'specially 
the laser 
o be per- 
specially 
ints 
rne laser 
thods for 
tion it 1s 
h is able 
nt points 
p. Bora 
ser beam 
penetrate 
ore reach 
ulse, i.e. 
ie ground 
ly reflec- 
ments do 
tion rate, 
ound sur- 
; in forest 
ich is the 
n rates of 
n conifer- 
'eas these 
ckermann 
or all sea- 
urements 
ch makes 
n surface 
the laser 
ng of the 
s the sep- 
face from 
flected by 
a Digital 
er points, 
pographic 
eliminate 
th can be 
'ound sur- 
he ground 
he ground 
algorithm. 
h provides 
ace out of 
the large 
ould work 
logical op- 
iles, good 
eason this 
] with the 
ning, first 
size is de- 
is deepest 
  
Figure 5: 3D-View of unfiltered laser data 
point is defined as a approximated ground point. The used 
band width corresponds to the measuring accuracy of the 
used laser sensor. In this way the window is moved by a 
certain step size over the whole data set. 
The main problem using this morphological operator is to 
define the optimal operator size, i.e. the size of the window 
that has to be examined. By the morphological processing 
not only vegetation, but also - especially in urban areas — 
large buildings have to be eliminated from the height data. 
Using a window entirely contained in a building’s outlines 
results in “roof-points“ which are falsely set as “ground 
points“. On the other hand the operator should be small 
enough to preserve smaller forms of the topographic sur- 
face. Using a too large window size raises the probability, 
that e.g. in areas of a rounded hilltops over a greater dis- 
tance no ground point can be found. Then it can happen, 
that in the preceding filtering and modeling step the roun- 
ded hilltop cannot be reconstructed, because there is a too 
large area with no approximation value in it, so that the 
hilltop could be cut off. In summary, the window size of 
the operator should be large enough to prevent the oper- 
ator from running into roofs or into the foliage, and on the 
other side the window should be small enough to preserve 
also smaller forms of the topographic ground surface. 
  
Figure 6: 3D-View of filtered laser data 
This motivated the application of a multi-level proced- 
ure. The morphological operator Openingis applied several 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
times with different window sizes starting with the smal- 
lest window size. Points which meet the condition to be 
within the band width higher than the deepest point in 
the applied window, get a certain weight depending on the 
window size; the larger the window size of the operator, the 
higher the weight for a laser point. After this step points 
which are likely to define the topographic surface posses a 
high, points which are likely to refer to object like build- 
ings or trees posses a low weight. All measured laser points 
with their certain weight are used in the last step of the 
process to compute the terrain surface. As filter algorithm 
is used a smoothing (filtering) by means of approximating 
cubic splines subject to given constraints. The constraints 
contribute directly to the smoothness by consideration of 
the standard deviation (measuring accuracy of the laser 
sensor) of the data to be smoothed [Fritsch 1991]. The 
result of this process is shown in figure 6. 
3.4 Derivation of further products 
An important application for airborne laser scanner sys- 
tems is the acquisition of three-dimensional databases for 
urban areas. Urban models consisting of a DTM and three- 
dimensional descriptions of buildings are e.g. required for 
tasks like 3D visualizations of urban scenes or for simu- 
lations like the propagation of electro-magnetic waves to 
plan optimal locations for transmitter stations. 
Utilizing the classification of laser points, described in the 
previous section, different descriptions of the surface of 
the sensed object surfaces can be derived. If all measured 
points are included a Digital Surface Model (DSM) can be 
calculated which represents the terrain surface including 
the surface of objects rising from the terrain like buildings. 
Besides the determination of laser points, which are likely 
to refer to the topographic terrain surface, the morpholo- 
gical processing which is described in the previous section 
can also be used to detect regions of the sensed surface, 
which are likely to define buildings. In order to recon- 
struct a 3D description of a building using the laser scanner 
  
  
E 
x 
© 
5 
  
UNE 
SETS