Full text: Proceedings, XXth congress (Part 3)

     
   
   
   
     
   
    
    
   
     
   
      
    
   
  
  
  
  
  
   
   
   
  
  
  
    
    
    
   
   
   
   
   
   
   
   
   
   
   
   
  
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
  
  
Internati 
quality if they are captured with large overlap and sidelap (such The 3-D 
2.1.3 Masking: Masking operation eliminates edges which as 80% and 60% ). Figure 4. shows a sample of a 3-D model of aerial 
does not overlap DSM mask. The remaining edges are used for with texture images cut and pasted from aerial photos. the Te-cr 
D I^ , re ~ 3 1 x 71 Oo ^, * SS 35. 
boundary extraction in the following processes shows à 
3 : ; ; al., 2004 
2.1.4 Hough Transformation: Hough transformation is 
executed to edge pixels in the masked edge image and straight 
lines are extracted as candidates of building boundaries. 
2.1.5 Line Segmentation: The extracted straight lines are 5% 7 A meth 
divided into line segments at the crossing points. After that, line X sd modellir 
segments that go through the DSM mask are eliminated. 3 data of 
  
2.1.6 Polygon Tracing: The line segments around a DSM 
mask are traced to form a polygon. If some parts around the à 
DSM mask have no line segments, DSM mask boundary 
segments are used. 
  
  
  
  
— 
Vnorm 
  
DSM DSM Masks M 
w^ “Face A 
  
  
  
  
| 
Among Img! — Img3, Imgl is selected for the texture image of Face 
A, since direction from Face A to projection center of Imgl have 
the least angle to the normal vector Vnorm. 
  
  
  
Figure 2. The schema of texture image selection 
vw uni m ret 
yo vr Mr 
wp der T er e 
sv se 
wri n we Lii 
ns dd ven a wy 
does tr n € prr 
nnm 
Saye boar br "e 
  
  
Masking 
   
  
   
Extracted Straight lines 
   
  
Hough 
Transformation 
  
  
Line Segmentation 
Line Segments 
| OO Or OO 
Building Polygons 
  
  
— 3. A 3. D mode! with texture images cut and PASTE] from 
aerial photos 
  
Figure 1. The schema of building extraction 
3. EXPERIMENTAL RESULTS 
2.2 3-D Modelling 
r bic | ane : We have applied the above algorithms to LIDAR data and 
After the building boundary extraction, 3-D-building model I$ aerial images of Shinjuku and Ginza area in Tokyo (50-cm 
created with these polygons such that each polygon forms the resolution). ALTM2033 of Optech is used as LIDAR sensor. A 
  
top of building with the elevation of corresponding DSM mask. prototype program has been developed as an application that 
Vertical walls from the top of building to the ground (DEM) are works on Windows2000-based personal computer. 
attached to the polygons. Texture images are pasted onto each Figure 4 shows the DSM image and the extracted polygons of 
building, including walls. The most appropriate texture image the test site of Shinjuku area. Figure 5 shows a sample of 3-D 
selected from aerial images according to geometry between city model of Shinjuku area. This figure shows that texture 
À building faces and external parameters of the photos as shown images on the wall area are pasted around the building, as well 
in Figure 2. External parameters of the photos can be directly as the top of the buildings. Figure 6 shows another sample of 3- 
measured by GPS/IMU systems, or calculated by aerial ^ [) city model of Ginza area. For both models, the processing 
triangulation. Aerial photos provides texture images of good time is about 2 hours for 1km” with a Pentium IV processor. 
278 
  
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.