Abstract-3D digital city models form the basis for flow simulations (e.g. wind flow and water runoff), urban planning, underand over-ground formation analysis, and they are very important for automated anomaly detection on man made structures. They consist of large collections of semantically rich objects which have many properties such as material and color. Such user's data structure perception is leading to complex storage schemas. The number of table relations to manage and the large data storage footprint drawbacks are then extended with the fact that not all the systems have a "real" 3D data type.In this work we would like to show our efforts to develop a new kind of Spatial Data Management System (SDBMS) where topological and geometric functionality for 3D raster manipulation will become part of the relational kernel and not an add-on. With it spatial analysis tailored to different use case scenarios is done on-demand and fast enough to support real-time interaction in modern risk management systems.
I. INTRODUCTIONDigital 3D city models play a crucial role in research of urban phenomena; they form the basis for flow simulations (e.g. wind streams and water runoff), analysis of underground formations and man made structures which provide crucial information for effective risk management systems.An urban scene, represented as a 3D city model, consists of large collections of semantically rich objects which have a number of properties such as use, function, and year. They are commonly reconstructed by segmenting and triangulating a point cloud thereby creating a surface representation. Representing urban objects (e.g. buildings, roads, trees, etc.) as surfaces has drawbacks while calculating intersections and volumes, and creating cross-sections is complex. Furthermore, modeling volumetric objects, such as walls, water, and underground, requires the deployment of complex shapes [18].Such users data structure perception is leading to complex storage schemes. The storage scheme designed for systems like Oracle Spatial, Grass, and PostGIS has limitations such as the management of many tables when the selection predicate is on the 3D city model semantics. The number of table relations to manage and the large data storage footprint drawbacks are then extended with the fact not all the systems have a "real" 3D data type. PostGIS, highly adopted in eScience projects, is a clear example.