International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
  
  
  
2.3 Data Pre-processing 
The raw data of all three digital elevation models is organised 
in a grid structure, where every point has an east, north and 
height value. The distance in cast and north direction is the 
same. For reusing the GI-System techniques it is necessary to 
create images including the information of a digital elevation 
model. We need to create a georeferenced image with each 
image pixel holding the height information of a grid point of 
the DEM. This height information is represented by the 
colour value of the image pixel. Thus it is necessary to find 
an image format, which colour depth value is large enough 
for handling the height resolution of the digital elevation 
model. It is necessary to define an interval including the 
height range of the digital elevation model to calculate the 
resolution. 
  
Calculation of height resolution: 
height resolution = (max height value — MIN height value) / 
(depth value — 1) 
  
  
  
One colour value should be reserved for grid points without 
defined data in the raw data set. All other colour values can 
be used for encoding the height information. The following 
shows the calculation of the colour value for one image pixel. 
  
Calculation of colour value: 
(depth value — 1) / 
(max height value MIN height value) 
height value * factor 
factor = 
  
  
colour value = 
  
Using this technique it is possible to create images including 
the height information. Additionally one file for each image 
is created including the bounding region of the image for the 
georeferencing. 
Using this technique for the integration of the digital 
elevation models in the UIS of Baden-Wiirttemberg we can 
consider the following conditions: 
a) The accuracy of the given height information in all 
raw datasets (50m, Sm and 1m) is +/- 0.1 m. 
b) All possible heights of Baden-Wiirttemberg fit in 
an interval from —500 m to 3500 m. 
c) That leads to a grey value depth of the image of 16 
bit (65 535). 
d) PNG-images with 16 bit depth are used to store the 
digital elevation model. 
2.4 Raster Tiling 
It is possible to treat the created images like other raster data: 
Use the given tiling, storing and accessing techniques for the 
management of the DEM data. 
In our example we use the already existing tiling algorithm of 
the base GIS GlSterm Framework. GlSterm includes classes 
and methods for tiling images, storing them into the database 
and accessing them. These methods include possibilities for 
managing data in raster pyramids in different resolutions 
(Hofmann/Veszelka/Wiesel 1999). This techniques 1s called 
raster tiling. It is possible to totally reuse these parts of 
GISterm. 
We can use the existing techniques for storing all three DEM 
data sets into the database. 
518 
2.5 Memory Consumption of the Example 
Reviewing the data handling section up to this part of the 
paper we firstly looked at the data basis (Raw ASCII data of 
digital elevation models in 50 m , 5 m and 1 m resolution). 
Secondly we created PNG-image files from the raw data. 
Thirdly we stored these images in the database using the 
raster tiling approach. The following table summarises the 
amount of memory consumption in all three states of the 
high resolution digital elevation model data needed for the 
entire area of Baden-Württemberg: 
  
  
  
  
  
  
  
  
  
1m 5m 
- 933 GB ~ 38 GB 
PNG ~ 11 GB ~ 1 GB 
PNG in database ~21 GB ~2GB 
  
Tab. 1: Memory Consumption for high resolution DEM data 
of Baden-Württemberg 
Thus, we can easily store the complete high resolution DEM 
of Baden-Württemberg into the database. The next part 
concentrates on accessing of the data in the database. 
2.6 Access to the Data in the Database 
DEM data are used in different applications. Thus we need a 
generic interface for getting the data out of the database. This 
interface is called Height-Service and offers classes and 
methods for providing the data for further processing 
(Hilbring 2002). The Height-Service is realised as newly 
created extension of the base GIS GlSterm for accessing 
DEM data in the database. 
The basic information needed by the Height-Service for 
accessing the data is the bounding region of the requested 
DEM part. Based on the size of this bounding region the 
Height-Service looks for an appropriate DEM in the 
database. 
2.6.1 What is an appropriate DEM? 
On the one hand the decision includes the aspects of the final 
displaying resolution. For example it is not sensible to load 
the entire DEM of Baden-Wiirttemberg in 1 m-resolution into 
a 3D scene, because the screen resolution is not high enough 
for displaying. On the other hand, loading such a large DEM 
needs huge amounts of memory and creates performance 
problems while handling the 3D object in a virtual scene. 
Practical tests showed that 500 000 base points can be 
handled by the application with reasonable response time. 
  
| Bounding region | 
| Decision for DEM resolution A 
| 
| Locate appropriate DEM in database? 
] 
[ | 
| True | False 
| | 
| Return access to DEM | | Take next resolution 
  
  
  
— 
  
Fig. 1: Choice of appropriate DEM 
Taking these aspects into account, the Height-Service decides 
from the size of the bounding box and the number of 
expected points on an appropriate resolution of the DEM and 
as 
puces 1 NM