Segmentation and classification of polarimetric SAR data based on the KummerU distribution

Olivier Harant, Lionel Bombrun, Michel Gay, Renaud Fallourd, Emmanuel Trouvé, Florence Tupin

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Thinner spatial features can be observed from the high resolution of newly available spaceborne and airborne SAR images. Heterogeneous clutter models should be used to model the covariance matrix because each resolution cell contains only a small number of scatterers. In this paper, we focus on the use of a Fisher probability density function (pdf) to model the SAR clutter. First, the benefit of using such a pdf is exposed. Covari-ance matrix statistics are then analyzed in details. For a Fisher distributed texture, the covariance matrix follows a KummerU pdf. Asymptotic cases of this pdf are presented. Finally, the KummerU pdf is implemented in both hierarchical segmentation and classification algorithms. Segmentation and classification results are shown on both synthetic and real data.

Original languageEnglish
Title of host publicationProceedings of the 4th International Workshop on Science and Applications of SAR Polarimetry and Polarimetric Interferometry, PolInSAR 2009
Publication statusPublished - 1 Dec 2009
Externally publishedYes
Event4th International Workshop on Science and Applications of SAR Polarimetry and Polarimetric Interferometry, PolInSAR 2009 - Frascati, Italy
Duration: 26 Jan 200930 Jan 2009

Publication series

NameEuropean Space Agency, (Special Publication) ESA SP
Volume668 SP
ISSN (Print)0379-6566

Conference

Conference4th International Workshop on Science and Applications of SAR Polarimetry and Polarimetric Interferometry, PolInSAR 2009
Country/TerritoryItaly
CityFrascati
Period26/01/0930/01/09

Keywords

  • Classification
  • Fisher pdf
  • High resolution PolSAR data
  • KummerU pdf
  • Segmentation
  • Texture

Fingerprint

Dive into the research topics of 'Segmentation and classification of polarimetric SAR data based on the KummerU distribution'. Together they form a unique fingerprint.

Cite this