An object oriented approach for quantitative assessment of building damage in urban areas using very high resolution images

Anne-Lise Chesnel (1) , Renaud Binet (1) , Lucien Wald (2)

1 DAM/DIF - DAM Île-de-France
2 CEP/Sophia

Very high resolution images are particularly well adapted to damage assessment methodology in urban area because on one hand it allows an analysis focused on the buildings solely through an object-oriented analysis, and on the other hand it permits a quantitative evaluation of this damage assessment using a visually established ground truth. We propose in this paper a method of damage assessment that uses these two benefits. First an original object oriented approach to register the images is presented. Then a simple and fast damage assessment method based on correlation is proposed and tested on the test-case of the earthquake of Bam in December 2003. Each building of a test-area is classified using Support Vector Machines. The performance of the method in each case is evaluated thanks to a manually constructed reference database that uses the European Macroseismic Scale. As a result, 75% of buildings are well classified among four different EMS damage grades.

Mots clés: earthquakes; geophysical signal processing; image resolution; remote sensing; support vector machines

Domaines: Sciences de l'environnement; Géographie; Traitement du signal et de l'image [eess.SP]; Traitement du signal et de l'image [eess.SP]

Anne-Lise Chesnel, Renaud Binet, Lucien Wald. An object oriented approach for quantitative assessment of building damage in urban areas using very high resolution images. IEEE Urban Remote Sensing Joint Event 2007, Apr 2007, Paris, France. pp.1-5, ⟨10.1109/URS.2007.371865⟩. ⟨hal-00464866⟩