developed. This topic is very complex and beyond the scope of
this paper.
Third, a closer inspection of the result (Figure 5) revels that some
extracted roads are not so smooth. Some of them even contain
branches (Figure 7). This problem is caused by the thinning
algorithm which can thin most parts of a road except corners
where small branches are found. Figure 7 shows the input
rectangle in the left and its corresponding output in the right.
Ideally, the output road should be a rectangular outline of one
pixel wide. However, it contains two tiny branches in two of the
four corners. The reason for this limitation is the order of iteration
during which the search is carried out in the order from the first
column of the top left corner down to the first column in the
bottom left, then the second column from the top left to the
bottom left, and so on until the bottom right. If the iteration order
is changed from right to left, then the output road will just be a
reversal of that shown in Figure 7. This problem can be overcome
through a further process of filtering to remove the branches
attached to a road.
Figure 7. A rectangular road block and its thinned output. The
output contains two tiny branches because of the sequence of
iteration. They will shift to the left if the order of iteration is
reversed.
S. SUMMARY AND CONCLUSIONS
In this study we developed an automatic method of extracting
road network information from hyperspatial resolution
multispectral IKONOS data in a densely populated urban area.
The introduction of spatial reasoning into the extraction is able to
overcome the problems commonly associated with existent
methods of road exaction, over which this proposed method based
on spatial reasoning has a few advantages. The first is that it is
highly flexible. No limit is imposed on the size of the area under
study. There is no assumption about the analyst's familiarity with
the research area. Second, it is comprehensive and covers the
entire process of road network extraction, from unsupervised
classification to create a binary image, to noise removal, road
segment joining and road thinning. This method has been
implemented in a flexible environment in which the user can
specify some parameters during the extraction, such as the
threshold for removing noise and the threshold for joining road
segments together. Finally, the method can be applied readily with
a wide range of image formats. Anyone who has access to an
unsupervised classification package can perform the extraction.
During undertaking of spatial reasoning the default file format is
the commonly used bitmap, which is widely available and
acceptable by most software packages. Apart from the IKONOS
images, this method also works with other types of satellite
imagery. Understandably, the finer the spatial resolution of the
image, the better roads show up on the image, and the more
accurate their extraction is. It must be pointed out, nevertheless,
that all extracted roads are restricted to the uniform width of only
one pixel.
References
Agouris, P., Stefanidis, A. and Gyftakis, S. 2001. Differential
snakes for change detection in road segments. Photogrammetric
Engineering and Remote Sensing, 67(12), pp. 1391-1399,
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
336
Barzohar, M. and Cooper, D. B. 1996. Automatic finding of
main roads in aerial images by using geometric-stochastic
models and estimation. /EEE Transactions on Pattern
Analysis and Machine Intelligence, 18(7), pp. 707 — 721.
Bhattacharya, U. and Parui, S. K., 1991. An improved
backpropagation neural network for detection of road-like
features in satellite imagery. International Journal of Remote
Sensing, 18(16), pp. 3379-3394.
Choy, S.S.0.. Choy, C. S.-T. and Siu, W.-C., 1995. New
single-pass algorithm for parallel thinning, Computer Vision
and Image Understanding, 62(1), pp. 69-77.
Fiset, R. and Cavayas, F. 1997. Automatic comparison of a
topographic map with remotely sensed images in a map
updating perspective: the road network case. International
Journal of Remote Sensing, 18(4), pp. 991-1006.
Geman, D. and Jedynak, B., 1996. An active testing model
for tracking roads in satellite images. IEEE Transactions on
Pattern Analysis and Machine Intelligence, 18(1), pp. 1-14.
Gruen, A. and Li, H., 1995. Road extraction from aerial and
satellite images by dynamic programming, /SPRS Journal of
Photogrammetry and Remote Sensing, 50(4), pp. 11-20.
Maillard, P. and Cavayas, F. 1989, Automatic map-guided
extraction of roads from SPOT imagery for cartographic
database updating. International Journal of Remote Sensing,
10(11), pp. 1775-1787.
Meisels, A. and Mintz, D., 1990. Symbolic reasoning in
object extraction. Computer Vision, Graphics, and Image
Processing, 52(3), pp. 447-459.
Nevatia, R. and Babu, K. R., 1980. Linear feature extraction
and description. Computer Graphics and Image Processing,
13(3), pp. 257-269.
Steger, C., 1996. Extraction of curved lines from images.
Proceedings of the [3th International Conference on
Pattern Recognition, 25-29 Aug. 1996, vol. 2, p. 251-255.
Stilla, U., 1995. Map-aided structural analysis of acrial
images. /SPRS Journal of Photogrammetry and Remote
Sensing, 50(4), pp. 3-10.
Tupin, F., Maitre, H., Mangin, J. F., Nicolas, J. M. and
Pechersky, E., 1998. Detection of linear features in SAR
images: application to road network extraction. [EEE
Transactions on Geoscience and Remote Sensing, 36(2), pp.
434 — 453.
Wang, J. Treitz, P. M., and Howarth, P. J., 1992. Road
network detection from SPOT imagery for updating
geographical information systems in the rural-urban fringe.
International Journal of Geographical Information Systems,
6(2), pp. 141-157.
Xiong, D., 2001. Automated road network extraction from
high resolution images, National Consortia on Remote
Sensing in Transportation: Safety, Hazards, and Disaster
Assessment, Albuquerque, New Mexico, May 2001, pp. 1-4.
KEY M
ABSTR
Automa
for buil:
or abse
classific
(SVMs)
techniqi
classific
classific
characte
multi-re
machin
and ach
same pl
classific
these te
suite of
The use
traditioi
fiducial
aerotria
is the
simple
features
despite
underst
Many a
from ac
research
central :
in 1995
2001 (E
man-me
Object
identific
and cla:
determi
research
tasks an
routines
