International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
ATKIS dataset leading to quality measures. Both tasks are
triggered by the ATKIS data being a valuable source of prior
information. The image analysis component interacts with the
knowledge-based component.
In our implementation the interactive GIS component runs on a
different computer than the automatic knowledge-based and
image analysis components which are implemented under
Linux. The automatic modules may be processed in batch mode,
communicating with the GIS component via data files.
3. AUTOMATIC IMAGE ANALYSIS MODULES
The verification of the topographic dataset is done by use of the
knowledge-based image interpretation system GeoAIDA
(Bückner et al. 2002, Müller et al. 2003) developed at the
Institut für Theoretische Nachrichtentechnik und
Informationsverarbeitung, Universität Hannover. GeoAIDA is
based on a semantic network that represents the scene to be
analyzed, it was designed for the interpretation of complete
scenes, and within this cooperation was modified and expanded
for GIS verification purposes.
3.1 Verification of Roads
The first step in road verification consists in defining a region
of interest for each road object from the database. More
precisely, a buffer around the vector representing the road axis
is defined, and the buffer width complies with the geometric
accuracy of the road object and the road width attribute in the
ATKIS database. If the latter value fails a plausibility test or is
not available at all, a predefined value is taken. Subsequently,
an appropriate road extraction algorithm to be executed in the
image domain of the buffer is selected. The selection includes a
control of the parameters considering the knowledge about the
given context region. We currently use the road extraction
algorithm presented in (Wiedemann and Ebner, 2000:
Wiedemann, 2002). This approach models roads as linear
objects in aerial or satellite imagery with a resolution of about 1
to 2 m. It should be noted that this algorithm was designed for
rural areas. Therefore, the following discussion and the results
for road extraction refer to rural areas only.
In the course of road extraction, initially extracted lines
(applying an approach given in (Steger, 1998)) are evaluated by
fuzzy values according to attributes like length, straightness,
constancy in width and constancy in grey values. The evaluation
is followed by a fusion of lines originating from different
channels. In our case we are using panchromatic imagery, but
the line extractor is applied twice: Firstly, using a bright line
model (line is brighter than the background) and secondly using
a dark line model (line is darker than the background). The last
step in road extraction as applied for verification is the grouping
of single lines in order to derive a topologically correct and
geometrically optimal path between seed points according to
some predefined criteria. The decision, if extracted and
evaluated lines are grouped into one road object, is taken
corresponding to a collinearity criterion (allowing a maximum
gap length and a maximum direction difference). All significant
and important parameters for road extraction can be set
individually. We adapted the described road extraction software
to our specific tasks, especially by applying individual
parameters for the given context areas and the extraction for
each road object separately. A road object from the ATKIS
database will be accepted, if the described road extraction in the
region of interest was successful and rejected otherwise.
738
The presented procedure is embedded in a two-stage graph-
based approach, which exploits the connection function of
roads and leads to a reduction of false alarms in the verification.
In the first phase the road extraction is applied using a strict
parameter control. leading to a relatively low degree of false-
positive road extraction, but also a high number of roads will be
rejected although being correct. For the second phase the latter
objects are examined regarding their connection function inside
the road network. It is assumed that accepted roads from the
first phase are connected via a shortest path in the network. All
rejected roads from the first phase fulfilling important network
connection tasks are checked again in a second phase, but with
a more tolerant parameter control for the road extraction.
For further information concerning the road verification refer to
(Gerke ct al., 2004). In (Gerke, 2004) a new approach to road
data verification is introduced. It incorporates local context
objects such as rows of trees to support the assessment of
ATKIS road objects and will be integrated into our system in
the near future.
3.2 Verification of Built-Up Areas
During the analysis two different complementary approaches
are followed. A textural analysis of the scene takes place to
decide between the classes agriculture, forest, industrial area
and settlement. A structural analysis of the image is carried out,
which searches for the most important items of settlement and
industrial areas, houses and industry halls.
3.2.1 Textural analysis The textural analysis uses a
segmentation algorithm described in (Gimel farb, 1997), it was
extended to multiresolution technique. Here the classes acreage
and grassland are combined to one unitary class agriculture,
because of similar texture. First, the algorithm has to learn the
properties of the classes with classified training regions, the
result of the learning process are four parameter files and an
evaluation matrix. The learning step determines the resolution
level on which a class has significant signatures. From the
evaluation matrix we derive in which resolution level a texture
can best be differentiated. The classification operator segments
the input image level by level with use of the 4 parameter files.
The resulting classification is a combination of all resolution
levels weighed with the evaluation matrix.
The learning step is a crucial part for the effectiveness and
correctness of the derived results. It is not necessary to train the
operator for each image, one time learning for a complete set of
images of a flight is sufficient. This step is preferably assisted
by a human operator, who manually defines and classifies
training regions for the desired classes. The borders of the
training regions can be taken from a GIS to speed up the
learning process. Since the fully automatic derivation of training
areas sometimes leads to training areas containing a mixture of
classes, the separability of the classes is not as good as with
manually defined areas. Another possibility to train the
classification operator is to take the borders and classes of a
GIS, this is only possible if the sample is large enough. The
advantage is a higher level of automation which is an important
feature of the system. For the results presented in Section 5
manually defined training areas were used.
3.2.2 Structural analysis The structural analysis is based on
finding buildings, which are modelled as complex structures
consisting of different parts. An illumination model is assumed,
Internc
thus a
zenith
knowk
genera
validat
one Oi
buildir
Since :
the gr
Additk
forest «
comple
(Rehrn
image.
hypoth
compa:
the last
to valk
of a sh
resultir
industr
3.2.3
structu
falsify
differei
is a h
classes.
using :
objects
evaluat
catalog
Both tH
a valid
verifiec
falsifie
decisio:
catalog
and ver
The m
quality
the int
operato
for the
The res
operato
the aut
workflc
automa
green
exclusn
ie. the
describ
accepta
indicate
accepte
rejected
when t
percent:
errors s
re. fal
rejected
by the }
