VHR and GIS data fusion for building damage assessment
- Paper number
IAC-07-B1.6.06
- Author
Mr. Mattia Stasolla, University of Pavia, Italy
- Year
2007
- Abstract
The need of rapid mapping after natural disasters has nowadays become one of the main objectives for many European and World Programmes, such as GMES (Global Monitoring for Environment and Security) and GEOSS (Global Earth Observation System of Systems). The strict time requirements for retrieving useful information about hazard effects drives to the employment of satellite technologies, particularly very high resolution imagery, which can easily and rapidly provide a large amount of data. However, the operational methodology for extraction of damage maps is mainly based on manual photo-interpretation, that is very time and cost consuming. The main challenge is thus the definition of automated processing chains that must be solid and widely usable.\\ This paper is devoted to the development of an automatic technique for fast building damage assessment in order to support and/or improve relief operations in case of earthquakes. The test case is the city of Bam, Iran, which was destroyed by a violent earthquake in 2003: a couple (pre and post event) of pan-sharpened Quickbird images at 0.6m resolution were employed, integrated with ancillary GIS data, for building detection and damage assessment. VHR images have enormous potentialities, but they show a consistent spectral variability, even within pixels belonging to the same objects, and purely spectral classification is not sufficient, so that segmentation techniques are mandatory. Nevertheless they require a substantial pre-processing that very often cannot rely on the original image itself: a data fusion approach - that merges information retrieved from different sources - is the only way to improve results. \\ In this work, we employ ancillary GIS data, namely the position of existing buildings before the earthquake, providing an a priori knowledge about the scene that was used to initialize the proposed segmentation algorithm, mainly based on the watershed technique. After a first step focusing on the extraction of buildings, the final part of the processing chain is dedicated to the computation of a damage assessment index, that is then converted into a very easily readable map of the destroyed buildings.\\ Digital Globe images and GIS data have been provided respectively by MCEER and Prof. Yamazaki's team, whose help is gratefully acknowledged.
- Abstract document