Improving Traditional Change Detection with DSM for Update Cartography in Urbanized Areas after Seismic Events

摘要

Earlier experimentations showed that some of the effects of a seismic event on buildings can be identified using change detection and multi-spectral classification algorithms [1]. The main change detection algorithms, are not always-effective where there are new buildings because there is the possibility that ground with the same radiometry of a roof is classified as a building and also the opposite is possible: a collapsed building after a seismic event may have the some radiometry of an undamaged building. The availability of stereopairs from high resolution satellite, allow to detect changes and transformations in the urbanized areas. Combine both techniques it's possible obviously only when stereoscopic and multispectral images are available. The situation of urban areas to be compared can be obtained easily from vectorial maps, containing the three-dimensional characteristics of the buildings. The result of the comparison is mainly influenced by the algorithms used for both the change detection and the DSM extraction. In the paper this approach was tested on L'Aquila earthquake site where a real high resolution satellite stereopair of the post event is still not available. For this reason a "pseudo stereopair" EROS-B was used to extract height information while multispectral classification was performed on a monoscopic WorldView2 image. To evaluate correctly height difference due to collapsed buildings a DSMs of the roofs was extracted from digital cartography excluding all the remaining parts of the territory. The work has to be considered as a "work in progress" but some methodological considerations on the advantage of combining the two techniques can be done.