Abstract: A method for detecting moving objects within a certain zone that includes initially providing multiple physical cameras around the zone and composing a panoramic video stream using the cameras. Upon composing the panoramic video stream, one or more virtual camera views are defined by selecting corresponding partial views of the panoramic views and by de-warping the selected, partial views into square views. Each de-warped view is fed into a deep learning neural network and then a corresponding part of the zone is detected.
Type:
Grant
Filed:
June 22, 2021
Date of Patent:
November 25, 2025
Assignee:
MOBILE VIEWPOINT B.V.
Inventors:
Michael Stuurman, Michel Alexander Bais
Abstract: A method for detecting moving objects within a certain zone that includes initially providing multiple physical cameras around the zone and composing a panoramic video stream using the cameras. Upon composing the panoramic video stream, one or more virtual camera views are defined by selecting corresponding partial views of the panoramic views and by de-warping the selected, partial views into square views. Each de-warped view is fed into a deep learning neural network and then a corresponding part of the zone is detected.
Type:
Application
Filed:
June 22, 2021
Publication date:
August 10, 2023
Applicant:
MOBILE VIEWPOINT B.V.
Inventors:
Michael Stuurman, Michel Alexander BAIS