Advanced

Polarimetric SAR feature selection using a genetic algorithm

Hadadi, G; Sahebi, Mahmoudreza and Mansourian, Ali LU (2011) In Canadian Journal of Remote Sensing 37(1). p.27-36
Abstract
One of the main applications of polarimetric synthetic aperture radar (POLSAR) images is terrain classification. In this study, an algorithm is presented to extract optimized features of POLSAR images that are required for classification. The proposed algorithm involves three main steps: (i) feature extraction using decomposition algorithms, including both coherent and incoherent decomposition algorithms; (ii) feature selection using a combination of a genetic algorithm (GA) and an artificial neural network (ANN); and (iii) image classification using the neural network. The algorithm is applied to a data set composed of different land cover elements, such as manmade objects, oceans, forests, and vegetation. The classification results... (More)
One of the main applications of polarimetric synthetic aperture radar (POLSAR) images is terrain classification. In this study, an algorithm is presented to extract optimized features of POLSAR images that are required for classification. The proposed algorithm involves three main steps: (i) feature extraction using decomposition algorithms, including both coherent and incoherent decomposition algorithms; (ii) feature selection using a combination of a genetic algorithm (GA) and an artificial neural network (ANN); and (iii) image classification using the neural network. The algorithm is applied to a data set composed of different land cover elements, such as manmade objects, oceans, forests, and vegetation. The classification results obtained by the GA-based feature selection method exhibit the highest accuracy. The best features from the extracted features were identified and used in the classification based on the proposed algorithm. (Less)
Please use this url to cite or link to this publication:
author
publishing date
type
Contribution to specialist publication or newspaper
publication status
published
subject
categories
Popular Science
in
Canadian Journal of Remote Sensing
volume
37
issue
1
pages
27 - 36
publisher
NRC Research Press
external identifiers
  • scopus:84863336517
ISSN
1712-7971
language
English
LU publication?
no
id
ca8c6f4a-914b-4e22-b12b-04ffbfdc3490 (old id 4779445)
date added to LUP
2016-03-21 10:55:29
date last changed
2017-11-05 03:20:38
@misc{ca8c6f4a-914b-4e22-b12b-04ffbfdc3490,
  abstract     = {One of the main applications of polarimetric synthetic aperture radar (POLSAR) images is terrain classification. In this study, an algorithm is presented to extract optimized features of POLSAR images that are required for classification. The proposed algorithm involves three main steps: (i) feature extraction using decomposition algorithms, including both coherent and incoherent decomposition algorithms; (ii) feature selection using a combination of a genetic algorithm (GA) and an artificial neural network (ANN); and (iii) image classification using the neural network. The algorithm is applied to a data set composed of different land cover elements, such as manmade objects, oceans, forests, and vegetation. The classification results obtained by the GA-based feature selection method exhibit the highest accuracy. The best features from the extracted features were identified and used in the classification based on the proposed algorithm.},
  author       = {Hadadi, G and Sahebi, Mahmoudreza and Mansourian, Ali},
  issn         = {1712-7971},
  language     = {eng},
  number       = {1},
  pages        = {27--36},
  publisher    = {NRC Research Press},
  series       = {Canadian Journal of Remote Sensing},
  title        = {Polarimetric SAR feature selection using a genetic algorithm},
  volume       = {37},
  year         = {2011},
}