International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004
entire objects and to individual geometric elements. This also
makes it possible to select, analyse or edit the geometry and the
appearance of objects based on semantic criteria.
2.2 Object geometry
Cultural heritage objects often have a very irregular complex
geometry. Thus, a good digital reconstruction requires a very
detailed 3D model with a lot of geometry elements. So there are
two main requirements to a 3D GIS. The first one is a support
for the acquisition and handling of large amounts of complex
and non-planar 3D geometry. The second one is the
visualisation of these objects which consists of a lot of
geometry elements.
2.3 Photo-realistic texture
For realistic virtual 3D models object textures are needed. It
must be possible to assign aerial or close-range imagery to the
individual geometry clements. Ideally, the resulting texture
information should also be managed within the 3D GIS. Major
challenges include the full automation of the texturing process
and the visualisation of objects with high-resolution texture
information.
3. THE DILAS 3D GIS
3.1 Overview
DILAS (Digital Landscape Server) is a comprehensive 3D GIS
platform for the integrated management of regional to national
3D landscape and city models and for the generation of web-
based geoinformation services (Nebiker, 2002a). DILAS™ is
the result of a joint research project and is now a commercial
product line of GEONOVA AG (www.geonova.ch).
The next section highlights some of the key concepts developed
and implemented as part of the DILAS project: a flexible 3D
object model, a multi-representation and multi-resolution
approach for the different object types, a storage concept for 3D
and raster objects and XML-based process rules.
3.2 DILAS Concepts
3.2.1 3D Object Model — One of the key concepts of the
DILAS project is a generic, fully object-oriented model for 3D
geo-objects. This object model incorporates a 3D geometry
model which is based on a topologically structured 3D
boundary representation and which supports most basic
geometry types (points, lines, planar and non-planar shapes as
well as a number of geometric primitives). It incorporates the
capability for multiple levels of detail (LOD) as well as texture
and appearance information. The 3D object model is suitable
for representing any spatial topic (e.g. buildings, bridges,
power-lines).
The DILAS 3D object type is supplemented by a number of
spatial data types used for representing very large mosaics of
high resolution terrain and texture data:
e raster maps
e orthoimagery
e terrain and surface models (regular grids)
e terrain and surface models (irregular point clusters), e.g.
for managing very large laser scanning height data sets
The concept for the DILAS 3D object model already
incorporates visualisation-related features such as viewpoints
and lighting information. In addition the model has been
explicitly defined to easily cater for future extensions, such as
animation paths etc.
3.2.2 Multi-Representation and Multi-Resolution — Two
key issues in the efficient management and visualisation of
large 3D models are multiple representations and multiple
resolutions. Different multi-representation strategies were
developed for the spatial object types used in DILAS. The
original multi-resolution approach for managing very large
raster mosaics (Nebiker, 1997) was further refined and
extended to all mosaic types listed above.
3D objects are represented by 3D bounding boxes, 2D object
boundaries and the actual 3D geometry (Nebiker, 2002c). The
first two representations are essential for efficient query
operations and are automatically derived from the main 3D
representation.
3.2.3 Storage Concept for 3D Objects — The goal for
handling and manipulating 3D objects was to provide an
optimum modelling flexibility in combination with an excellent
object query and retrieval performance. The developed concept
is based on the following components:
e a 3D object representation in Java and XML
e a3D object serialisation and de-serialisation
e apersistence framework built on top of the DBMS
e spatial data types for 3D and raster objects within an
object-relational environment
A number of these mechanisms are adapted from modern
object-oriented programming environments. The object
serialisation approach, for example, permits to map very
complex objects to a simple, but highly efficient storage
mechanism. The storage mechanism is based on a type
extension for 3D objects which encapsulates the actual large
object based (LOB) object storage.
The persistence framework developed in DILAS adapts
concepts from the Java Data Objects (JDO) extension. It
permits a fine-grained control over changes to the 3D object
properties.
3.24 XML-based Process Rules — The processes of
importing, structuring, generating and validating 3D city
models are quite complex and typically differ from organisation
to organisation, e.g. different level assignments, exchange of
geometry only versus exchange of actual 3D objects etc.
The goal of accommodating these diverse requirements led to
the development of a mechanism using "XML-based process
rules’. The benefits of this rule-based approach are:
e The possibility of formally specifying valid processing
options (e.g. data import options) through the means of
different XML Schemas.
e The easy adaptation of process rules or the creation of new
process rules by a project leader or system administrator
and the possibility of easily integrating these rules into the
user interface.
e A rigorous validation of user-defined process rules by
means of standard XML tools and mechanisms.
3.3 DILAS System Architecture
One of the design goals of the DILAS project was to rely on
state-of-the art commercial database technologies. The current
system is using an Oracle 91 DBMS. The DILAS system
Internat
Men
consists
DILAS
2002b).
The S
compor
manage
represei
The 3D
of Ben
mappin
MicroS
Java A
geomet
import/
Current
the edi
buildin;
interact
The DI
web-ba
models
softwar
based |
objects
3.3.2
factors
technic
landsca
enviror
In DIL
OGC S
spatial
yields :
e th
ac
9 th
ati
oe th
or
3.4 Sr
In addi
DILAS
