Geoscience Reference
In-Depth Information
seamless 3D dataset no matter their current location in the country or specific area of
interest. The ability to dynamically interchange datasets (from detailed to generalised)
also adds the potential to incorporate application-specific 3D data to a generic App.
This would help to move 3D modelling towards the situation currently in place for
2D, where generalisation (and in particular object removal or exaggeration) takes full
account of the end user's requirements, resulting in a situation where the 3D model
would show detail for popular landmarks for tourism but de-emphasize these and focus
on the detail of proposed buildings for planning-related applications. The dynamic
approach also has the advantage that the user can immediately see changes made by
other users, important for collaborative working. It gives the App designer the option
to provide enhanced support for spatial literacy issues as suggested by Hildebrandt and
Timm (
2013
) and (Fling
2009
) in
Sect. 2.3
.
In both the dynamic and full cases, the quantity of downloaded data could
also be reduced by the use of the indexing approach described in Ellul and
Altenbuchner (
2014
). This would in turn reduce the XML Parsing and Mesh Setup
times which currently form a significant portion of the overall display time for
the dataset. Additionally it may be possible to combine the work described here
with additional pre-preparation of datasets and implementation of the techniques
described in
Sect. 2.1
such as data compression (van Essen
2008
) and mesh sim-
plification (Sester
2007
) over a block of buildings to test for optimal performance.
Techniques to reduce the volume of transmitted data and the time required to parse
this data also include the use of (Crockford
2006
) Javascript Object Notation
(JSON) as a smaller footprint data transmission mechanism. As with the visual
aspects of the data, in the case of performance the end user requirements are also
paramount. For example, definitions such as “an application should respond within
2 s to provide users with a feeling of interactivity” cannot be applied universally
and in mobile applications, 2 s is too long for an application that communicates
with a driver (Marsh and Haklay
2010
).
The pre-creation of the triangles for rendering potentially resulted in overall
performance improvements. However, this currently means that the triangles are
de-coupled from the original data, meaning that changes in the latter will not be
reflected. The “Ear Clipping” triangulation algorithm is a simple and robust algo-
rithm which works well for geospatial polygons (Cozzi and Ring
2011
). It is under-
pinned by an assumption that the polygons which represent the building roofs in
spatial datasets are “simple polygons” (i.e. do not have internal holes or self-inter-
sections), iterating around the Nodes making up the polygon removing each trian-
gular “ear” as it goes. Triggers could be developed inside the Oracle database to
automatically re-triangulate the data when the underlying buildings change.
More generally, the datasets used for this test, the multiple methods used
to generate 3D data, the increasing availability of Building Information Models
(BIM, which go beyond LoD 4 to model a digital representation of physical and
functional characteristics as a shared resource to be used throughout a building's
life-cycle (Smith
2011
), and are now becoming mandatory in large government
construction in the UK (Ngo
2012
) ) and the wide range of potential applications
that make use of 3D mobile mapping and City Models all highlight the fact that it
