• DocumentCode
    3368082
  • Title

    An edge-based mesh segmentation method for engineering objects

  • Author

    Liu, Min ; Ramani, Karthik

  • Author_Institution
    Dept. of Precision Instrum. & Mechanology, Tsinghua Univ., Beijing, China
  • fYear
    2010
  • fDate
    26-28 June 2010
  • Firstpage
    511
  • Lastpage
    514
  • Abstract
    This paper proposes a triangle mesh segmentation technique for conventional engineering objects. We exploited two key aspects of engineering geometry: (1) the intrinsic discontinuities existing in boundary surfaces and (2) the anisotropic behavior of true discontinuities. A two run non-linear diffusion algorithm based on an optimal estimation theory is designed to filter out undesired discontinuities. The first run is for treating C1 continuities where the crease edges are detected using anisotropic re-estimation of triangle normal. The second run is for handling C2 continuities in which transition edges are derived through diffusion of curvature tensor associated with each triangle. Morphological operator and geometric snake model are implemented for interpolating smooth and closed feature loops. Results indicate that our method can determine surface patch layout satisfactorily.
  • Keywords
    computational geometry; edge detection; estimation theory; image segmentation; interpolation; mesh generation; object detection; curvature tensor; edge detection; engineering geometry; engineering object; geometric snake model; interpolation; morphological operator; nonlinear diffusion algorithm; optimal estimation theory; triangle mesh segmentation technique; Algorithm design and analysis; Anisotropic magnetoresistance; Estimation theory; Filtering theory; Filters; Geometry; Solid modeling; Surface morphology; Surface treatment; Tensile stress; mesh segmentation; reverse engineering;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Mechanic Automation and Control Engineering (MACE), 2010 International Conference on
  • Conference_Location
    Wuhan
  • Print_ISBN
    978-1-4244-7737-1
  • Type

    conf

  • DOI
    10.1109/MACE.2010.5536720
  • Filename
    5536720