ect. 
ons 
orld 
»yer 
tion 
  
  
nap 
n & 
on 
data 
been 
port 
t of 
n on 
d on 
> the 
bute 
ing. 
soft 
ross 
for 
) are 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B4, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
One important aspect is compatibility to other geospatial data 
services. Therefore, we chose to use the OpenStreetMap tile 
layout for the tile storage. This makes it very easy to also use 
our processed data in 2d applications like OpenLayers. 
A general data processing workflow has been created to 
simplify data processing from small to very large datasets. The 
OpenWebGlobe processing toolkit consists of a number of 
command line tools. A typical processing workflow is shown 
below: 
1. The first step in data processing is determining the extent 
of your dataset. You can use the tool called 
"ogCalcExtent" to calculate the extent of your data in 
WGS84 and tile coordinates. 
2. Once the extent of the data is known, a new layer must be 
created. This is done using the "ogCreateLayer" tool. 
3. After creation of the layer, data can be added. This is done 
using the "ogAddData" tool. 
4. When elevation data is being added the "ogTriangulate" 
tool is used in this step to create geometry from the 
previously added data. À large scale delaunay triangulation 
is used as shown by (Christen & Nebiker, 201 1a). 
5. When adding data is finished, levels of detail can be 
calculated. This can be done using the "ogResample" tool. 
Recently new tools have been developed. They are used for 
rasterizing vector data to image tiles using Mapnik, for creating 
hillshading and normal maps, as well as OpenStreetMap data 
with transparent background for overlays (Wüest, 2012). An 
introduction and a step by step tutorial for data processing with 
the OpenWebGlobe processing tools can be found in (Christen, 
2012) 
4. SELECTED APPLICATIONS 
Among the 3d web applications created with the 
OpenWebGlobe SDK two are especially notable. The first one 
is a 3d viewer of Switzerland displaying high resolution 
imagery (25cm/pixel) and elevation data (25m grid). The raw 
data size was around 1.2TB. All data were directly 
preprocessed in the cloud. 26 million tiles were generated and 
stored on an amazon S3 instance in a pyramidical folder 
Structure with 19 LOD steps. This platform was created for the 
Federal Office of Topography (Swisstopo) and used as a 
demonstrator project for the Open WebGlobe technology. Also 
some textured and untextured 3d models of swiss towns and 
selected buildings were integrated. Additionally a special 
image layer shows the country borders. The application also 
provides a localisation search service which allows searching 
and flying to a given town in Switzerland. A Screenshot of this 
application is shown in figure 7. The platform is accessible 
online under swiss3d.openwebglobe.org. 
OpenWebGlobe DemoViewer Switzerland 3D 
   
  
  
Figure 7. Screenshot of the 3d Viewer Switzerland 
swiss3d.openwebglobe.org 
Another application for rapid 3d mapping and geolocalization 
was created in summer 2011 as part of a research project 
(Ladetto et al, 2011). The application offers possibilities to 
integrate and compare different 3d point clouds captured from 
laserscanning or stereo vision mobile mapping systems. Bulk 3d 
point cloud data was preprocessed prior to its integration into 
OpenWebGlobe. Data from a stereo vision SLAM device such 
as point cloud fragments, position and orientation data is 
continuously loaded and displayed by the application. Figure 8 
shows a Screenshot of the application with a point cloud of the 
building exterior in an overview. Different predefined or 
dynamic view positions allow the user to investigate the 
building in more detail or to dynamically follow and observe 
moving objects or targets. Viewing modes like the 'follow 
mode' in which the user sees the virtual globe in the view of an 
individual or the 'third person mode' which allows observing a 
scene from a predefined distance and automatically follows any 
moving target are other useful features. 
  
E 
Figure 8. Screenshot of the rapid 3d mapping application 
5. CONCLUSIONS AND OUTLOOK 
In this paper we presented an open source framework for the 
creation of large scale virtual 3d globes with highly detailed 
contents and for their interactive visualization directly within a 
broad spectrum of web browsers. While commercial virtual 
globes are largely unrivalled in terms of global coverage and 
189