Comparison of MACLAW with several attribute selection methods for classification in hyperspectral images

Author

Blansche, Alexandre ; Wania, Annett ; Gancarski, Pierre

Author_Institution

LSIIT - AFD, Louis Pasteur Univ., Strasbourg

fYear

2006

fDate

Dec. 2006

Firstpage

231

Lastpage

236

Abstract

MACLAW is a clustering algorithm with local attribute weighting performed through cooperative coevolution. In this paper, we will compare the attributes weights obtained by MACLAW with several relevance indices for band selection on DAIS remotely sensed image which registers spectral object information in 79 bands of at least 2 nm. MACLAW capacities are also assessed by comparing its results to a supervised classification method for feature extraction proposed by the software ENVI (RSI Inc.). The MACLAW results are satisfying. Classification results are similar to the results of the supervised method. Supervised classification results are slightly improved using only a feature subset identified by MACLAW

Keywords

feature extraction; image classification; remote sensing; DAIS remotely sensed image; ENVI software; MACLAW; attribute selection methods; clustering algorithm; cooperative coevolution; feature extraction; hyperspectral images; spectral object information; Classification algorithms; Clustering algorithms; Clustering methods; Data mining; Feature extraction; Hyperspectral imaging; Hyperspectral sensors; Image analysis; Image classification; Remote sensing;

fLanguage

English

Publisher

ieee

Conference_Titel

Data Mining Workshops, 2006. ICDM Workshops 2006. Sixth IEEE International Conference on

Conference_Location

Hong Kong

Print_ISBN

0-7695-2702-7

Type

conf

DOI

10.1109/ICDMW.2006.47

Filename

4063630