Three-dimensional multispectral classification and its application to early seismic damage assessment

The recent seismic events that have affected different areas of the planet point out the need to respond quickly to emergencies caused by these events; for this reason a semi-automatic method was experimented which would provide information concerning the damaged buildings in the study area through satellite images. In September 1997 an earthquake hit the study area, causing significant damage to a number of towns and villages in the area. This research attempts to: a) improve the digital surface model extracted from Ikonos satellite images covering an area of central Italy (Foligno, Umbria), through the pre-processing of raw images and manual editing b) study the best DSM models to improve the detection of height differences, mainly in urban areas, and evaluate the results of the land cover classification as further data for detecting changes in building distribution. DSM obtained by three-dimensional maps have been compared with DSM extracted directly from aerial stereo pairs using different approaches. The innovative aspect of the experiment is that of wanting to evaluate whether the combined use of multispectral classification techniques and altimetric aspects taken from high-resolution satellite images can make the recognition of changes to buildings affected by the earthquake more robust. The same methodology can be used also for updating existing medium-scale maps; in this case as well, the comparison of data regarding the same area but for different periods is important.