i Gn MEME C uL a BESTE 
  
  
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
  
vehicle candidate. Also this vehicle candidate is fitted by 
rectangular polygon fitting algorithms. 
  
Figure 7 example of our algorithm of moving vehicle detection 
By using this polygon, correlation of area-based stereo 
matching algorithm is calculated to measure degree of same 
vehicle matching on different directional TLS images because a 
moving vehicle is at different same position on Nadir, Forward 
or Backward TLS image. Finally, moving vehicle candidates 
with their rectangular polygon and correlation of matching, 
which are agreed with moving vehicle model, are detected 
moving vehicles. 
6. PACKED VEHICLE DETECTION 
Parking/Idling Vehicle Detection is our algorithm to 
discriminate parked and signals wait vehicle types. From all of 
four vertexes of each stopped vehicle polygon to both side of 
road edge known from pre-processing stage, on space domain, 
perpendicular distances are solved. 
Edge Line Equation 
0 = AxO+By0+c 
IBIE-! meter 
  
Dr» |(Ax*By*C)|(A^2*B^2)^0.5 
Stopped Vehicle is parked vehicle if Dr <= Dt | 
Figure 8 Concept of Parked Vehicle Detection 
At least, one of all vertexes of each stopped vehicle polygon is 
agreed with Road . Edge-to-Parked-Vehicle ^ Distance 
Thresholding. Road Edge-to-Parked-Vehicle-Distance 
Thresholding is defined by the observation of on-street parking 
and signals waiting vehicles from the ground true survey. 
  
Figure 9 some result example of Parked/ Idling Vehicle 
Classification. Green polygon as Parked Vehicle and Red as 
Idling Vehicle 
7. SURMERRY 
Finally, our contribution presented new algorithms of vehicle 
detection by using new aerial image of Three Line Scanner. The 
algorithms perform the promising results. The improvement of 
algorithms is based on as below; 
l. Stopped Vehicle Detection algorithm by using multi TLS 
image has been developed. 
2. Moving Vehicle Detection algorithms by using multi TLS 
image has been created. 
3.  Parked/ Signals Waiting Vehicle Classification has been 
developed. 
All of our new proposed vehicle detection algorithms perform 
robustness with promising results. 
ACKNOWLEGNE 
Authors would like to specially thank for STARLABO Co.Ltd 
which distribute Three Line Scanner Images and all our 
members of Spatial Information Engineering Laboratory, 
Centre for Spatial Information Sciences, the University of 
Tokyo who encourage and support our works directly and 
indirectly. 
References 
1. Chellappa R., Zheng Q., Davis L., Lin C., Zhang X., and 
Rofield A., 1994, Site model based monitoring of aerial 
images , Image Understanding Workshop, pp.295-318 
  
  
    
    
   
   
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
   
   
      
      
      
   
   
  
    
   
   
   
    
  
  
  
   
    
  
  
  
  
  
  
  
  
   
   
  
  
   
Intern 
2 Ch 
Accur: 
(TLS) 
Semin 
Japan, 
lab.g.c 
(acces. 
3. Hin 
[Image 
for I 
Unvis: 
http:// 
/site. lj 
22 Ap: 
4. Hi 
Vehic! 
Techn 
Remo 
Germe 
2004) 
5. Na 
Mappi 
the 2 
Noven 
http:// 
ssed 2 
6. Shi 
from 
Remo 
7. Son 
TLS n 
the url 
2002, 
tokyo. 
ccesse 
8. Zha 
Aerial 
http:// 
(acces