ibul 2004 
by least 
>xtracted, 
oY à 1S 
(7) 
servation 
1, and the 
ng out of 
errors for 
alue, and 
removed 
matching 
:alculated 
tly. 
he above 
ly for all 
sfore, the 
1g. 
lines for 
Jar lines, 
condition 
the each 
rientation 
After the 
mage was 
sequently, 
sing these 
1) method 
performed 
of Koma 
ground or 
the result 
acquired 
However, 
gnized for 
formed by 
d surfaces 
(Kunii and 
e opening 
   
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004 
  
  
Figure 6. Result of line matching 
  
Figure 7. Result of surface extraction 
Furthermore, in order to perform 3D modelling for the Koma 
house, camera calibration for the first image and the last image 
were performed by combined adjustment (Chikatsu and Kunii, 
2002). Therefore, 3D data for the Koma house could be 
calculated efficiently. In addition, TIN for each surface was 
generated by using end points of the matched lines, and wire 
frame was reconstructed. In addition, texture mapping could be 
performed for each surface. Figure 8 shows wire frame model, 
and Figure 9 shows texture mapped model for the Koma house. 
Figure 9. Texture mapped model 
  
6. CONCLUSION 
This paper investigates mainly 3 issues regarding 3D modelling 
for historical structure by using image sequences: (1) efficient 
and robust line matching method using optical flow, trifocal 
tensor and epipolar matching (2) calculation of accurate trifocal 
tensor, (3) efficient remove of the useless lines during the line 
matching procedure and followings main results were obtained: 
+ Line matching was improved by similarity function with 
threshold value which acquired automatically. 
+ LMedS realized improvement of trifocal tensor and remove of 
the useless lines. 
Thus, it is concluded that the line matching method comprised 
optical flow, trifocal tensor and epipolar matching is useful 
method for 3D modelling of historical structure. However, there 
are still the following issues to be resolved before this method 
becomes operational. 
+ Efficient texture mapping. 
3D modelling for more complicated object. 
References from Journals: 
Canny, J., 1986. A computational approach to edge detection, 
IEEE Transaction on Pattern Analysis and Machine 
Intelligence, Vol.PAMI-8, No.6, pp.679-697. 
Gruen, A., 1985. Adaptive least squares correlation: a powerful 
image matching technique, South Africa Journal of 
Photogrammetry, Remote Sensing and Cartography, Vol.l4, 
No.3, pp.175-187. 
Ong, E., P. and Spann, M., 1999. Robust optical flow 
computation based on Least-Median-of-Squares regression, 
International Journal of Computer Vision, Vol.31, No.1, pp.51- 
82. 
References from Books: 
Rousseeuw, P., J. and Leroy, A., M., 1986. Robust Regression 
and Outlier Detection, Wiley 
References from Other Literature: 
Beardsley, P., Torr, P. and Zisserman, A., 1996. 3D model 
acquisition from extended image sequences, 4/h European 
Conference on Computer Vision, pp.683-695, Cambridge, UK. 
Chikatsu, H. and Kunii, Y., 2002. Performance evaluation of 
recent high resolution amateur cameras and application to 
modeling of historical structure, Internationals Archives of 
Photogrammetry and Remote Sensing, Vol XXXIV Part5, 
pp.337-341, Cortu, Greece. 
Kunii, Y. and Chikatsu, H., 2003. Efficient 3D modeling for 
historical structure by image sequences analysis. International 
Archives of Photogrammetry and Remote Sensing, Vol.XXXIV, 
Part5/W 12, pp.201-204, Ancona, Italy. 
Lucas, B., D. and Kanade, T., 1981 An iterative image 
registration technique with an stereo vision, DARPA Image 
Understanding Workshop, pp.121-130, Washington D.C., USA