ight
nt, and
(17)
ce <0,6.
when a
ne it as
another
metric.
(sm)
r which
pose all
J. la
ij
Hee Ju Park
line pair is corresponding to another, then it will have a high probability concerning{ SM 1, } and {SM 2, }. So both
results are consistent, and we select the pairs and continue with another set of pairs. We call this { P }. Then we discard
p less than a threshold, for example the lowest 5% .We receive a final set of matched line pairs.
4 RESULTS OF TEST AND DISCUSSION
We tested the proposed matching strategy with the Avenches dataset described in [Mason et al, 1994]. We checked the
results by human visual inspection with stereoscope. Figure 7 shows one example of our tests. We used the SE operator
[Heitger, 1995] as edge detection operator. Our epipolar imagery generation using the proposed epipolar line equation
was successful in all cases. Concerning point matching there was no blunder among 1050 matched point pairs when we
set the mask size for the Foerstner Interest Operator W=5. Concerning line matching 323 line pairs were matched and
no blunder was found. At least 2 pairs of lines are matched for each house. Generally the ridge line was matched. The
use of previous matched point sets supported line matching useful. Available matched points reduced the number of
candidate line pair in comparison to the case without previous matched points. So far the changing topology problem
was not solved though our method although there is no blunders in matched line pairs. For example sometimes the roof
line and the bottom line are seen as one line on one image, but has two separated corresponding lines in an other view.
In this case it can happen that the roof line is matched to the bottom line or to the roof line. We need more studies to
solve this remaining problem successfully.
5 CONCLUSIONS
In this paper we propose a new method which is useful for image matching, for DTM generation and house extraction.
The proposed new epipolar line equation which is determined by orientation parameters is easy to implement and useful
for epipolar line search and epipolar imagery generation. Concerning image matching two new ideas were proposed.
One is blunder suppression to matched point pairs based on positional relationship between possible matched point
pairs in point matching. The other one is the use of matched point sets as constraints for line matching in two ways -
one to define position constraints and the other to define a colour similarity constraint for overlapping flanking regions.
The matching strategy of combining line shape, flanking region similarity, positional and connectivity relationship
between corresponding lines and corresponding points, corresponding lines and its neighbouring corresponding lines
works fairly well with our test dataset. But there is still problem in solving the changing topology case. More thorough
evaluation for various imagery and and search for a solution for the changing topology case will be continued in future
studies.
REFERENCES
Foerstner, W., Guelch, E., 1987. A fast Operator for Detection and Precise Location of Distinct Points, Corners of
Circular Features, Proc. of ICFPDD, pp. 281-305.
Gruen, A., Wang, X., 1998. CC-Modeler: A topology generator for 3-D city models, International Archives of
Photogrammetry and Remote Sensing, Vol.32, Part 4, pp.188-196.
Gruen, A, Baltsavias E., Henricsson, O., 1997, Automatic Extraction of Man-Made Objects from Aerial and Space
Images (II), Birkhaeuser Verlag, Basel.
Hannah, M.J, 1989. A System for Digital Stereo Image Matching, PE & RS, Vol. 55, No. 12, pp. 1765-1770.
Heitger, F., 1995. Feature Detection using Suppression and Enhancement. Technical Report TR-163, Image Science
Lab, ETH-Zurich, Switzerland.
Henricsson O., 1996, Analysis of Image Structures using Color Attributes and Similarity Relations, PhD Thesis No.
11663, Swiss Federal Institute of Technology.
Mason, S., Baltsavias M., Stallman, D., 1994. High Precision Photogrammetric Data Set for Building Reconstruction
and Terrain Modelling, Internal Report, Institute for Photogrammetry and Geodesy, ETH, Zurich.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 703