US 11,900,706 B1
Object of interest detector distance-based multi-thresholding
Gang Qian, McLean, VA (US); Allison Beach, Leesburg, VA (US); Sima Taheri, McLean, VA (US); Sravanthi Bondugula, Vienna, VA (US); Sung Chun Lee, Tysons, VA (US); and Narayanan Ramanathan, Chantilly, VA (US)
Assigned to ObjectVideo Labs, LLC, Tysons, VA (US)
Filed by ObjectVideo Labs, LLC, Tysons, VA (US)
Filed on Sep. 24, 2020, as Appl. No. 17/030,965.
Claims priority of provisional application 62/924,286, filed on Oct. 22, 2019.
Int. Cl. G06T 7/20 (2017.01); G06N 20/00 (2019.01); G06T 7/70 (2017.01); G06V 40/10 (2022.01); G06F 18/2113 (2023.01)
CPC G06V 40/10 (2022.01) [G06F 18/2113 (2023.01); G06N 20/00 (2019.01); G06T 7/20 (2013.01); G06T 7/70 (2017.01); G06T 2207/20081 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A system comprising one or more computers and one or more storage devices on which are stored instructions that are operable, when executed by the one or more computers, to cause performance of operations comprising:
detecting an object depicted in a first image frame captured by a camera with detection confidence indicated by a confidence score;
determining a predicted physical distance between the object and the camera;
selecting, using the predicted physical distance between the object and the camera and from a plurality of predetermined confidence thresholds each of which (i) are a threshold for a different range of physical distances in a plurality of ranges and (ii) define a different range of confidence scores to be satisfied for an object to be classified as an object of interest, a confidence threshold corresponding to a range of physical distances that includes the predicted physical distance;
determining, using the confidence threshold selected using the predicted physical distance and the confidence score, that the object is not likely an object of interest; and
in response to determining that the object is not likely an object of interest using the confidence threshold selected using the predicted physical distance:
determining to skip detection for an object in one or more image frames subsequent to the first image frame; and
analyzing, for detection of an object of interest, a second image frame captured subsequent to the one or more image frames for depiction of a second object.