Peter Lohmann 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Figure 2: Processing chain 
e Upper left: Window of DSM (Black points = Missing data) 
Upper mid: Standard Deviation Image 
Upper right: Marked regions 
Mid left: Filled regions 
Mid mid: Automatic threshold 
Mid right: Best fitting rectangle 
Lower left: Regions marked to cut 
Lower mid: Removed non-ground regions 
Lower right: Filtered and interpolated DTM 
s. PRACTICAL RESULTS AND COMPARISON OF METHODS 
This paper deals with data, which has a spacing of 1 meter in x- and y- direction. In forest areas the change in elevation 
can be up to 20 meters or more at a point distance of 1 meter. The effect of these rapid differences in elevation has been 
investigated. Previous work in this area has only been carried out using a 10 m grid (Koch, 1999, Lohmann et al., 1999). 
The filtering of the DSM with help of DTMCOR (see chapter 3) is analysed on the basis of a test site. The area is 
partially covered with dense coniferous forest. 
Figure 3 shows the color coded image of the DSM including two roads crossing the area. The peaks represent the 
treestands. In the middle and in the western part of the area most of the vegetation can be found. The size of the area is 
1000 x 1000 meters. The minimum and maximum elevation in the DSM is 82,5 m and 139,6 m, respectively. 
The following filter settings were used: 
  
Tolerance factor /,,, 0,6 m 
Tolerance factor /,, 2,5m 
C(0) of covariance 0,7 
Mesh size 7m 
  
544 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 
  
  
  
25845( 
Figure 
The L 
height 
filtere 
(Figur 
points 
the fil 
filter 
influe 
that e: 
5.1 
A cor 
test s 
elevat 
of 15 
black 
The 1 
(9,88 
7749 
be no 
veget 
edges 
belon 
in the 
small 
The 1 
ramp 
gene 
non- 
The 
How 
build 
terra 
topo. 
impl 
curv 
with