Fitting Surfaces to Point Clouds


FarFields's FastRBFTM toolbox enables surfaces to be fitted to point cloud data. A point cloud is compactly represented by a signed distance function which is smooth and continuous.

  • Accuracy may be specified globally or locally
  • Normals may be specified with the point cloud data to further constrain the fitted surface
  • A unique interpolant that minimises a surface-energy functional is fitted with FastRBFTM
  • A surface mesh can be constructed at any desired resolution from the RBF representation
  • Mesh optimisation produces uniform facets of good aspect ratio (small and elongated triangles are avoided)
  • Manifold surface guaranteed
  • Mesh can be clipped at specified bounding planes
  • Mesh can be closed at specified bounding planes (watertight surface)
  • Noisy data can be approximated or the fitted RBF smoothed with low pass filtering during evaluation (meshing)

Polhemus FastSCANTM examples
Laser scanned point cloud of an ear Fitted RBF surface
Laser scanned point cloud from a hand (27,000 points) A closed RBF surface
consisting of 2,600 terms

Fitting to standard models

(Courtesy of Stanford Computer Graphics Laboratory)

*Data courtesy of Stanford computer graphics laboratory
Raw point cloud data without normals  Automatically fitted surfaces 

Further examples