Christoph Kaeser 
  
ATOMI - AUTOMATED RECONSTRUCTION OF TOPOGRAPHIC OBJECTS FROM AERIAL IMAGES 
USING VECTORIZED MAP INFORMATION 
Christoph EIDENBENZ *, Christoph KÁSER ', Emmanuel BALTSAVIAS gi 
* Swiss Federal Office of Topography 
Seftigenstr. 264, CH-3084 Wabern, Switzerland 
Tel.: +41-31-963 2311, Fax: +41-31-963 2459, {Christoph.Eidenbenz , Christoph.Kaeser } @1t.admin.ch 
** Institute of Geodesy and Photogrammetry, 
ETH-Honggerberg, CH-8093 Zurich, Switzerland 
Tel.: +41-1-633 3042, Fax: +41-1-633 1101, manos @ geod.baug.ethz.ch 
Working Group IC IV/III.2 
KEY WORDS: Data fusion, Mapping, Model-based Processing, Feature Extraction, Edge Matching, Object 
Reconstruction, Roads, Buildings. 
ABSTRACT 
The project ATOMI is a co-operation between the Federal Office of Topography (L+T) and ETH Zurich. The aim of 
ATOMI is to update vector data of road centerlines and building roof outlines from 1:25,000 maps, fitting it to the real 
landscape, improve the planimetric accuracy to 1m and derive height information (one representative height for each 
building) with 1-2 m accuracy. This update should be achieved by using image analysis techniques developed at ETH 
Zurich and digital aerial imagery. The whole procedure should be implemented as a stand-alone software package, able 
to import and export data as used at L+T. It should be quasi operational, fast, and the most important reliable. We do 
not aim at full automation (ca. 80% completeness is a plausible target). The paper will present in detail the aims, input 
data, strategy and general methods used in ATOMI. We will also present an overview of the results achieved up to now, 
and problems faced in building and road reconstruction. More detailed investigations of partial aspects for buildings and 
roads will be presented in two other papers at this Congress. 
1 INTRODUCTION 
The last period National Mapping Agencies (NMA), especially in Europe, try to generate digital landscape models that 
conform to reality and do not include any map generalisation effects. This process allows the integration of additional 
object classes and information compared to the ones in traditional topographic maps and also inclusion of the qu 
dimension. In addition, the demand for digital data, especially of buildings and roads, for various applications is 
increasing and the requirements for their accuracy, completeness and up-to-date status are also raised. NMAs have to 
face these new challenges, on top of production of their usual products, often in shorter revision periods, and with 
financial and personnel restrictions. To cope with higher product demands, increase the productivity and cut cost and 
time requirements, automation tools in the production should be employed. As aerial images are a major source of 
primary data, it is obvious that automated aerial image analysis can lead to significant benefits. 
Automated aerial image analysis has been a topic of active research for decades. This includes various aspects like 
DTM and DSM generation, aerial triangulation, interior and relative orientation, orthoimage generation and mosaicking 
etc. The last 5 years research has focussed on detection and reconstruction of topographic objects, especially buildings 
and roads, and generation of 3-D city models. A quite good overview of such research is given in Gruen et al. (1995, 
1997), Fórstner and Plümer (1997) and in special issues of some journals (ISPRS Journal - April 1998, CVIU - 
November 1998, PERS - July 1999). However, automated methods fail to provide good quality results and are very 
slow. Thus, semi-automated methods have been developed and often used successfully in various projects (Gülch et al., 
1999; Gruen and Wang, 1998; Lammi, 1998; Halla and Brenner, 1998; Vosselman and Veldhuis, 1999). Some of them 
have matured and are available as commercial systems. These methods however, require substantial human interaction, 
in most cases do not make use of a priori information (maps, plans etc.) and are rather focussed on limited areas and 
object modelling with a high level of detail. New airborne sensors like laser scanners (Lemmens et al., 1997; Hug and 
Wehr, 1997; Axelsson, 1999; Maas and Vosselman, 1999; Haala and Brenner, 1998; Stilla and Jurkiewicz, 1999; 
Chilton et al., 1999; Rieger et al., 1999) and to a less degree interferometric SAR (Burkhart et al., 1996; Henderson and 
  
462 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.