Centre de documentation
scientifique

Advanced polarimetric SAR data classification for cartographic information extraction / Manfred F. Buchroithner (31/05/1999)

Public contenu dans Remote sensing in the 21st century : economic and environmental applications / José Luis Casanova (2000) Titre : Advanced polarimetric SAR data classification for cartographic information extraction Type de document : Article/Communication Auteurs : Manfred F. Buchroithner, Auteur ; E. Kraetzschmar, Auteur ; M. Hellmann, Auteur Editeur : Lisse et Amsterdam : Balkema (August Aimé) Année de publication : 31/05/1999 Conférence : EARSeL 1999, 19th symposium, Remote sensing in the 21st century : economic and environmental applications 31/05/1999 02/06/1999 Valladolid Espagne Importance : pp 533 - 539 Note générale : Bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Acquisition d'image(s) et de donnée(s) [Termes IGN] analyse comparative [Termes IGN] classification [Termes IGN] classification floue [Termes IGN] extraction automatique [Termes IGN] image radar [Termes IGN] image SIR-C-X-SAR [Termes IGN] réseau neuronal artificiel [Termes IGN] valeur propre Résumé : (Auteur) Within the last decade several studies using polarimetric SAR data for bio-/geo-physical feature extraction have been reported and have significantly improved the understanding of polarimetric scattering mechanisms. In a comparative view, the approach presented here, which is based on Shane Cloud's Decomposition Theorem, seems to be most promising : the Entropy H and Angle Classification has been extended by using not only the parameters H and but also the first eigenvalue1. A big advantage of this approach is the high correlation between its results and the physical properties of different landsurface materials, which makes this recently adapted method well-suited for both supervised and automated landcover classification. It is also most useful for the derivation of topographic as well as thematic map information, the scale naturally depending on the sensor's spatial resolution. The well-known DLR Oberpfaffenhofen study site in southern Germany served as a testbed for the exploitation of spaceborne SIR-C/X-SAR data and of L-band data from DLR's airborne experimental ESAR. For validation purposes the classification results have been high-precision geocoded and compared to both ground truth and recent map data, also superimposing the digital national topographic geodata of Germany, ATKIS. The classification proved to be of extraordinary accuracy, as far as can be judged by visual inspection. An exact quantification is underway. The approach presented is being further developed and will - in contrast to the flat test-site used so far - be applied to data flown by a new airborne SAR sensor over mountainous terrain at the northern rim of the Bavarian Alps. First results in this respect look very promising. Numéro de notice : C1999-051 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Communication Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=65817