d national
SCAN for
and single
nition and
sportation
of ca. 2.5-
ologically
1, size and
OR25 data
ty include
Regarding
he general
ctors may
image on
r with any
layers in a
) be stored
stem used
| platform.
> so-called
lected and
N
e lowlands
ges. Other
25 m pixel
raphy and
al length),
JTM, road
Christoph Kaeser
centerlines, building roofs) were measured at a BC3 analytical plotter. The images cover 2x2 blocks in characteristic
areas (generally hilly) with forests, agricultural fields, villages, water surfaces, but also modern urban centres with more
dense and high buildings.
Figure 2. Examples of VECTOR25 data for roads.
4 GENERAL STRATEGY AND METHODOLOGY
In a first step, we aim at detecting existing objects, while objects that do not exist anymore or new ones will be treated
later. Extraction of roads and buildings are treated by two separate researchers, but common input and other derived
data, are used by both and at a final stage due to the complementarity of road and building objects the results will be
fused. To increase the success rate and the reliability of the results we strongly rely on three aspects:
|. Use and fusion of multiple cues about the object existence and of existing information sources
The information provided by these sources should not be only complementary, but also redundant to account for
errors and incomplete results in the low-level image analysis. The basic cues used are DSM blobs, colour, texture,
shadows, and edges, while some secondary ones, like motion of vehicles on the roads, signalisation strips, context
between buildings and roads etc. will be also used. All these cues have associated relevant attributes.
2. Use of existing knowledge, "rules" and models
They are used to restrict the search space, treat each object subclass differently, check the plausibility of multiple
possible hypotheses, and derive reliability criteria. Using the known object class, e.g. highway, 1st class road etc.,
different possible value ranges for the attributes of these objects and certain rules can be used, e.g. road width,
horizontal and vertical curvature of roads, signalisation, no road can cross a highway at the same level etc. The
knowledge data base is automatically updated and refined using information gained from image analysis (only from
the reliably correct solutions), as well as at the stage of the manual editing and correction of the results at the L+T.
The road model includes geometric, radiometric, topological and contextual attributes.
3. Object-oriented approach in multiple object layers
A first, hierarchical, object layer can divide an object class, e.g. transportation network, to various subclasses, e.g.
road classes, railway lines, bridges, pathways etc. A second object layer can divide the objects in subclasses
according to the landcover and terrain relief, since these factors influence considerably the ability of object
reconstruction. E.g. roads will be divided in: inside forest, at forest border, open rural areas, urban areas (and
possibly city centres), while use of the terrain relief can be used in determining the plausibility of some object
parameters like horizontal and vertical curvature etc. Each subclass of the above two layers will have appropriate
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 465