Abstract
In this article we address the initial phase of the surface reconstruction process in which the input point cloud is processed. The specific task we solve is given by an unorganized finite set of points in the space.
The output is the reconstructed surface so that the points of the input set lie on or close to surface. Several types of data approximations using line in two dimensional space and line and plane in three dimensional space are used for the initial description and preprocessing of the input point set.
Further we address the problem of noise reduction, removal of redundant points and points that do not belong to the object. Spatial orientation of data and symmetries of the point set are also investigated.
In our experiments we apply surface reconstruction on special types of objects. Suggested methods are evaluated on point clouds representing synthetic data, geometric models, and simple parts of real buildings.
All the methods are implemented in the language provided by the MATLAB environment.