Perceptron nonlinear blind source separation for feature extraction and image classification

Author

Boussema, Mohamed Rached ; Naceur, Mohamed Saber ; Elmannai, H.

Author_Institution

Lab. de Teledetection et Syst. d informations a Reference spatiale, Ecole Nat. D´´Ing. de Tunis, Tunis, Tunisia

fYear

2012

fDate

15-18 Oct. 2012

Firstpage

259

Lastpage

263

Abstract

In this paper, we aim to classify remotely sensed images for land characterisation. The major goal is approaching the natural nonlinear mixture for band observation and then dimension reduction by supervised classification. After that, an unsupervised method combining feature extraction and SVM in investigating to discriminate the land cover for SPOT 4 satellite image. In this technique, training data base are wavelet features that are extracted from a subset of sources.

Keywords

blind source separation; feature extraction; geophysical image processing; image classification; learning (artificial intelligence); multilayer perceptrons; remote sensing; support vector machines; terrain mapping; SPOT 4 satellite image; SVM; band observation; dimension reduction; feature extraction; image classification; land characterisation; land cover; natural nonlinear mixture; perceptron nonlinear blind source separation; remote sensing image; supervised classification; support vector machines; Accuracy; Bayesian methods; Feature extraction; Source separation; Support vector machine classification; Wavelet transforms; Bayesian Inference; Blind Source Separation; Feature extraction; Multilayer perceptron; Wavelet transform; image classification;

fLanguage

English

Publisher

ieee

Conference_Titel

Image Processing Theory, Tools and Applications (IPTA), 2012 3rd International Conference on

Conference_Location

Istanbul

ISSN

2154-5111

Print_ISBN

978-1-4673-2585-1

Type

conf

DOI

10.1109/IPTA.2012.6469537

Filename

6469537