ibul 2004 
  
y a special 
iburg. The 
DTM. The 
ON image 
nainly used 
first step a 
tracting the 
oothed by a 
g algorithm 
After the 
ied until the 
ne maxima 
the results 
iin concept. 
lusions can 
ree heights 
nents. This 
erous 
stereo 
image 
correl. 
C US C^ 
3 U2 UJ 
AI I On 
1.81 
2.08 
0.30 
0.50 
0.51 
0.90 
y different 
r Scanner. 
correlation. 
nd Tree 
the ground 
arlier. The 
s and the 
. Also the 
ights. The 
red to the 
flight was 
only 0.2 to 
the stereo 
| with field 
ual to the 
he image 
tion to the 
ous stand, 
rger in the 
' same for 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
  
  
  
  
  
  
  
Figure 1: DSM of the old growth stand 22 - laser scanner 
the deciduous stand, about 0.8m larger in the coniferous stand 
and reaches the high value of 5.37m in the old growth stand. 
The relationship between ground height and laserscanner height 
in terms of the correlation coefficient is weaker if compared to 
other studies (Persson et al., 2003). The reason for this is that 
the height interpolated at the stem position was taken as the tree 
height instead of the largest difference between DSM and DTM 
closest to the stem postion. Most tree tops are not perpendicular 
to the stem positions, moreover, especially coniferous trees 
show a conical shape. Therefore, a small shift in planimetry 
implies a significant height difference and leads - possibly - to 
coefficients indicating the weaker statistical relationship (Table 
3). 
4.2 Measuring of profiles 
The mean differences in the old growth stand between the stereo 
measurements and the values of the laser DSM are 0.53m for 
the deciduous trees and 0.35m for the coniferous trees. This 
shows that the laser DSM lies beneath the stereo measurements 
which seems to be reasonable because the DMC flight took 
place one year after the laser scanner flight. However, the 
Figure 2: DSM of the old growth stand 22 - DMC 
difference is larger than the annual growth of approximately 0.2 
to 0.25m. The standard deviations are almost the same and are 
in the range of 0.8m and 0.9m. When comparing the differences 
between stereo measurements and image correlation it can be 
found that the mean value is three times larger than for the 
laserscanner DSM. This means that the DSM derived by image 
matching lies far beneath the stereo measurements. The data 
shows that the mean for the tops of deciduous trees has a similar 
value than the laser data. The main difference is caused by the 
mean values of tops of the coniferous trees. In these cases the 
DSM derived from DMC data is at average 5m beneath the 
stereo measurements. Figure 5 is showing the reason for this 
underestimation of coniferous trees in the old growth stand. The 
tops of large spruce trees which stick out far above the beech 
trees are cut by the algorithm. The beech trees are cut slightly 
by the algorithm, too. Also the standard deviations are much 
larger. They range between 2.34m for the deciduous and 3.83m 
for the coniferous tops. The mean differences between stereo 
measurements and laserscanner data in the coniferous stand are 
larger than in the other stand types. The highest differences are 
found for the edges of the crowns and the tree tops. However, 
the ground surface is better determined by 
  
  
  
  
3 
Figure 3: DSM of the deciduous stand 60 - laser scanner 
  
xi 
  
  
  
Figure 4: DSM of the deciduous stand 60 - DMC 
  
xS 
IEEE 
dec aRe 
i 
x N 
ME, 
  
a