Spatial database research has traditionally been considered
in a 2D space. The Earth is not flat, and there are many spatial
applications that must consider a 3D terrain model, which can
typically be approximated by a triangular mesh with millions of
3D points. Terrain data is very large,
and surface-based spatial query processing can be very costly
in terms of both computational cost and I/O cost. In this talk,
we will introduce the concept of multiresolution terrain
modelling, which supports on-the-fly derivation of approximate
surface models at any required level of details. We will discuss
various database techniques we have developed to support
efficient processing of view-dependent visualization queries and
k nearest neighbour queries. We will also outline other
challenging research issues along this research direction. |