×

Method of real-time tracking of moving/flexible surfaces

  • US 8,693,730 B2
  • Filed: 11/15/2012
  • Issued: 04/08/2014
  • Est. Priority Date: 11/15/2011
  • Status: Active Grant
First Claim
Patent Images

1. A method on a computer for real time tracking of a moving flexible surfaces, comprising:

  • a) synchronously acquiring left and right image frames from left and right camera members of a stereo camera, of a region of interest on a moving flexible at a time K;

    b) computing, from said left and right image frames 3D point cloud data in a coordinate frame of said stereo camera at said time K;

    c) fitting a parametric surface to said 3D point cloud at said time K;

    d) extracting a location of at least three features of interest in said left image frame at said time K in 2D pixel coordinates, extracting a location of each at least three feature of interest in said right image frame at said time K in 2D pixel coordinates, storing said 2D pixel coordinates of said locations of said features of interest in said left and right image frames at said time K in a vision processor memory storage;

    e) computing, using said stored 2D pixel coordinates of said locations of said features of interest in said left and right image frames at said time K, a 3D location of said features of interest at said time K in a coordinate frame of said camera and storing said 3D locations in said vision processor memory storage;

    f) mapping said 3D location of said features of interest onto said parametric surface at time K and storing the parametric surface with mapped 3D locations in the vision processor memory storage;

    g) retrieving from the vision processor memory storage 3D locations of said features of interest from previous left and right image frames synchronously taken at a time K−

    1;

    h) computing a best fit rigid transform using said retrieved 3D locations of the features of interest at time K−

    1 and the 3D locations of the features at time K;

    i) retrieving from the vision processor memory storage a parametric surface with mapped 3D locations at time K−

    1, applying said rigid transformation to said retrieved parametric surface with mapped 3D locations at time K−

    1 to obtain an intermediate parametric surface with mapped 3D locations;

    j) matching the 3D locations of features of interest on the intermediate parametric surface at time K−

    1 to the same corresponding 3D locations on the stored parametric surface at time K and from said correspondences computing a stretching transformation; and

    k) using said rigid and stretching transformations to calculate at time K a location of any region on said moving flexible surface based on a location of the same region on said moving flexible surface at time K−

    1 stored in said vision processor memory storage.

View all claims
  • 10 Assignments
Timeline View
Assignment View
    ×
    ×