International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
3. STARIMAGER/TLS (THREE LINE SENSOR) 
STARIMAGER/TLS (Three Line Sensor) is an optical 
sensor for aerial survey. TLS is composed of three linear 
CCD arranged in parallel, and it can acquire three images 
of each direction (forward, nadir and backward) at the 
same time as Figure.1 shows. The sensor is oriented on 
an aircraft perpendicularly to flight direction and scans 
the ground plane. Therefore, a triple stereo image of a 
ground object can be acquired. 
TLS images are ultra high-resolution (spatial resolution: 
3cm) images, and it can acquire triple stereo images at 
the same time. Additionally, TLS images have more 
advantages such as less influence of occlusion and 
information from roofs as well as walls. 
  
  
   
Flight direction 
  
    
Backward 
Forward 
Figure. 1. Method of getting data with TLS 
4. EXISTING 3D DATA 
3D building texture data, which is assumed as existing 
data, are prepared beforehand. The data include not only 
roof textures but also wall textures. If the data do not 
exist, they can be generated from TLS images, which are 
acquired at different time period than the TLS data for 
change detection. 
Additionally, 3D data with high representation are 
required to refer textures in TLS images correctly in this 
method. For example, Figure.2 shows a result of 
overlaying low-level 3D data on a TLS image. As this 
figure shows, the low-level 3D data cannot be referred 
corresponding texture correctly. 
  
bei 
Figure. 2. Low-level 3D data 
  
Figure. 3. High-level 3D data 
Though many 3D data exist in urban areas, high-level 3D 
data such as TLS-scale 3D data exist seldom. Based on 
these, we have developed the snake-aided 3D data 
refinement method for this processing [1]. The 
followings show examples of the SNAKE-aided 3D data 
refinement (Figure.4-Figure.6). 
  
  
  
  
   
   
SM UE. : : pes, 
Figure. 5. Refined 3D data 
  
  
Figure. 6. Refined 3D data 
5. CHANGE DETECTION 
Therefore, high-level 3D data must be prepared such as 
Figure.3. Urban dense areas have various sizes of buildings. 
Therefore 'coarse change detection? and ‘fine change 
1325