| Summary: | In 3D-optimisation of today a decrease in the amount of surfaces to render is often handled by replacing aneighbourhood of surfaces with a simplified and similar set of surfaces. In some cases when the surfacestructure is complex and highly symmetrical, the neighbourhood simplification might produce anomalies thatbreak the symmetry in a visually detrimental way. By knowing the underlying primitive shape of the 3Dstructure, simplified reproductions of a surface structure can be produced without risking the symmetricalintegrity of the object.This report details the possibility of detecting the primitive shapes in a 3D-structure of voxels by analysingthe available data. This is done by looking at neighbourhoods of voxels and using the gathered information toestimate normals and curvature data that indicates a primitive shape. Through categorisation, clustering andfiltering of the data, primitive shapes can be detected with a degree of certainty.Utilising a few different approaches an analysis is performed on the efficiency in the detection of primitiveson the surface of voxel structures. The report concludes that with a sequence of easy-to-follow steps reliabledetection of primitives can be performed.
|
|---|
