Geoscience Reference
In-Depth Information
zoom and pan timings for the “Dynamic Dataset” refer to operations on an extract
of the dataset, with the zoom data having an average extent of 500 m and the pan
data of 385 m, compared to the 1,000 m extent of the entire dataset. The time for
a pan operation varies between 2,319 and 10,204 ms, and for a zoom operation
between 4,196 and 24,509 ms. The latter value corresponds to a total of 298,629
nodes, i.e. the user has zoomed out to view almost the entire generalised dataset.
6 Discussion and Further Work
The results for the “Full Dataset” test show the potential of downloading an entire
dataset into the device's memory, as once this is done the interaction with the
resulting model is instantaneous. However, while this approach was successful on
the relatively high specification test device, when testing on two other devices (an
Acer Iconia 10.1 Tablet with 1 GB of RAM and a Samsung XCover 2 Smartphone
with 800 MB of RAM) it was not possible to visualise the dataset at the full 1 km
extent due to an 'Out of Memory' error. Further attempts to reduce the memory
consumed by the App (specifically, by writing the downloaded XML file to disk
rather than storing it in memory) did not resolve this issue and also added to the
rendering time, due to time taken to first write the file to disk and then re-read the
data to generate the 3D scene.
However, the results obtained using the “Dynamic Dataset” did not prove
entirely satisfactory either, as while initial download time was halved, the zoom
and pan times were greatly extended. This approach also requires additional net-
work bandwidth, which may incur a cost. Overall, therefore, the results obtained
can be said to confirm that, as suggested by Nielsen and Budiu (
2012
), there is
a compromise to be made between providing detailed information to allow users
to self-localise and navigate efficiently and the requirements of devices with low
memory working on low bandwidth networks. The adequacy of 3D detail, the vis-
ual impact of the resulting 3D dataset, the suitability of the response times and
the overall usability of the 3D model is subjective and will depend on the spe-
cific application for which the 3D City Model is to be used i.e. the context of use
(Kjeldskov and Graham
2003
).
The results of the “Dynamic Dataset” test do, however, confirms the techni-
cal potential of a more dynamic approach to 3D City Modelling on mobile devices,
showing that it is possible to store data in a spatial database using real-world (British
National Grid) coordinates and visualise it on a mobile device via a series of dynamic
requests based on a pinch-zoom movement as suggested by Nielsen and Budiu (
2012
)
and a pan event combining the various swipe operations (Nielsen and Budiu
2012
).
Using this approach, the initial download and setup time can be reduced by focusing
the 3D dataset on the user's current location (or an area of specific interest that they
specify) and providing an initial, generalised, dataset for display. The ability to change
detail in the dataset depending on the view extents contributes to the usability and self-
localisation requirements identified in
Sect. 2.2
, and will permit end users to access a
