First implementation results and open issues on the Poincaré-TEN data structure

Abstract

Modeling 3D geo-information has often been based on either simple extensions of 2D geo-information modeling principles without considering the additional 3D aspects related to correctness of representations or on 3D CAD based solutions applied to geo-information. Our approach is based from the scratch on modeling 3D geo-information based on the mathematically well-defined Poincaré-TEN data structure. The feasibility of this approach still has to be verified in practice. In this paper, the first experiences of loading a reasonable sized data set, comprised of about 1,800 buildings represented by nearly 170,000 tetrahedrons (including the 'air' and 'earth'), are discussed. Though the Poincaré-TEN data structure is feasible, the experience gained during the implementation raises new research topics: physical storage in one (tetrahedron only) or two tables (tetrahedron and node), effective clustering and indexing improvements, more compact representations without losing too much performance, etc.