Back to results
US 10,706,561 B2Granted
Systems and methods to track objects in video
Assignee612 Authentic Media
InventorGabriel Cheifetz
Life of this patent
Estimating the term…
Abstract
Systems and methods to track objects in video are disclosed. Exemplary implementations may obtain a user entry and/or selection of a depiction of an object visible within a first frame of a video and process the depiction of the object for use as a current reference. For subsequent frames, the systems and methods may determine whether the depiction of the object is visible. If the depiction of the object is no longer visible within a particular frame, image features of subsequent frames are compared with the latest reference of the depiction of the object until the depiction of the object is visible again. If and when the object is visible, an object indicator may be displayed on the relevant frames and/or otherwise presented to a viewer.
Claims18
A system configured to track objects in video, the system comprising:
- one or more hardware processors configured by machine-readable instructions to:
- obtain view information for a video comprising multiple consecutive frames, the view information defining visual content, the visual content including depictions of objects and/or surfaces, the view information defining the visual content in the form of images, an individual image forming an individual frame of the video;
- obtain a user entry and/or selection of a depiction of an object visible within a first frame;
- identify a pattern of motion of the depiction of the object between a first frame and a second frame that appears subsequent to the first frame to track a location of the depiction of the object from the first frame to the second frame, the second frame being adjacent to the first frame;
- in response to the depiction of the object within the second frame being viable as a reference, process the depiction of the object for use as a reference;
- track the depiction of the object through one or more frames between the second frame and a third frame that appears subsequent to the second frame, wherein such tracking includes:
- in response to the depiction of the object being visibly present in the one or more frames between the second frame and the third frame and not being viable as a reference in the one or more frames between the second frame and the third frame, track the depiction of the object through the one or more frames between the second frame and the third frame is performed using the reference from the second frame, and
- in response to the depiction of the object being viable as a reference in one of the frames in the one or more frames between the second frame and the third frame, process the depiction of the object that is viable as a reference for use as an updated reference;
- determine whether the depiction of the object is no longer visible within the third frame using the updated reference;
- in response to determining the depiction of the object is no longer visible within the third frame, extract image features from a fourth frame that appears subsequent to the third frame;
- perform a comparison of the image features of the fourth frame with the updated reference;
- determine, based on the comparison, whether the depiction of the object is visibly present in the fourth frame;
- determine whether the depiction of the object within the fourth frame is viable as a reference;
- in response to the depiction of the object within the fourth frame being viable as a reference, process the depiction of the object for use as a further updated reference;
- in further response to determining the depiction of the object is no longer visible within the third frame using the updated reference,
- extract image features from the third frame;
- and
- perform a comparison of the image features from the third frame and the further updated reference to further attempt to identify the depiction of the object in the third frame;
- track the depiction of the object in one or more frames subsequent to the fourth frame using the further updated reference in response to the depiction of the object being visible in the one or more frames subsequent to the fourth frame but not being viable as a reference;
- and
- in response to the depiction of the object in one of the one or more frames subsequent to the fourth frame being viable as a reference, process such depiction of the object from the one of the one or more frames subsequent to the fourth frame as a still further updated reference.
18 more elements
A computer-implemented method to track objects in video, the method comprising:
- obtaining view information for a video comprising multiple consecutive frames, the view information defining visual content, the visual content including depictions of objects and/or surfaces, the view information defining the visual content in the form of images, an individual image forming an individual frame of the video;
- obtaining a user entry and/or selection of a depiction of an object visible within a first frame;
- identifying a pattern of motion of the depiction of the object between a first frame and a second frame that appears subsequent to the first frame to track a location of the depiction of the object from the first frame to the second frame, the second frame being adjacent to the first frame;
- in response to the depiction of the object within the second frame being viable as a reference, processing the depiction of the object for use as a reference;
- tracking the depiction of the object through one or more frames between the second frame and a third frame that appears subsequent to the second frame, wherein such tracking includes:
- in response to the depiction of the object being visibly present in the one or more frames between the second frame and the third frame and not being viable as a reference in the one or more frames between the second frame and the third frame, tracking the depiction of the object through the one or more frames between the second frame and the third frame is performed using the reference from the second frame, and
- in response to the depiction of the object being viable as a reference in one of the frames in the one or more frames between the second frame and the third frame, processing the depiction of the object that is viable as a reference for use as an updated reference;
- determining whether the depiction of the object is no longer visible within the third frame using the updated reference;
- in response to determining the depiction of the object is no longer visible within the third frame, extracting image features from a fourth frame that appears subsequent to the third frame;
- performing a comparison of the image features of the fourth frame with the updated reference;
- determining, based on the comparison, whether the depiction of the object is visibly present in the fourth frame;
- determining whether the depiction of the object within the fourth frame is viable as a reference;
- in response to the depiction of the object within the fourth frame being viable as a reference, processing the depiction of the object for use as a further updated reference;
- in further response to determining the depiction of the object is no longer visible within the third frame using the updated reference,
- extracting image features from the third frame;
- and
- performing a comparison of the image features from the third frame and the further updated reference to further attempt to identify the depiction of the object in the third frame;
- tracking the depiction of the object in one or more frames subsequent to the fourth frame using the further updated reference in response to the depiction of the object being visible in the one or more frames subsequent to the fourth frame but not being viable as a reference;
- and
- in response to the depiction of the object in one of the one or more frames subsequent to the fourth frame being viable as a reference, processing such depiction of the object from the one of the one or more frames subsequent to the fourth frame as a still further updated reference.
17 more elements
Dependencies are inferred from the claim text.