Automatic three dimensional feature extraction: the case study of l'Aquila for collapse identification after April 2009

This paper illustrates an innovative methodology for post-earthquake collapsed building recognition, based on satellite-image classification methodologies and height variation information. Together, the techniques create a robust classification that seems to yield good results in this application field. In the first part of this study, two different feature extraction methodologies were compared, based respectively on pixel-based and object-oriented approaches. Then the classification results of the most accurate classification methodology, obtained on an eight band WorldView-2 monoscopic image, were completed with height variation information before and after the event. The height difference is calculated, comparing a photogrammetric DSM, obtained using a photogrammetric rigorous orbital model on some EROS-B 0.7 metre across-track stereopairs with a ‘roof model’ before the earthquake. Keywords: image classification, L’Aquila earthquake, collapses, EROS-B, World-View-2