• DocumentCode
    2775733
  • Title

    Self-Organizing Feature Map (SOFM) based Deformable CAD Models

  • Author

    Igwe, Philip C. ; Knopf, George K.

  • Author_Institution
    Univ. of Western Ontario, London
  • fYear
    0
  • fDate
    0-0 0
  • Firstpage
    3798
  • Lastpage
    3804
  • Abstract
    An adaptive modeling approach that uses a self-organizing feature map (SOFM) to create deformable hexahedral meshes for interactive geometric modeling is presented in this paper. The technique uses the nodes of a three-dimensional SOFM to represent discrete point masses that comprise a solid object. Although the geometry of the resultant mass-spring mesh will change under the influence of inputs applied through a haptic tool and interface, the relative connectivity of neighboring nodes in the time-varying mesh are maintained under the external and internal forces. The initial mesh can either be retrieved from a library of primitive shapes, or created by automatically fitting the topology preserving SOFM to selected surface points. The designer reshapes the virtual object by applying external forces and pressure to the initial mesh. The accuracy of the system depends on the mathematical equations used in formulating the model behavior. The model behavior can be altered by changing the material properties in the underlying mathematical equation. Examples of shape deformation are provided to illustrate the concepts introduced.
  • Keywords
    computational geometry; mesh generation; self-organising feature maps; 3D SOFM; adaptive modeling; deformable CAD model; deformable hexahedral mesh; discrete point mass; haptic interface; haptic tool; interactive geometric modeling; model behavior; resultant mass-spring mesh; self-organizing feature map; solid object; time-varying mesh; topology; virtual object; Deformable models; Equations; Geometry; Haptic interfaces; Libraries; Mathematical model; Shape; Solid modeling; Surface fitting; Topology;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Neural Networks, 2006. IJCNN '06. International Joint Conference on
  • Conference_Location
    Vancouver, BC
  • Print_ISBN
    0-7803-9490-9
  • Type

    conf

  • DOI
    10.1109/IJCNN.2006.246873
  • Filename
    1716621