Detection of buildings from Landsat-7 ETM+ and SPOT panchromatic data in Beijing, China

The city of Beijing is going through a rapid developing period, both in the suburbs and inside the old city. To detect the buildings developed in and around the city at present, Landsat-7 ETM+ and SPOT panchromatic imagery of 2001 were collected. To make use of higher spatial resolution of the SPOT panchromatic imagery and higher spectral resolution of the ETM+ imagery, the two data sources were fused. To reduce the confusion between urban built-up areas and non built-up areas, an unsupervised classification was performed first. Agricultural and large urban green and open areas such as parks and lakes were separated from other built-up areas. Major roads were mapped from ETM+ band8. Buildings were then extracted from the SPOT PAN imagery within the built-up areas using a set of proposed high-pass filters. Traditional Chinese-style houses, which are small in size and unidentifiable individually from the satellite imagery, were not detected. Our result shows that many medium-to-large buildings of regular arrangement can be mapped from the fused dataset. The buildings extracted from the satellite imagery can provide useful information for urban change detection and environmental management.