Matching ‘sticks, plates and blobs’ objects using geometric and relational constraints

https://doi.org/10.1016/0262-8856(84)90003-9Get rights and content

Abstract

A recognition scheme using relational and rough geometric information about three-dimensional man-made objects to recognize instances of the objects in single perspectivenormal views of scenes is described. Experiments performed using the matching scheme show that, in most cases, the object in the view can be identified correctly and reasonable estimates can be made of the unknown camera position responsible for generating the given view.

The technique is based on the fact that the camera position constrains the appearance of the various parts of the object. The propagation of these constraints from one planar object surface to another through the projection equations is worked out. This constraint propagation guides the matching scheme in the development of the interpretation of the scene. The results provide an estimate of the camera position within 20° of the actual location.

References (26)

  • CD Feustel et al.

    The nearest neighbor problem in an abstract metric space

    Pattern Recognition Lett.

    (December 1982)
  • LG Shapiro et al.

    A generalized blob model for three-dimensional object description

  • HB Voelcker et al.

    Geometric modelling of mechanical parts and processes

    Computer

    (December 1977)
  • T Binford

    Visual perception by computer

  • R Nevatia et al.

    Description and recognition of curved objects

    Artif. Intell.

    (1977)
  • D Marr et al.

    Spatial disposition of edges in a generalized cylinder representation of objects that do not encompass the viewer

  • MG Thomason et al.

    Database representations in hierarchical scene analysis

  • Shouxuan Zhang

    A Chinese character recognition system based on pictorial database techniques

  • LG Shapiro et al.

    Organization of relational models for scene analysis

    IEEE Trans. Pattern Anal. Mach. Intell.

    (November 1982)
  • T Pavlidis

    A review of algorithms for shape analysis

    Comput. Graphics Image Process

    (April 1978)
  • LG Shapiro et al.

    Decomposition of two-dimensional shapes by graph-theoretic clustering

    IEEE Trans Pattern Anal. Mach. Intell.

    (1979)
  • DB Gennery

    A stereo vision system for an autonomous vehicle

  • RA Brooks

    Model-based three-dimensional interpretation of two-dimensional images

    Tech. Rep. Stanford University

    (1980)
  • Cited by (15)

    • Learning to generate novel views of objects for class recognition

      2009, Computer Vision and Image Understanding
    • Navigational functionalities

      1995, Computer Vision and Image Understanding
    • Recognition by functional parts

      1995, Computer Vision and Image Understanding
    • Function-based generic recognition for multiple object categories

      1994, Computer Vision and Image Understanding
    View all citing articles on Scopus
    View full text