Geoscience Reference
In-Depth Information
and Graham
2003
). Because of this, as with 2D maps, the scale at which the data
is visualised is an important factor to consider, with detail being more appropriate
for large scale visualisation but perhaps appearing as “clutter” at smaller scales,
where a simplified dataset could be said to be more visually appealing, without
significant loss of detail or general shape of the overall building.
To contribute to the understanding of the performance implications of these
larger datasets, and investigate the potential of scale-depending rendering, this
chapter compares the rendering performance obtained when visualising a 3D data-
set using two different approaches. Firstly, an entire city model is pre-loaded onto
the device and interaction tests conducted, and secondly a subset of the model is
dynamically loading depending on the user's view extent, with the data re-loaded
each time the user changes the viewpoint, making use of a more generalised data-
set for larger viewing extent (smaller scale) data requests.
2 Background
3D datasets have varying levels of detail (LoD) (Kolbe et al.
2005
)—ranging from
LoD 0 (a digital terrain model), through LoD 1 (a block model without any roof
structures), LoD 2 (a city model having explicit roof structures and potentially
associated texture) and moving up to LoD 4, which includes interior structures.
Such data can be generated from digital ortho-photos, 2.5D image draping, extru-
sion, Computer Aided Design (CAD) models (Batty et al.
2001
), LiDaR (Light
Detection and Ranging) point clouds, applications such as PhotoSynth (
2012
) and
terrestrial Laser Scanning. Single sources of data have been used (Tse et al.
2008
);
however, it is more common to combine multiple sources of data (van Essen
2008
;
Richmond and Romano
2008
; Wang and Sohn
2011
). Extrusion provides an auto-
mated method to generate a 3D model—combining 2D topographic mapping
with height information derived from LiDaR data. This gives a rapid mechanism
for generating an entire City Model to LoD 1. Importantly, where detailed topo-
graphic mapping also exists at a national level, the potential now exists to create
a national 3D dataset at Level of Detail 1. This process, however, results in very
large, detailed, 3D datasets. This section presents an overview of both the tech-
nical (system architecture) issues and usability issues—specifically interaction
design and spatial literacy—to be taken into account when visualising such data-
sets on mobile devices.
2.1 Improving Performance: Examining System Architecture
Research that has been conducted into city modelling for mobile devices is per-
haps limited due to the fact that devices have only recently become powerful
enough to render 3D graphics. Indeed, mobile computation still faces various
