nbul 2004
ion could
| only a
but also a
ult. These
ion of the
. Another
d quality
1SCS.
>r Linder
Amtlich
nssystem
ed fuzzy
rmation.
sing, 58
ng in an
tives for
reve, K
, Politics
1, 555 ~
urban
ery. in
Remote
Sensor
ler R-
Baden-
nmetric
ition of
try and
sensors
n GIS
egrated
iational
ng data
etry &
mated
y and
ensing
or the
A FRAMEWORK FOR ROAD CHANGE DETECTION AND MAP UPDATING
Qiaoping Zhang ^ *. and Isabelle Couloigner?
* Department of Geomatics Engineering University of Calgary — 2500 University Drive N.W. — Calgary, Alberta,
Canada — T2N IN4 — (qzhang, couloigner)@geomatics.ucalgary.ca
KEY WORDS: Road extraction, Change detection, Feature matching, Database updating, Spatio-temporal modeling
ABSTRACT:
The updating of road network databases is crucial to many Geographic Information System (GIS) applications such as navigation,
urban planning, etc. This paper presents a comprehensive framework for image-based road network updating, in which the following
three tasks are performed sequentially: road extraction from imagery, road change detection and updating, and spatio-temporal
modeling. For road extraction a multi-resolution analysis approach is used in combination with a novel road junction detection
method. The road change detection and updating is one of the typical issues in the map conflation field. Feature matching techniques
are applied to determine the changed and unchanged portions of the road network. A conflation step is then used to create an updated
road network in which the attributes will be transferred from the existing database to the new database based on the conjugate
features resulting from the feature matching step. For a pragmatic road updating system, a spatio-temporal modeler should be
encompassed to efficiently and effectively store and make use of both the updated and old databases. The proposed methodology has
been tested on updating the Canadian National Topographic DataBase (NTDB) based on road extraction from remotely-sensed
imagery.
1. INTRODUCTION
Keeping the road network database up-to-date is important to
many Geographic Information System (GIS) applications (e.g.,
traffic management, emergency handling, etc). In the geomatics
community, there are several options to update a road network
map, including ground surveying, vector map comparison,
image-based updating, etc. Image-based updating is based on
feature extraction from remotely-sensed imagery, which has
become even more important recently because of the high
spatial resolution (1-4 meters), fast orbit repeatability, rich
multi-spectrum information and stable, affordable acquisition
cost of satellite imagery [Hu & Tao, 2002].
Research on image-based geospatial change detection is rather
limited, at least compared to the body of work on object
extraction [Agouris et al, 2001] Among the methods
developed, we can mention particularly the work of Klang
(1998). He developed a method for detecting changes between
an existing road database and a satellite image. First, he used
the road database to initialize an optimization process using a
snake approach to correct road location. Then, he ran a line
following process using a statistical approach to detect new
roads, starting from the existing network. In Fortier et al.
(2001), the authors extend the above approach by using road
intersections. Road intersections improve matching between the
road database and the lines on the image. Hypotheses for new
road segments are generated from these line junctions. To avoid
the pitfalls in GIS updates that result in storing multiple slightly
different representations of an object that has actually remained
unchanged, Agouris et al. (2001) extend the model of
deformable contour models (snakes) to function in a differential
mode. They introduce a new framework to differentiate change
detection from the recording of numerous slightly different
versions of objects that may remain unchanged.
Although the snake model has been successfully applied in
feature extraction from imagery and change detection as well,
the review of the previous work shows that a snake model-
based approach to road change detection and updating has the
following problems:
1) It requires the initial position of every snake which makes
it of little use in finding new roads;
2) It is very sensitive to the initial position. An undesirable
initial snake will lead to an inaccurate result;
3) In a typical road change detection scenario, there is an
existing road database which provides an initial version of
the snake. Very often, however, the end points of the
existing road polyline will not be accurate enough to lead
the road nodes to the desired positions after snake
deformation;
4) It is also sensitive to the noise in the digital numbers along
the road lines, which is often the case due to the complex
radiometric background.
Usually a road updating processing will involve several issues,
such as 1) the improvement of the weak positional accuracy of
the existing road locations that remain unchanged; 2) the
updating of the changed road; 3) the detection of the new roads;
and 4) the transfer of attribute data from the previous version of
the road database. These issues are among the typical concerns
in map conflation research area. Map conflation is the process
of creating a new database based on two or more different
databases covering the same area [Cobb et al., 1998]. The new
database is superior to any single one in the whole, with high
accuracy, rich attribute information, up-to-date, etc.
This paper will present a framework for image-based road
change detection and map updating.
Corresponding author. Qiaoping Zhang — qzhang@geomatics.ucalgary.ca
729
