Geoscience Reference
In-Depth Information
From 3D City Modelling to 3D Country
Modelling: First Tests on Scale-Dependent
Rendering on Mobile Devices
Claire Ellul and Julia Altenbuchner
Abstract
To date, 3D City Models have been confined to well defined city
boundaries, and as yet have not been considered in a multi-city integrated man-
ner, at country scale and beyond as would be the case for 2D spatial information.
However, increasingly National Mapping and Cadastral Agencies (responsible
for generating country-wide mapping) are investigating the possibility, financial
feasibility and utility of making such datasets available over a broader area. The
potential size and detail of such 3D datasets has led to a challenge when rendering
such data on mobile devices and understanding the limitations to the usability of
such models on these devices is particularly important given the broadening range
of applications—such as pollution or noise modelling, tourism, planning, solar
potential—for which these datasets and resulting visualisations can be utilized.
This chapter describes preliminary work carried out to compare two approaches
to rendering 3D data on mobile devices for the test area of the city of Sheffield
(in the UK Midlands). The data is stored in a 3D spatial database as triangles
and then extracted and served as a web-based data stream which is queried by
an App developed on the mobile device (using the Android environment, Java
and OpenGL for graphics). Current approaches to rendering suggest loading the
full city model onto the device and then interacting (e.g. zooming and panning)
with the city model stored in memory. However, it is suggested that, in particu-
lar for lower specification devices, this may cause memory issues and we com-
pare the standard approach to a second, dynamic, approach where a subset of data
is downloaded and rendered each time the user zooms or pans around the model,
with generalised data being used for larger extents as would be the case for 2D
spatial data. Results of 24 s (dynamic dataset) versus 54 s (full dataset) for initial
