×

System and method for deformable object recognition

  • US 8,260,059 B2
  • Filed: 05/05/2008
  • Issued: 09/04/2012
  • Est. Priority Date: 01/18/2008
  • Status: Active Grant
First Claim
Patent Images

1. A method for recognizing a model object in images, under general nonlinear deformations, comprising the steps of:

  • (a) acquiring in electronic memory an image of the model object;

    (b) transforming the image of the model object into a multi-level representation consistent with a recursive subdivision of a search space, said multi-level representation including at least the image of the object;

    (c) generating at least one precomputed model of the model object for each level of discretization of the search space, said precomputed model consisting of a plurality of model points with corresponding direction vectors, said model points and direction vectors being generated by an image processing operation that returns a direction vector for each model point;

    (d) generating a subdivision of said plurality of model points into a plurality of parts, where a deformed instance of the model is represented by transforming the parts;

    (e) acquiring in electronic memory a search image;

    (f) transforming the search image into a multi-level representation consistent with the recursive subdivision of the search space, said multi-level representation including at least the search image;

    (g) performing an image processing operation on each transformed image of the multi-level representation that returns a direction vector for a subset of image points within said search image that corresponds to the range of nonlinear transformations for which the at least one precomputed model should be searched;

    (h) computing a global match metric for each affine model pose at the coarsest discretization level of the search space, said global match metric combining the results of a local metric, where for the local metric the parts of the model are searched in a restricted range of local affine transformations close to the precomputed model and the maximal fit of each part is taken as the contribution of said part to the global match metric;

    (i) determining those model poses whose match metric exceeds a user-selectable threshold and whose match metric is locally maximal, and generating a list of instances of the at least one precomputed model in the coarsest discretization level of the search space from the model poses;

    (j) computing a nonlinear deformation transformation that describes the local displacements of the parts;

    (k) tracking said instances of the at least one precomputed model in the coarsest discretization level of the search space through the recursive subdivision of the search space until a finest level of discretization is reached;

    (l) computing at each level the respective nonlinear deformation transformation and propagating said deformation transformation to the next level; and

    (m) providing the nonlinear deformation transformation of the instances of the model object on the finest level of discretization.

View all claims
  • 1 Assignment
Timeline View
Assignment View
    ×
    ×