3D Shape Description with Principal Curve

The unordered and non-uniform shape distribution of 3D models is converted to an ordered and uniform one of 1D manifold.By projecting 3D models into 1D smooth principal curves,equally segmentation and sampling along the length of curve are implemented to extract shape feature vectors,then one can retrieve 3D model according to the matching result of shape feature vectors of 1D manifold.The experiments prove that this shape description is invariant under transformations and the errors brought by multi resolution,local morph,and noises etc are depressed.