Full text: XVIIIth Congress (Part B3)

    
igital Image 
ric Week. 
otrie, Heft 9, 
respondence 
. & RS., Vol. 
986. 
3 for Point 
ogrammetric 
erg, 1995. 
1e Noise and 
Int. Arch. of 
1, 1994. 
Grouping for 
oc. 5" ICCV 
andemodelle 
sertation at 
<ommission, 
llierung mit 
1esis at the 
nsing at TU 
ogrammetric 
.): Optical 3- 
ann Verlag, 
al Elevation 
ogrammetry. 
Institute of 
10, 1995 
hniques. In: 
lotes of that 
try at Bonn 
stem SCOP 
ch. of P. & 
) 
übertragung 
ligitale Bild- 
y. Deutsche 
hen, 1992 
ree search 
Journal of 
50, Number 
992: Global 
Proceedings 
grammetry, 
92 PR-10 795 
AUTOMATIC DIGITAL TERRAIN MODEL GENERATION 
USING AERIAL IMAGES AND MAPS. 
Michel Roux, Jaime Lopez-Krahe*, Henri Maître 
Ecole Nationale Supérieure des Télécommunications 
Département IMA, 46 rue Barrault 
F 75634 Paris Cedex 13 
email : mroux@ima.enst.fr 
* now at Université Paris-8, Institut Intelligence Artificielle 
Commission III, Working Group 3. 
KEY WORDS: DEM/DTM, Vision, Cartography, 
Aerial Image Analysis, Urban Scene Interpretation. 
ABSTRACT: 
In this paper we present a new method for the ge- 
neration of Digital Terrain Models (DTM) on urban 
areas using simultaneously a stereoscopic pair of aerial 
images and a scanned map. This method relies on the 
information given by the map on the position of the 
road network. Roads and crossroads are places where 
information on terrain elevation can be extracted di- 
rectly from the aerial images. 
The effectiveness of this approach is demonstrated with 
complex imagery on a urban area containing a large 
variety of different urban block types. The results are 
evaluated with respect to ground truth data. 
1 Introduction 
The automated generation of 3D cartographic data- 
bases from aerial images has become a major field of 
interest, with many applications in cartography, na- 
vigation, telecommunications, urbanism, etc [1]. Ho- 
wever in the case of very complex scenes like dense 
urban areas the automatic process of the only aerial 
images leads to ambiguous solutions (many feasible in- 
terpretations can be made from the same elementary 
features). In order to help the image interpretation, 
existing maps provide useful information on the pre- 
sence, shape and localization of various features in the 
scene [2, 3]. 
In this paper we present a new method for the ge- 
neration of a Digital Terrain Model (DTM) of a urban 
scene using simultaneously a stereoscopic pair of aerial 
images and a scanned map. The map provides useful 
information on the places where the ground could be 
seen in the aerial images: mostly the road network and 
the surfaces without buildings (white surfaces of the 
map). We first describe briefly the analysis of the scan- 
ned map (road network extraction and urban block 
classification), then we depict in details our method 
for the generation of a DTM using the road network 
extracted from the scanned map. Results on a scene 
This work was sponsored by the CNET-Belfort under Re- 
search Contract 94 PE 7216. Special thanks to MMs. Rene 
Mignone and Rafı Deryeghiyan from the CNET-Belfort. 
Fig. 1- Portion of the scanned map. ©IGN 
The complete scene is 2km x 2km. 
in the suburb of Paris are presented and compared to 
ground truth data. 
2 Map Analysis 
For this application commercially available maps at 
the scale of 1:25000 have been purchased. They have 
been scanned at 300 dpi which gives a pixel approxi- 
matively equivalent to 2 meters on the ground. Be- 
cause of the importance of color for the representation 
of various cartographic features (main roads, contour 
lines, forest, ...), maps were scanned in full color with 
24 bit /pixel. 
Information on various structures can be extracted 
from the map. For our application, the road network 
and the urban blocks (regions delimited by roads) are 
the most interesting features. Word extraction is also a 
source of information on natural or man-made features 
present in the scene. In the remaining of this section, 
road network extraction and urban block classifica- 
tion are briefly described. More details on road and 
crossroad extraction algorithms can be found in [4]. 
697 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
   
   
  
   
    
   
  
  
  
  
   
    
     
    
   
  
  
    
    
    
    
   
    
    
    
    
      
   
    
    
   
   
   
   
   
    
  
    
    
    
     
   
   
	        
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.