International Archives of the Photogrammetry, Remote Sensing and Spatial Information 
real world object corresponds to an equivalent pixel in the 
image. 
  
  
  
Figure 3. Illustration of projection geometry with directions of 
used coordinate frames 
Thereafter, parameters of interior and exterior orientation of 
the camera are known. Assuming a known interior camera 
geometry, for an observed point P the following equation 
describes the relation between image space and object space 
(Figure 3): 
P" zC" e (P- C)" 2 C" 4g. gp" (1) 
Xp 
dt CPC oco vp |. 
"ip 
where 
= Xp, Vpare coordinates of P in image frame (i) 
=»  fmeans focal length of the camera, 
= (Cis the camera projection centre, 
= ®,0,¥ (Roll, Pitch, Heading) are Euler angles for 
the transformation from navigation frame (b) to 
geographic frame (g). 
J describes an appropriate rotation for 
transformation from frame (k) to frame (1). 
Thus it 1s possible to project an image to the digital map frame 
(m) in a world coordinate system and vice versa (Figure 4 and 
  
ul { : $, AE N | 
: 1 e d | 
va po gee SFE a 
Y up a i 4 z^ e 
TU. D aU lod Je . j — I 
^ - = a]. fie tte 
/ FIRE me 
7 "m. { p V - 
/ : ue 7 i à s 
/ Nd À 2 ps i d A 
7 A D i & 
J Na x x 1 i X 
7 i X | | 
/ 1 E cS i 3 
# 7 ry i 
/ hs + , 
/ 2 Eo eer N 
, LA à v 
/ t j n x. 
7 1 @ X 
/ B 
/ “ i AA À 
/ 2 - 1 
ur P X 
aUo eren PLA x 
tt / ^ Î x. 
~~ j x T ^ 
/ 
7 ; A SM 
Figure 4. Georeference of the recorded image within the digital 
map 
    
A 
Figure 5. Projection of the recorded image within the graphical 
representation of the digital map 
4.2 Image processing 
After deriving the relations between object space and image 
space, relevant traffic objects have to be detected in the image 
data. Thematic image processing is the most demanding part of 
the project. Different algorithms were developed and tested 
(e.g. Hetzheim et al 2003). 
The preferred approach (Ernst et al 2003) deals with a digital 
road map (produced by Navtech) of the relevant area. Using 
the a priori knowledge about roads and their parameters 
(width, lanes, and directions), images can be masked 
considering a margin depending on the accuracy of different 
data (etc. attitude data, maps). The roads are now the only 
image sections to be investigated. Histogram based estimators 
can additionally limit the car search region. In this manner 
search area and calculation time per image can be reduced 
significantly (Figure 6). 
A 
      
Figure 6. The expected road area based on the digital map 
information is masked out on the image 
To get accurate knowledge about the mapped roads all street 
segments of a sufficient area around the recorded region have 
to be tested regarding their intersection with the image. This 
information enables the aggregation of vehicle data from image 
sequences later on. 
The vehicle detection is done on the reduced image area from 
the previous phase. The pixel sizes of the expected vehicle 
classes are dynamically adapted to the current navigation data 
Sciences, Vol XXXV, Part Bl. Istanbul 2004 
    
   
   
  
  
  
   
    
  
  
  
   
    
    
   
   
    
   
    
    
     
   
    
   
  
   
    
  
    
    
    
   
   
    
  
    
   
   
  
     
Inte 
—— 
alt 
veh 
sing 
esti 
flig 
The 
ima 
env 
of 1 
sha 
The 
suit 
If 
furt 
cro 
the 
inf 
or 
Ev: 
for 
con 
  
ve 
Th 
m 
De 
frc 
ob