CIP A 2003 XIX th International Symposium, 30 September - 04 October, 2003, Antalya, Turkey 
X 
Y 
Z 
RMSE 
1.291 
1.932 
2.468 
Table 2. RMSE of 3D ground coordinates of the check points in 
digital image (units are mm) 
The rectification was conducted before the stereo matching to 
decrease the parallax which could induce the disparities of the 
pixels of the same position. Right image is rectified to left 
image using ERDAS IMAGINE 8.4. Inverse transform equation 
is following equation : 
X'= 1.3026X + 0.172982 Y + 54.8103 (]) 
Y'= 0.0862574 X + 1.04515 Y - 79.7225 
where X, Y = rectified pixel position 
X\ Y— inverse transformed pixel position 
cross correlation method for stereo matching was performed. 
Then we can get the 3D point cloud pixel using the analytical 
space intersection with interior, exterior orientations and 
inverse transformed pixel position. 
3D model can be generated using Rapidfom2002 (Figure 5.). 
Figure 6. DSM extracted from the data acquired from laser 
scanning (left) and stereo photogrammetry (right) 
3.2 Comparison 
Dark area means an accordance of the two DSMs. White area is 
the region where the error is over 74.368 mm. Coefficient of 
correlation of these regions was represented low value. 
It is possible to classify the two regions (Figure 8.). One is for 
the shadows and the other is for the large parallax. The region 
affected by shadow is arised because the direction and intensity 
of sun light on getting the left image were different from those 
on getting the right image (region B). 
The region affected by parallax is arised because the parallax 
can be too large at some part, like A and B. It is possible to see 
the effect of parallax on the image matching from Table 3. 
Figure 5. Screen capture from Rapidform2002 program for 
building the digital elevation model using the data 
acquired from stereo photogrammetry 
3. COMPARISON WITH TWO RESULTS 
3.1 3D modeling 
It is possible to the difference between the two methods from 
the results. Figure 6 shows results from each method. 
DSM by S25 laser scanner is used as the reference data for the 
analysis of result by digital stereo photogrammetry. S25 ensures 
an accuracy that is the standard deviation of 0.21mm in 0.8-10m 
measuring distance. 
Heightss of both DSMs can be detected by overlapping the 
DSM from digital stereo photogrammetry on the DSM by laser 
scanning (Figure 7.). 
Figure 7. Height differences between laser scanner DSM and 
photogrammetry DSM (units are mm) 
Figure 8. Target region