Radar clutter as an indicator for vegetation classification using a single dual polarimetric TSX-1 image

Marina Hetz, Dan G. Blumberg, Stanley R. Rotman, Roee Enbar, Ariel Benou

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

1 Scopus citations

Abstract

This paper investigates the ability to classify different vegetation types covering a semi-arid Mediterranean vegetated area using single-polarization SAR images from the TerraSAR-X (TSX-1) satellite. Based on statistical moments such as mean, standard deviation (STDEV), skewness and kurtosis, we found textural differences useful for the classification of the vegetation types. The research site, located near Zafit Hill, Israel, includes several different vegetation types such as pines and cypress forests with shrubs as an underlying vegetation layer (understory), olive orchards, eucalyptus clusters, natural grove areas with the presence of stones and smooth rocks, a wet cotton field, and smooth agricultural fields after harvest. In each vegetation type area, 40 equal polygons (10*10 pixels each) were identified on an optical image and defined on the TSX-1 image; 280 polygons in total were identified. The aforementioned statistical parameters were produced for each polygon, and co-variance matrices of combinations of two, three, or all four parameters together were produced. It was found that using the Mahalanobis distance of the mean-STDEV-skewness combination after applying a mode filter (5 *5 in size) was the best way to classify the vegetation types in the research area.

Original languageEnglish
Title of host publicationSAR Image Analysis, Modeling, and Techniques XIII
DOIs
StatePublished - 9 Dec 2013
EventSAR Image Analysis, Modeling, and Techniques XIII - Dresden, Germany
Duration: 24 Sep 201326 Sep 2013

Publication series

NameProceedings of SPIE - The International Society for Optical Engineering
Volume8891
ISSN (Print)0277-786X
ISSN (Electronic)1996-756X

Conference

ConferenceSAR Image Analysis, Modeling, and Techniques XIII
Country/TerritoryGermany
CityDresden
Period24/09/1326/09/13

Keywords

  • Co-variance matrix
  • Forest classification
  • SAR
  • TerraSAR-X
  • Vegetation

Fingerprint

Dive into the research topics of 'Radar clutter as an indicator for vegetation classification using a single dual polarimetric TSX-1 image'. Together they form a unique fingerprint.

Cite this