bul 2004
ment.
r content
racy, its
addition
y GPS
absolute
imework
üntained
ch lower
| With
mporary
ig to be
stomers.
light be
will be
w years.
mmetric
or as
urce for
:vitably,
al world
pvement
> data to
| (his is
ndful of
ositional
ment of
k data.
tion are
ve been
1
in US
of shift
oftware
he task,
systems
PAI to
well be
mation,
or less
e of the
strained
cometry
here? If
rator
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
should be capable of a degree of solution sufficient to cut
timescales and costs.
4.3 Support for Persistent Public Feature Identifiers.
A major aspect of the new generation of data models being
adopted by national mapping agencies and other data
suppliers is the shift to meaningful features (often termed
‘objects’, at this level of discourse the terms are effectively
synonymous) with unique and persistent identifiers. This
model provides strong support for a change-only or
incremental update service to users. In the geospatial
community this technique was first effectively adopted in the
S57 standard of the International Hydrographic Organisation
(1HO). It is now well established for topographic framework
data, and is used for example by Ordnance Survey Great
Britain in providing an incremental update service for its
national MasterMap and Integrated Transportation Network
datasets. Progress in this area can be monitored via the
regular Joint ICA/ISPRS/EuroGeographics Workshops on
Incremental Updating and Versioning of Spatial Databases,
one of which precedes this Congress (ICA, 2003).
Identifiers also play a key role as the ‘hooks’ by which user
data can be related to framework data. Because they are
relied upon by user applications, they force a more rigorous
approach to the semantics of update. Lifecycle rules have to
be defined across the tasks of creation, deletion, splitting and
merging of objects and the modification of their geometric,
thematic, topological and temporal descriptions. These have
to cover identifiers, and the circumstances in which identity
is retained or lost.
All update processes need to be aware of these rules and to
enforce them if update is to be efficient. They also need to
respect existing identifiers and be capable of issuing new
ones, either by access to a central registry or by a surrogate
mechanism, leaving the final assignment to the commit stage.
In all events the trend towards this form of data model
strengthens the case for tighter integration between the
database and the photogrammetric systems.
It is worth noting that if framework data had provided
identifiers as hooks from the outset, the consequential
problems for user data from PAI would have been avoided.
4.4 Support for Topology
The desirability of active topology maintenance in update
processes has been well documented and the lack of it is one
of the major contributors to the extended timescales and
repeated round trips of the early generation of systems
(Edwards, 2000). For 2D data, support for topology is
becoming increasingly available. The recently released
Oracle 10g provides support for the storage of topology.
Laser-Scan’s Radius Topology complements this with active
server-side topology maintenance (Laser-Scan, 2004).
ESRI’s ArcGIS (ESRI, 20023) provides topology support on
the client side. There is still debate, as yet unresolved, as to
which approach provides the better overall efficiency.
The essential point for the discussion of integration is that the
technology for active topology management in 2D is well-
established and available. A level of integration which
exploits this will deliver substantial productivity gains, as
slivers and overshoots become things of the past. A key
763
overall architecture decision to be made is whether active
topology maintenance needs to be available only through the
GIS interface, or to any application that uses the standard
database interface.
Photogrammetry is of course intrinsically concerned with 3D
data. Support for topology in 3D is a much more open issue
and is left to the next section.
4.5 Further Benefits
The use cases described in this section, which all represent
major aspects of the overall task, all argue towards a greater
degree of integration between photogrammetry and the
database. Further benefits are potentially achievable with
better data management and closer alignment with the IT
mainstream. Photogrammetry system vendors and GIS
vendors seek to off-load transaction management, history
management and archiving to the DBMS. Furthermore, it is
arguable (Garland, 2004) that the greatest savings may arise
from the adoption of workflow management technology,
which becomes accessible with improved data management
environments.
5. STANDARDS AND 3D SUPPORT
5.1 Standards
There is a common feeling that the integration of
photogrammetry and GIS databases is not yet at a
sufficiently mature stage to be the subject of Standardisation.
In particular the lack of adopted and implemented standards
for 3D data is an obstacle. GML2 from the Open GIS
Consortium has proven its worth for transferring and serving
2D data and for supporting incremental update. It is being
widely adopted. Implementations of GML3, which supports
3D data and topology, are at a very early stage. Some
photogrammetry vendors have reported early experiments
with GML (Olhof et al, 2004).
Nevertheless, from the wider perspective, the shift to
database-centric environments stands to benefit from the
progress towards interoperability and better access to
information through the application of OpenGIS and ISO
specifications.
As an aside, the progress towards the Sensor Web (OGC,
2004) as reported in a special session of this Congress will
have significant effects on front-end data and imagery
gathering processes.
5.2 Stages towards 3D support
Databases which support spatial data typically include
support for 3D data (although with significant restrictions in
areas such as indexing and query). There is no difficulty in
storing 3D information, although many national mapping
agencies still do not retain z-values even when they are
captured. This is changing as market demand for 3D vector
data in addition to DEM increases. In practice 2.5D vector
data (z as a single valued function of (x,y)) can be created
from 2D data and sufficiently high quality DEM data.
Issues arise in the handling of multiple z-values and topology
in 3D. The implementation of full 3D topology, whilst
defined in the ISO standards, is a long wav off. Many
observers, the author included, would advocate seeking this
