WIT Press


Endmember Extraction From Hyperspectral Images Using Self-organizing Neural Network

Price

Free (open access)

Paper DOI

10.2495/MIS000241

Volume

24

Pages

10

Published

2000

Size

818 kb

Author(s)

P.L. Aguilar, P. Martinez Cobo & R.M. Perez

Abstract

The present work exploits the possibility of using a Self-Organizing Neural Network to obtain the endmembers on hyperspectral images. The Self- Organizing neural network has the advantage that results by competitive procedures the class prototypes (endmembers). This ability can be used for determining the endmembers on hyperspectral images. We propose a neural network for processing the spectral information for each pixel. The neural model consists on Self-Organizing Neural Network. This net has an input neuron for each image channel. The output neuron number is related with the endmember number. Different distances and learning functions are used to obtain a better endmember extraction. The result discussion also includes the neighborhood function and the i

Keywords