building), the overall Fisher-test is often rejected, while all 
tests of individual correspondences are accepted. The lack of 
"perfect" correspondences reduces the precision that can be 
reached. To check this precision, relative orientation of both 
pairs was determined through least-squares adjustment of 20 
manually measured points. The deviations in the orientation 
of the detected epipole are below 2 degree (Table 2). As 
expected the computational burden of the proposed method is 
considerable. For the first pair computational time is in the 
order of 10 minutes on a modern PC. 
  
  
  
  
  
  
Figure 7: The detected correspondences of both pairs. 
5. CONCLUSIONS 
A new method for automatic relative orientation has been 
presented. It relies on the extraction of straight image lines 
and their vanishing point labelling. Vanishing point detection 
is a crucial step in the procedure that results in an ambiguous 
orientation of the images relative to the building. The epipole 
detection shows many similarities with the vanishing point 
—232— 
detection. Both are based on clustering of rigorous statistical 
tests and adjustment of constraints on the observations. 
Experiments show that relative orientation can be detected 
successfully between two images with an angle of 65 degree 
between the optical axes (see section 4, first image pair), 
while the difference in orientation with a manually 
determined relative position vector was less than 2 degree. 
The proposed method can be regarded as a first step towards 
automated reconstruction because the model coordinates of 
the corresponding points and the parameters of the plane in 
which they recede become available as a by-product. 
REFERENCES 
Baumberg, A., 2000. Reliable feature matching across widely 
separated views, Computer Vision & Pattern Recognition 
(CVPR) 2000, pp. 774-781. 
Fischler, M.A. and Bolles, R.C., 1981. Random sample consensus: a 
paradigm for model fitting with applications to image analysis 
and automated cartography. Communications of the ACM, Vol. 
24(6), pp. 381-395. 
Fórstner, W., 2000. New orientation procedures. International 
Archives of Photogrammetry and Remote Sensing, Vol. 33 part 
3, pp. 297-304. 
Fórstner, W. and Gülch, E., 1999. Automatic orientation and 
recognition in highly structured scenes. J. of Photo-grammetry 
and Remote Sensing, Vol. 54, pp. 23-34. 
Habib, A. and Kelley, D., 2001. Automatic relative orientation of 
large scale imager over urban areas using modified iterated 
Hough transform. J. of Photogrammetry and Remote Sensing, 
Vol. 56, pp. 29-41. 
Heipke, C., 1997. Automation of interior, relative, and absolute 
orientation. ISPRS J. of Photogrammetry and Remote Sensing, 
Vol. 52, pp. 1-19. 
Matas, J., Urban, M. and Pajdla, T., 2001. Unifying view for wide- 
baseline stereo matching. In: B. Likar (Editor), Computer Vision 
Winter Workshop. Slovenian Pattern Recorgnition Society, pp. 
214-222. 
Pollefeys, M., Koch, R., Vergauwen, M. and Van Gool, L., 2000. 
Automated reconstruction of 3D scenes from sequences of 
images. ISPRS J. of Photogrammetry and Remote Sensing, Vol. 
55(4), pp. 251-267. 
Pritchett, P. and Zisserman, A., 1998. Wide baseline stereo matching, 
Sixth International Conference on Computer Vision, pp. 754 - 
760. 
Teller, S., 2001. Scalable, controlled imagery capture in urban 
environments. Report 825, MIT Laboratory for Computer 
Science. 
Tuytelaars, T. and Van Gool, L., 2000. Wide Baseline Stereo 
Matching based on Local, Affinely Invariant Regions, British 
Machine Vision Conference BMVC'2000, pp. 412-425. 
van den Heuvel, F.A., 1998. Vanishing point detection for 
architectural photogrammetry. In: H. Chikatsu and E. Shimizu 
(Editors). International Archives of Photogrammetry and 
Remote Sensing, Vol. 32 part 5, pp. 652-659. 
van den Heuvel, F.A., 1999. Estimation of interior orientation 
parameters from constraints on line measurements in a single 
image. In: P. Patias (Editor). International Archives of 
Photogrammetry and Remote Sensing, Vol. 32 part SW11, pp. 
81-88. 
van den Heuvel, F.A., 2001. Object reconstruction from a single 
architectural image taken with an uncalibrated camera. 
Photogrammetrie, Fernerkundung, Geoinformation, Vol. 
2001(4), pp. 247-260. 
Wang, Y., 1998. Principles and applications of structural image 
matching. ISPRS J. of Photogrammetry and Remote Sensing, 
Vol. 53, pp. 154-165. 
  
  
  
FT" La a 
AY (a) LL] mM peed