ranbul 2004 
e center; an 
Iternational, 
The United 
& Culture. 
| Zayed Al 
hief of Staff 
ng the UAE 
represented 
onal ground 
u Alainain, 
nse for his 
appreciated. 
AUTOMATIC CHANGE DETECTION OF GEOSPATIAL DATABASES 
BASED ON A DECISION-LEVEL FUSION TECHNIQUE 
F. Samadzadegan “, R. A. Abbaspour * *, M. Hahn b 
“Dept. of Surveying and Geomatics, Faculty of Engineering, University of Tehran, Tehran, Iran —(samadz, 
abaspour)@ut.ac.ir 
"Dept. of Geomatics, Computer Science and Mathematics, Stuttgart University of Applied Sciences, Stuttgart, Germany 
- m.hahn.fbv @ fht-stuttgart.de 
Commission ICWG IVIV 
KEY WORDS: Urban Geospatial Database, GIS, Information Fusion, Change Detection, Artificial Intelligence, Computer Vision 
ABSTRACT: 
Monitoring of changes in topographic urban geospatial databases (UGDs) is one of the main requirements of urban planners, urban 
decision-makers and managers. In this paper, an attempt has been made to design an automatic change detection (ACD) solution. 
The approach presented here takes advantage of fusion of descriptive and logical information represented on two levels. That is, 
information fusion to exploit the multi-level characteristics of the objects and logic fusion for enhancing the learning and with this 
the recognition abilities of the system. The proposed ACD process utilizes two types of data sets. Firstly, aerial and satellite images 
are used as data sources for the generation of a Digital Surface Model as well as for extracting textural and spectral information. 
Aerial images, because of their geometric stability, provide metric information, and satellite imageries, due to their wealth of spectral 
information, generate spectral data. The second type of data is the topographic urban geospatial databases. These data sets provide 
reference information, whereas the aerial and the satellite images serve to generate the more recent information and changes. The 
change detection process includes object identification, object extraction, object recognition and change detection phases. The 
proposed automatic change detection methodology was tested on a 1:1000 scale digital map of the city of Rasht in Iran by using of 
1:4000 aerial photos and a pan-sharpened IKONOS scene. Visual inspection of the obtained results demonstrates the high capability 
of the proposed method. 
I. INTRODUCTION the object definition. regarding its complexities, should be 
devised. c) Learning capabilities to modify the defects 
High and accelerating rate of urban changes, in particular in accompanied by the objects definition needs to be 
the developing countries, calls for an efficient and fast considered. This will enhance the recognition potentials 
technique for mapping the changes with the required when encountering new and undefined objects. 
accuracy for updating the existing topographic urban 
geospatial databases (Dowman, 1998; Armenakis et. al, Nevertheless, most of the existing methods for doing change 
2002; Kim and Muller. 2002). detection process are basically optimized to use information 
of one sensor imagery and in addition, by employ parametric 
However. in practice most of the processes for analysing the methods, object’s fuzziness behaviour and the possibility for 
changes are the manual methods like on-screen change introducing training potentials are basically neglected. In this 
detection that are time consuming and expert dependent. The paper, for the first time, an attempt has been made to design a 
availability of the new generation commercial one-meter system that integrates all above features in a total and 
resolution satellite imageries, due to their wealth of comprehensive automatic change detection (ACD) solution. 
information content, have opened a new era in the problem of The approach presented here takes advantage of the concept 
automatic change detection and consequently the UGDs of fusion in two levels of descriptive information and logical 
updating. Therefore. automatic change detection has been an planes. That is, information fusion to exploit the multi-level 
area of major interest in remote sensing and GIS for the last . characteristics of the objects and logic fusion for enhancing 
few years (Peled, 1993; Darvishzadeh, 2000; Dowman, 1998; the learning and hence recognition abilities of the system. 
Armenakis ct. al., 2002: Kim and Muller, 2002: Schiewe 
2002: Shi and Shibazaki. 2000). 2. PROPOSED METHODOLOGY 
Regarding the human potentials in object perception and As shown in the Figure 1. the proposed ACD process utilizes 
recognition. it seems that a comprehensive automatic UGD two categories of the data sets. In the first category, raster 
change detection system should be able to integrate the aerial and satellite images are used as data sources for the 
capabilities of: a) All available descriptive information generating a DSM (using stereo digitized aerial photographs), 
components of an object (such as: structure information, textural information (using both image and satellite images) 
textural information and spectral responses) must be and spectral information (using satellite image). That is, 
simultaneously taken into account. b) A fuzzy formulation for aerial images, because of their geometric stability, provide 
  
Corresponding author. 
489