by a device using a first sensor, first data representing a plurality of obstacles in an environment of the device;
determining, based on the first data, second data, the second data including a first cluster of points corresponding to a first object from among the plurality of obstacles;
determining, using the second data, a first location associated with the first object;
generating, using a camera of the device, first image data representing a portion of the environment;
determining that the first image data includes a representation of a first person;
determining, using the first image data, first posture data associated with the first person;
determining, using the first image data, a second location associated with the first person;
determining that the second location is within a threshold distance of the first location;
in response to determining that the second location is within the threshold distance of the first location, associating the first posture data with the first object;
determining, based on the first posture data, that the first object is mobile; and
generating, using the first data, map data indicating a subset of the plurality of obstacles, wherein the subset does not include the first object.