Urban Building Inventory Development using Very High-Resolution Remote Sensing Data for Urban Risk Analysis

要旨

Disaster risk analysis is important not only to estimate the losses from future events but also to make recommendations for prevention, preparedness and response. Building inventories are essential for all types of disaster risk analysis models. With a slight difference in characterization of building types, all models require an estimate of number of buildings or total square footage (Eguchi et al., 2000). Land use information is very important for disaster risk analysis in urban areas. Traditional land surveying methods, such as field surveys, aerial photography, etc. are costly and time consuming. In order to rapidly derive detailed land use information in broad areas, it is necessary to use remote sensing techniques. In the last few years fine spatial resolution satellite imagery has become widely available. Such as the QuickBird satellite imagery and several new satellite sensors being developed are capable of generating imagery with spatial resolutions as fine as 0.6m in panchromatic mode and 2.8m in multi spectral mode. Many details such as buildings, roads, and other component elements of urban structures can be clearly identified from these high-resolution satellite images and this has opened a new window for urban land use information studies. A few studies, in the past years, were done to use satellite imagery to develop building and other infrastructures inventories for the selected areas. Yamazaki et al. (2000) investigated the capability of developing building inventory used for seismic risk analysis using satellite images from LANDSAT, IRS, JERS-1, ADEOS and IKONOS. They used principal component analysis and found this method as a possible solution to classify urban structures. However, Sande et al. (2003) proposed a segmentation and classification approach for IKONOS-2 imagery for land cover mapping to assist flood risk and flood damage assessment using object oriented image analysis technique. The paper presents a methodology for development of an up-to-date building inventory using information obtained from remote sensing data analysis and existing databases. The main objectives of the study were; i) analysis of very high-resolution satellite remote sensing data for deriving urban building features, and ii) development of an inventory for buildings for an urban area for disaster risk analysis. The outcomes from the object oriented classification methodology were very satisfactory with 90% accuracy in building classification and an overall accuracy of 85%. The shapes of the roads and water bodies were found as regular shapes. However, shapes of buildings were not well representative.