ul 2004
al road
her by
ising à
/stem).
s have
of the
hod is
taken
ther at
inge is
pad to
many
hange
vay to
ite the
anged
|. This
abase.
years
s itis
ich to
poral
en the
about
action
level
„A Abe
have
ct the
This
years.
ques,
iction
The
iction
g the
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
The methodology for road change detection and map updating
based on map conflation technology is discussed in the next
section.
4. ROAD CHANGE DETECTION AND MAP
UPDATING BASED ON MAP CONFLATION
TECHNOLOGY
Map conflation techniques are used in the change detection and
updating stage in this research due to the following reasons.
1) Feature matching is a well-known map conflation
technology to determine the conjugate features between
two different versions of geographic databases. Both node
feature matching and linear feature matching technologies
could be used in road change detection processing;
2) Through feature matching we can not only identify the
conjugate features so that the relative accuracy of the two
versions of databases can be determined, but we can also
determine which parts of the road network have changed
and which parts remain unchanged. In addition, conflation
operations can be performed to transfer attributes to the
new database based on conjugate features;
3) Map conflation is also useful in the case where the
improvement of the positional accuracy of the original
version of the road database is necessary or desired. Both
node-based conflation (e.g. TIN approach in [Saalfeld,
1993]) and polyline-based conflation (e.g. polyline
mapping method in [Filin and Doysther, 1999; 2000])
could serve to correct the old version of the road database;
4) Map conflation is originally an editing operation in GIS to
reconcile the position of related features. So the results
from map conflation will have a good consistency between
the changed and the unchanged road features.
After a successful extraction of road features from the imagery,
feature matching is performed to identify the conjugate road
nodes and road centerlines from the two versions of the road
datasets. A conflation procedure then follows to obtain a new,
more accurate road database. The main steps are detailed in the
following subsections.
4.1 Node matching
Node matching, i.e., identifying the conjugate road nodes in the
two versions of the road database, is usually the first step in
map conflation.
Both distance and topological similarity measures should be
used to find possible conjugate nodes. The calculation of
distance similarity is usually based on the positional
discrepancy between two points. The topological similarity is
based on the "Spider code" of a road node which was originally
presented by Saalfeld (1988, 1993). It is a measure of the
structure information of a node based on the number of linked
arcs and corresponding directions.
4.2 Point-based conflation
In this step, a point-based conflation is performed on the
original road database in order to reduce the positional
731
discrepancy between the two versions of the road database. The
typical point-based conflation procedure is based on a
piecewise local transformation determined by two TINs
constructed from the conjugate nodes (see Cobb et al [1998] for
a detailed description).
4.3 Polyline matching
Polyline matching is mainly used to identify the conjugate road
lines, but at the same time, it could also be used to detect road
changes.
Although both distance and shape similarity measures could be
used for polyline matching, none of these measures is well
defined for polylines.
There are some definitions for distance between two polylines,
such as Hausdorff distance [Hangouet, 1995], L, distance
[Saafeld, 1993], etc. Walter and Fritsh (1999) used buffering to
assess the distance similarity between two lines in road feature
matching. However, all these distance measures failed to give a
meaningful similarity index for partially-matched line pairs. A
modified intermediate area approach is used in this research
which could solve the multiple matching issues. An appropriate
shape similarity measure for road polylines may also be helpful
in feature matching, however, it is still under development.
Based on the results from polyline matching, a road will be
categorized into one of the following cases:
1) Unchanged, if the road arcs are successful in finding
conjugate features;
2) Disappeared, if the road arcs in the original version of
database failed to find conjugate features;
3) Created, if the road arcs in the new version of database
failed to tind conjugate features;
4) Changed, if the road arcs are successful in finding
conjugate features but the positional discrepancy is
significant;
5) Partially changed, it the road arcs are partially successful
in finding conjugate features.
All this information will be helpful in updating the road
database and will be used in the next processing step.
4.4 Polyline-based conflation
Polyline-based geometric conflation could be used to improve
the accuracy of the original GIS data. This step is optional if the
correction of the positional data of road polylines is not desired.
Filin and Doysther (1999; 2000) introduced the polyline
mapping method to correct the old version of the road database
based on matched polylines.
4.5 Attribute transferring
Transferring the attribute data from the old version database
into the new one is an important function for a map updating
system. This step is simple for unchanged road features because
these features will have 1:1 matching in the polyline matching
step. However, problems may occur for those that have been
recognized as partially changed. This usually should be done
with the aid of user interaction. An appropriate "transferring"
