PolSAR covariance structure detection and classification based on the EM algorithm

Sudan Han, Pia Addabbo, Filippo Biondi, Carmine Clemente, Danilo Orlando, Giuseppe Ricci

Research output: Chapter in Book/Report/Conference proceedingConference contribution book

13 Downloads (Pure)

Abstract

This paper proposes a new method for clustering polarimetric synthetic aperture radar images by leveraging the peculiar characteristics of the polarimetric covariance matrix (PCM). Specifically, the feature used for classification is the PCM structure. To this end, the problem of detecting and classifying spatial variations in PCM structure is formulated as a multiple hypothesis test, where one null hypothesis and multiple alternative hypotheses are present. The estimation problems are solved by resorting to hidden random variables representative of covariance structure classes in conjunction with the expectation-maximization algorithm. These estimates are then used to form a penalized likelihood ratio test. The effectiveness of the proposed detection strategies is demonstrated on real polarimetric SAR data.
Original languageEnglish
Title of host publication2022 IEEE 9th International Workshop on Metrology for AeroSpace (MetroAeroSpace)
Place of PublicationPiscataway, NJ
PublisherIEEE
Pages254-258
Number of pages5
ISBN (Electronic)9781665410762
DOIs
Publication statusPublished - 18 Aug 2022
Event2022 IEEE 9th International Workshop on Metrology for AeroSpace (MetroAeroSpace) - Pisa, Italy
Duration: 27 Jun 202229 Jun 2022

Publication series

NameIEEE International Workshop on Metrology for AeroSpace (MetroAeroSpace)
PublisherIEEE
ISSN (Print)2575-7482
ISSN (Electronic)2575-7490

Conference

Conference2022 IEEE 9th International Workshop on Metrology for AeroSpace (MetroAeroSpace)
Abbreviated titleMetroAeroSpace 2022
Country/TerritoryItaly
CityPisa
Period27/06/2229/06/22

Keywords

  • expectation-maximization
  • polarimetric radar
  • radar
  • synthetic aperture radar
  • phase change materials
  • urban areas
  • scattering
  • metrology
  • radar tracking
  • radar polarimetry
  • polarimetric synthetic aperture radar (PolSAR)

Fingerprint

Dive into the research topics of 'PolSAR covariance structure detection and classification based on the EM algorithm'. Together they form a unique fingerprint.

Cite this