Statistical Learning in Surface Reconstruction

Yannis Ivrissimtzis


We present a new algorithm for surface reconstruction from unorganized point clouds. The algorithm is based on a special type of Neural Networks, the Growing Cell Structures, starting with an initial Neural Network and learning the geometry of target space in a "competitive learning" process. The network expands by splitting its most active vertices and its quality is constantly improved by removing the least active vertices. The topology changes with statistics based operations which create boundaries and merge them to create handles.
Thursday 18th September 2003, 14:00
Robert Recorde Room
Department of Computer Science