Geoscience Reference
In-Depth Information
possible in order to achieve satisfactory performance. A number of approaches can
be identified. (Dollner et al.
2005
) briefly review optimization strategies such as
view-frustum culling, occlusion culling, and back-face culling that operate on gen-
eral graphics primitives. Additional approaches to reducing the volume of data to
be rendered client-side for a 3D City Model include data compression (van Essen
2008
) and mesh simplification (Sester
2007
).
Combining Server Side and Client Side Approaches
The results of sever
side mapping are generally static (i.e. a simple image, or raster). This approach
has the advantage of limiting the amount of data to be downloaded, as well as
allowing the images to be pre-prepared for instant rendering. However, such map-
ping does not provide the interactivity of a vector map, which stores the individ-
ual buildings as separate objects rather than as part of a single image. Although
vector datasets can be overlaid onto 3D raster maps, they often overlap the
buildings as shown in Fig.
2
rather than being correctly placed. An image based
approach cannot be easily navigated around or updated when underlying data-
sets change and does not provide click-and-identify information for the building
objects. Generating a fully interactive map (similar to a GIS, where the user can
click on a point to find out information, i.e. a vector map) (Mitchell
2008
) can be
achieved by making use of the client-side rendering approaches described above,
with, however, the limitation imposed by the hardware on the amount of data that
can be rendered, as well as on the amount of data to be downloaded over what is
frequently a low bandwidth link.
Hybrid Approaches combine server-side and client-side rendering, with the aim
of limiting data to be served over the network and hence processed for render-
ing on the client. For example, (Prieto et al.
2012
) note the importance of render-
ing city models on web-based platforms, and describe a process to take a detailed
point cloud and transform it into a CityGML database structure. They suggest that
the user can be presented with 2D map on which individual 3D buildings can be
requested by clicking. The data is then presented as a single building in a custom-
ised JSON (JavaScript Object Notation) format (Prieto et al.
2012
). Additionally,
(Quillet et al.
2006
) developed a system that extracts feature lines of building
facades on the server side and streams data on demand to the client. The approach
forces the point of view to street level. Ellul (
2012
) describes the use of topologi-
cal data structures to remove shared and hidden internal walls in a model gener-
ated by extrusion, rendering the reduced dataset on the client. (Ellul and Joubran
2012
) also tested an aggregation approach, looking at the impact of generalisation
on rendering performance in Google Earth.
An additional hybrid approach is proposed in this chapter, where the dataset is
pre-structured (triangulated) on the server in order to avoid executing a triangula-
tion process on the client device. Common Nodes (x, y, z points) are identified
server-side and re-used in order to minimise the size of the dataset for transmis-
sion to the client. The proposed approach is described in further detail in
Sect. 4
.
A hybrid approach requires additional processing overhead on the client side, but
has the advantage of interactivity which in turn improves the usability of a 3D
mobile application.
