Registration of multitemporal landsat images: a local correlation approach

• Autorzy: Gusella L.
• Konferencja: 7th Bilateral Geodetic Meeting Italy-Poland (7; 22-24.05.2003; Bressanone, Italy)
• Języki publikacji: EN

Abstrakty

EN

The study of multitemporal dataset is a typical task of remote sensing analysis. The areas subjected to catastrophic events, such as earthquake, flooding, fire, can be empbasized by visual interpretation of images acquired on different periods. Seeing that the instant of a natural event is known, the acquisition and the subsequent interpretation of remotely sensed data can be focused on pre and post event image, calling master the first one, representing the normal situation, and slaves the post event images. Multiple techniques of enhancing can be used, like bands ratios, vegetation indexes comparison, principal components analysis and coherence calculation. Since these are essentially pixel-to-pixel image operations, a good registration between slaves (post-event images) and master (pre-event image) is the first result to achieve. The work presents a possible solution of registration problem in multitemporal data sets. The event studied is the earthquake occurred in the urban area surrounding the gulf of Izmit, in Marmara Sea (Turkey), August 17, 1999. Earthquake damage assessment by interpretation of remotely sensed data is carried out generally by visual recognition of differences in high-resolution images, and the main geometrical task is the georectification of datasets for mapping purpose. In this case, a very precise registration between image data sets is not essential to achieve significant results. While Landsat images have 30 meters of resolution, visual recognition of single, damaged buildings is not possible, but the study of multispectral bands can be helpful to find variation of suńace reflectance, caused by building collapse and presence of debris seattered on the roads. However, different position of satellite during acquisition (different view angles) and different sub-pixel features cause duplication/shifting of objects. This fault is particularly visible making color composites using the whole dataset. A local correlation filter can be used instead usual registration algorithms, trying to improve feature by feature matching and local coherence, which can be helpful to interpret variation on reflectance due to damage.

Słowa kluczowe

EN

satellite geodesy; hydrogeological catastrophe

• Wydawca: Wydział Geodezji i Kartografii Politechniki Warszawskiej
• Czasopismo: Reports on Geodesy
• Rocznik: 2003
• Tom: z. 2/65
• Strony: 269--278
• Opis fizyczny: Bibliogr. 4 poz., schem. fot.

Twórcy

autor

Gusella L.

• University of Bologna, Italy

• University of Bologna, Italy,

Bibliografia

• [1] Campbell, J. B., 1996, Introduction to Remote Sensing - Second Edition, Taylor & Francis, London
• [2] CEOS Disaster Management Support Group, 2002, The use of Earth Observingbn satellites for Hazards support; assessment & scenarios Dutch Interferometry Group, 2003, Doris User manual, http://www.geo.tudelft.nl/fmr/research/insar/dig/dig.html
• [3] Estrada M., Matsuoka M., Yamazaki F., 2000, "Damage Detection from the Space", EDM Technical Report No. 6, Report on the Kocaeli, Turkey Earthquake of August 17 Eurimage, 2002, Product & service specification
• [4] Mucciarelli M, Camassi R., Gallipoli M.R., 1999, Rapporto preliminare sulla missione in Turchia per lo studio del terremoto del 17 agosto 1999 (Preliminary report on mission in Turkey for studying earthquake in August 17,1999), http://emidius.mi.ingv.it/GNDT/T19990817/report.html