Local update of B-spline surfaces by Kohonen neural network

This paper is devoted to the problem of creating and updating a B-spline surface based on a set of scattered data, i.e. totally unorganized points locally distributed over a 3D area. The crucial point of the problem is the parametrization of the given points, where earlier approaches introduced a base surface and the spatial points were mapped onto this surface to gain parameter values. The base surface can be created automatically only in the most simple cases but generally additonal data used to find it. Our method applies an artificial neural network to solve this problem using only the original data set and giving an alternative way to find the best fitted B-spline surface.