International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004
the relationship between the building template and the real
buildings as the similarity transformation (equations 1 and 2).
The same least squares matching process can be applied.
Through matching, the position and orientation of the building
template is refined. Figure 9 explains this template matching
process.
4. RESULTS AND DISCUSSION
The road detection algorithm proposed here was tested with Im
resolution IKONOS images. Figure 10 shows one example of
road centerline extraction on a typical highway. In the example,
road orientation was estimated automatically by applying the
line extraction algorithm. Once a user point was given, a series
of matching was applied rightwards and leftwards. It shows that
the least squares template matching we designed works.
Figure 11 shows the results of road extraction over the whole
test area. In this example, an operator provided a series of input
points and road orientation was estimated using the input points.
By the series of operator's input points, the centrelines from all
major roads are extracted. There are still many roads in the
figure that are not extracted. These are mostly small roads
without centerlines. Such roads can be extracted by measuring
start and end points of the road and by connecting the two
points with a straight line.
Figure 12 shows the results of building extraction from an
IKONOS image by the proposed method. The left image is the
results of the estimation of building rectangles by orientation
and position voting process. By clicking one point on building
roof, building rectangles are extracted automatically. No
manual edition was applied. As mentioned before, short sides of
buildings are not well detected. Sometimes, we observe false
detections of even long sides of buildings, typically for those
vertically aligned buildings in the left part of the image. This is
due to the line extraction failure. Our method works only if
there are valid line responses corresponding to the true building
sides. For most cases our method correctly found the
orientation and location of long sides of buildings. We argue
that producing such results by only one input point from a user
is very promising.
The right image in figure 12 is the result of template matching
using previously extracted building rectangle. First we apply
the automated extraction and manual editing. We then used this
result as building template. By clicking one point on building
roof, template matching is initiated and the results shown in the
image are obtained. These results are without manual rotation
and translation. Only the manual editing of scaling was
performed to adjust the different building length.
We can see that the orientation of some buildings is slightly
wrong. It seems the template matching we designed did not
work properly as we intended for such cases. This is the current
limitation of our algorithm. Nevertheless, by reusing the
previously extracted building rectangles we generated those
results almost automatically, which is also very promising.
There are other limitations of our building extraction algorithm.
It was not designed to work on small house buildings. Figure 13
illustrates this. When we click on small buildings, sometimes a
group of small buildings were extracted and sometimes
arbitrary rectangles. Also, a user should click a point within
building roof in order to get valid results. The right image in
figure 13 says that even if a point is clicked on a road, the
building extraction algorithm is still initiated and somehow
generate rectangles.
So far, we described the algorithm we developed to extract two
major map objects from Im resolution images. Due to the
limitation of page length, we could only briefly mention the
theory, procedures and performance of the algorithms
developed. We believe, nevertheless, that we have shown a fair
amount of information can be retrieved from only one single
image with very little manual intervention with carefully
devised line analysis and template matching.
REFERENCES
Baltsavias, E., S. Mason, and D. Stallmann, 1996 "Use of
DTMs/DSMs and Orthoimages to Support Building Extraction",
Automated Extraction of Man-made Objects from Aerial and
Space Images, (edited by A. Gruen et al), pp.199-210,
Birkhauser
Burns, J.B., A.R. Hanson and E.M. Riseman, 1986, "Extracting
Straight Lines", IEEE Trans. Pattern Analysis and Machine
Intellegence, 8(4):425-445
Cochran, S.D. and G. Medioni, 1992, "3-D Surface Description
from Binocular Stereo", IEEE Trans. on Pattern Analysis and
Machine Intelligence, 14(10):981-994
Doucette, P., P. Agouris, A. Stefanidis, and M. Musavi, 2001,
Self-organised clustering for road extraction in classified
imagery, ISPRS Journal of Photogrammetry and Remote
Sensing, 55(2001):347-358
Gruen, A. and H. Li, 1997, Semi-automatic linear feature
extraction by dynamic programming and LSB-snakes,
Photogrammtric Engineering and Remote Sensing, 63(8):985-
995
Gruen, A., P. Agouris and H. Li, 1995, Linear feature
extraction with dynamic programming and globally enforced
least squares matching, Automatic Extraction of Man-made
Objects from Aerial and Space Images (A. Gruen, O. Kuebler
and P. Agouris, editors), Birkhauser, Basel, pp. 83-94
Hu, X, Z. Zhang, J. Zhang, 2000, An approach of
semiautomated road extraction from aerial image based on
template matching and neural network, International Archives
of Photogrammetry and Remote Sensing, XXXIII(B3/2):994-
999
Huertas, A. and R. Nevatia, 1988, "Detecting Buildings in
Aerial Images", Computer Vision, Graphics, and Image
Processing, 41:131-152
Katartzis, A., H. Sahli, V. Pizurica and J. Cornelis, 2001, A
model-based approach to the automatic extraction of linear
features from airborne images, IEEE trans. on Geoscience and
Remote Sensing, 39(9):2073-2079
Kim, T. and J-P Muller, 1999, "Development of a Graph-based
Approach for Building Detection", Image and Vision
Computing, 17(1):3-14
Kim, T. and J-P Muller, 1998, “A Technique for 3D Building
Extraction”, Taejung Kim and Jan-Peter ‘Muller,
Photogrammetric Engineering and Remote Sensing, 64(9):923-
930
Kim, T. and J-P Muller, 1996, "Automated Urban Area
Building Extraction from High Resolution Stereo Imagery",
Taejung Kim and Jan-Peter Muller, Image and Vision
Computing, 14(2):115-130
Kim, T., S-R Park, M-G. Kim, S. Jung, K-O Kim, Tracking
road centerlines from high resolution remote sensing images by
least squares correlation matching, , Photogrammetric
Engineering and Remote Sensing, (in press), 2004
Intel
May
link
Phot
Shu!
Cue:
Com
Und
