Context-based mapping of damaged buildings from high-resolution optical satellite images
2010 (English)In: International Journal of Remote Sensing, ISSN 0143-1161, E-ISSN 1366-5901, Vol. 31, no 13, 3411-3425 p.Article in journal (Refereed) Published
In the early stages of post-disaster response, a quick and reliable damage assessment map is essential.As time is a critical factor, automated damage mapping from remotely sensed images is the expected solution to drastically reduce data acquisition and computation time. Recently, high-resolution satellite images, such as QuickBird data, have been in high demand by damage assessment analysts and disaster management practitioners. However, the existing automated mapping approaches hardly accommodate such high-resolution data. This research aims at developing a new context-based automated approach for earthquake damage mapping from high-resolution satellite images. Relevant contextual information (including structure, shape, size, edge texture, spatial relations) describing the damage situation is formulated and up-scaled on a morphological scale-space. Speed optimization is achieved by parallel processing implementation. The developed approach was tested with two QuickBird images acquired on 26 June 2005 and 3 June 2008 over YingXiu town, Sichuan, China, which suffered the devastating 12 May 2008 earthquake. In comparison to the reference, the developed mapping approach could achieve over 80% accuracy for computation of the damage ratio. Future research is planned to test the approach on various disaster cases for both optical and radar images using a grid-computing platform towards a cost-effective damage mapping solution.
Place, publisher, year, edition, pages
2010. Vol. 31, no 13, 3411-3425 p.
IdentifiersURN: urn:nbn:se:kth:diva-29465DOI: 10.1080/01431161003727697ISI: 000280282000006ScopusID: 2-s2.0-77954849517OAI: oai:DiVA.org:kth-29465DiVA: diva2:395893
QC 201102082011-02-082011-02-022016-04-20Bibliographically approved