Relative pose estimation of non-overlapping cameras using the motion of subjects in the camera fields of view
AssigneeDisney Enterprises, Inc.
InventorsEric Foote, G. Peter K. Carr, Yaser Sheikh, Iain Matthews
Life of this patent
Estimating the term…
Abstract
A relative pose between two cameras is determined by using input data obtained from the motion of subjects, such as pedestrians, between the fields of view of two cameras, determining trajectory information for the subjects, and computing homographies relating lines obtained from trajectories in the first image data to lines obtained from the trajectories in the second image data. The two fields of view need not overlap.
Claims20
A method for use by a system including a processor, the method comprising:
- receiving, by the processor, a first image data representing a first field of view of a first video camera;
- receiving a second image data representing a second field of view of a second video camera, wherein the second field of view is non-overlapping the first field of view;
- determining a motion representing each of at least four trajectories moving between the first field of view and the second field of view;
- and
- computing a relative pose between the first video camera and the second video camera using the motion, wherein the computing includes backprojecting a plurality of foot position points onto a ground plane, and determining if the backprojected plurality of foot position points matches the motion.
A system comprising:
- a processor configured to:
- receive a first image data representing a first field of view of a first video camera;
- receive a second image data representing a second field of view of a second video camera, wherein the second field of view is non-overlapping the first field of view;
- determine a motion representing each of at least four trajectories moving between the first field of view and the second field of view;
- and
- compute a relative pose between the first video camera and the second video camera using the motion, wherein the processor is configured to compute the relative pose by backprojecting a plurality of foot position points onto a ground plane, and determining if the backprojected plurality of foot position points matches the motion.
A method for use by a system including a processor, the method comprising:
- receiving, by the processor, a first image data representing a first field of view of a first video camera;
- receiving a second image data representing a second field of view of a second video camera, wherein the second field of view is non-overlapping the first field of view;
- determining a motion representing each of at least four trajectories moving between the first field of view and the second field of view;
- and
- computing a relative pose between the first video camera and the second video camera using the motion, wherein the computing of the relative pose comprises:
- determining a plurality of head position points in the first image data and a plurality of head position points in the second image data using the motion;
- determining a plurality of foot position points in the first image data and a plurality of foot position points in the second image data using the motion;
- determining at least eight lines in the first image data and at least eight lines in the second image data using the plurality of head position points and the plurality of foot position points;
- computing a first planar homography and a second planar homography using the at least eight lines in the first image data and the at least eight lines in the second image data;
- and
- computing the relative pose using the first planar homography and the second planar homography.
A system comprising:
- a processor configured to:
- receive a first image data representing a first field of view of a first video camera;
- receive a second image data representing a second field of view of a second video camera, wherein the second field of view is non-overlapping the first field of view;
- determine a motion representing each of at least four trajectories moving between the first field of view and the second field of view;
- and
- compute a relative pose between the first video camera and the second video camera using the motion, wherein the processor is configured to compute the relative pose by:
- determining a plurality of head position points in the first image data and a plurality of head position points in the second image data using the motion;
- determining a plurality of foot position points in the first image data and a plurality of foot position points in the second image data using the motion;
- determining at least eight lines in the first image data and at least eight lines in the second image data using the plurality of head position points and the plurality of foot position points;
- computing a first planar homography and a second planar homography using the at least eight lines in the first image data and the at least eight lines in the second image data;
- and
- computing the relative pose using the first planar homography and the second planar homography.
Dependencies are inferred from the claim text.