Istanbul 2004 
  
  
investigations 
on 2 a short 
bed. The data 
he processing 
SETUP 
measurement 
n setup that 
s build up for 
jects using a 
fa generation 
in Section 4. 
s depicted in 
nulation or a 
use of a 3-d 
sure 2-1) and 
position and 
) get a high- 
re 2-3). The 
d we want to 
temporal and 
ling the laser 
of the laser 
gence of the 
= 
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
laser beam are used for convolving the high-resolution range 
image with the temporal waveform and weighting the high- 
resolution reflectance image with the spatial energy distribution 
(Figure 2-5). The spatial undersampling of the high resolution 
data is comparable to focusing the laser beam on the detector of 
the receiver (Figure 2-6). This kind of processing generates an 
intensity cube where the values correspond to the intensity of 
the signal. The simulation should give a realistic description of 
what we would expect from the measurement (Figure 2-7). 
To improve the estimate of the building's edge position the 
intensity cube is processed (data analysis). Processing starts 
with the matched filtering for signal preprocessing of the 
intensity cube to increase the accuracy of the range 
measurement and improve the detection rate (Figure 2-8). Then 
cach detected pulse is investigated for characteristic properties 
with pulse property extraction to get an enhanced description of 
the illuminated surface (Figure 2-9). By the use of the 
determined pulse properties a region based segmentation 
algorithm generates range and power images for region 
descriptions to handle the ambiguity of multiple reflections at 
the same spatial position (Figure 2-10). Analyzing the pulse 
power in the associated spatial neighborhood of the region 
edges estimates the edge position and edge orientation with sub 
pixel accuracy (Figure 2-11). Finally, for evaluation the 
received parametric boundary function of the region edges has 
to be compared with the 3-d object model (Figure 2-12). 
3. DATA GENERATION 
For simulating the temporal waveform of the backscattered 
pulses a scene model (1) and a sensor model (ii) is required. 
3.1 Scene modeling 
3.1.1 Scene representation 
For a 3-d scene representation, our simulation setup considers 
geometric and radiometric features of the illuminated surface in 
the form of 3-d object models with homogeneous surface 
reflectance. Figure 3 shows a 3-d object model with a 
homogeneous surface reflectance of the building in Figure | 
based on VRML (Virtual Reality Modeling Language). 
Figure 3. 3-d object model with homogeneous surface 
reflectance 
3.1.2 Sampling 
The object model with homogeneous surface reflectance is then 
ampled higher than the scanning grid we simulate and process, 
ecause with the higher spatial resolution we simulate the 
spatial distribution of the laser beam. Considering the position 
ind orientation of the sensor system we receive a high- 
resolution range image (Figure 4) and reflectance image. 
S 
109 
Depending on the predetermined position and orientation of the 
sensor system, various range images can be captured. 
  
Figure 4. High-resolution range image 
3.2 Sensor modeling 
The sensor modeling considers the specific properties of the 
sensing process: the position and orientation of the sensor, the 
laser pulse description, scanning and the receiver properties. 
3.2.1 Orientation 
To simulate various aspects a description of the extrinsic 
orientation of the laser scanning system with a GPS/INS system 
is used. 
3.2.2 Laser pulse description 
The transmitted laser pulse of the system is characterized by 
specific pulse properties (Jutzi et al., 2003a). We assume a 
radial symmetric Gaussian spatial distribution and a temporal 
exponential function as an approximation for the laser pulse. 
3.2.3 Scanning 
Depending on the scan pattern of the laser scanner system, the 
grid spacing of the scanning and the divergence of the laser 
beam a sub-area of the high-resolution range image is 
processed. By convolving the sub-area of the range image with 
the temporal waveform of the laser pulse, we receive a high- 
resolution intensity cube. Furthermore the corresponding sub- 
area of the high-resolution reflectance image is weighted with 
the spatial energy distribution of the laser beam (Gaussian 
distribution at the grid line +26) to take into account the amount 
of backscattered laser light for each reflectance value. Then we 
have a description of the backscattered laser beam with a higher 
spatial resolution than necessary for processing. 
3.24 Receiver 
By focusing the beam with its specific properties on the 
detector of the receiver, the spatial resolution is reduced and 
this is simulated with a spatial undersampling of the sub-areas. 
Finally we receive an intensity cube that considers the scanning 
width of the simulated laser scanner system and the temporal 
description of the backscattered signal. Because each 
reflectance value in the sub-area is processed by 
undersampling, multiple reflections arc considered with the 
backscattered signal.