US 12,406,489 B2
System and method for automatic detection of visual events in transportation environments
Ilan Naslavsky, Newton, MA (US); Aditya Gupte, Cambridge, MA (US); Moran Cohen, Jerusalem (IL); and David J. Michael, Waban, MA (US)
Assigned to ShipIn Systems Inc., Newton, MA (US)
Filed by ShipIn Systems Inc., Newton, MA (US)
Filed on Jul. 25, 2022, as Appl. No. 17/873,053.
Claims priority of provisional application 63/225,652, filed on Jul. 26, 2021.
Prior Publication US 2023/0040565 A1, Feb. 9, 2023
Int. Cl. G06V 10/94 (2022.01); G06V 10/25 (2022.01); G06V 10/764 (2022.01); G06V 10/774 (2022.01); G06V 10/82 (2022.01)
CPC G06V 10/94 (2022.01) [G06V 10/25 (2022.01); G06V 10/765 (2022.01); G06V 10/774 (2022.01); G06V 10/82 (2022.01)] 25 Claims
OG exemplary drawing
 
1. A system for detecting visual events in a transportation environment having one or more locations of interest in which the events occur comprising:
a plurality of cameras arranged to image each of a plurality of activities relevant to the transportation environment, in which each camera of the plurality of cameras respectively acquires images of a location of interest and transmits image data thereof to a processor; and
a visual detector associated with the processor arranged to include,
(a) at least one visual change detector that identifies changes between at least two of the images,
(b) at least one pre-trained visual deep learning model operating on at least an individual frame of images exclusively transmitted based on the identified changes, and generating a deep learning inference output associated with at least one person or object depicted at least in part in the individual frame, and
(c) at least one rule-based classifier that produces events or alerts from the deep learning inference output associated with the at least one person or object, and that is trained with at least one rule facilitating safe, rule-conforming operation of at least one of crew behavior, crew safety, ship maintenance, ship environment, or active cargo monitoring.