3 Detection of Recently Constructed Multi-storey Buildings Using SPOT Panchromatic and Landsat TM/ETM+ Data Qiaofeng Zhang Department of Geosciences, Murray State University Murray, KY 42071-3311, USA E-mail: robin.zhang@murraystate.edu Jinfei Wang Department of Geography, The University of Western Ontario London, ON, N6A 5C2, Canada E-mail: ifwang@uwo.ca Peng Gong Department of Environmental Science, Policy, and Management University of California, Berkeley, CA 94720, USA Peijun Shi Institute of Resources Science, Beijing Normal University Beijing, 100875, China Abstract In this paper, methods were developed to detect recently constructed buildings in the city of Beijing using remotely sensed data. The city of Beijing is going through a rapid developing period, both in the suburbs and inside the old city. To detect the building development in and around the city, Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper (ETM+) and SPOT High Resolution Visible (HRV) panchromatic (PAN) images were collected in 1994 and 2001. The two data sources were merged using a spectral-preserving fusion method. An unsupervised classification was performed first on the merged dataset to separate urban built-up areas with non built-up areas. Buildings in the built-up areas were then extracted using a classification method based on high-pass filtered SPOT images. Our results show that over 83% of multi-storey buildings can be detected using the merged dataset. The comparison of buildings extracted for the two years indicated the locations of building development between 1994 and 2001. Introduction Remote sensing is especially valuable for areas that change rapidly, such as a developing city like Beijing, China. During the last decades, the city of Beijing has experienced rapid population growth, vast urban expansion and large-scale infrastructure improvement. Zhang et al. (2002) demonstrated that Landsat Thematic Mapper (TM) data were useful for locating urban expansion at the urban-rural fringe. Buildings are among the most important features in an urban environment. In Beijing, buildings have been constructed rapidly in new subdivisions in the suburbs, while inside the old city, some traditional one-storey houses were replaced with multi-storey apartments in redeveloped neighbourhoods. Up-to-date information about building locations and their attributes are needed for cartographic mapping, urban planning, and environmental and business applications using a Geographic Information System (GIS). Usually buildings are digitized manually from airborne orthophotographs or paper maps into a GIS system. However, it is very time-consuming and expensive to manually digitize specific features, such as buildings and roads, from remotely sensed data. Automated feature extraction is the key to future productivity of image-based spatial databases (Firestone et al., 1996). Spatial resolution is one of the most important considerations when extracting features, such as buildings from remotely sensed data. To illustrate the differences in spatial resolution for the purpose of building detection, the term “low resolution” is adopted in this paper to indicate spatial resolution of 30 m or coarser, “medium resolution” indicates spatial resolution of 5-30 m and “high resolution” refers to spatial resolution finer than 5 m. Using low-resolution remotely sensed data, such as data collected by Landsat Multi-Spectral Scanner (MSS, 80-meter resolution) and TM (30-meter resolution), Geocarto International, Vol. 20, No. 1, March 2005 E-mail: geocarto@geocarto.com Published by Geocarto International Centre, G.P.O. Box 4122, Hong Kong. Website: http://www.geocarto.com