TELKOMNIKA, Vol.13, No.3, September 2015, pp. 976~984
ISSN: 1693-6930, accredited A by DIKTI, Decree No: 58/DIKTI/Kep/2013
DOI: 10.12928/TELKOMNIKA.v13i3.1805 976
Received March 25, 2015; Revised June 14, 2015; Accepted June 30, 2015
RVM Classification of Hyperspectral Images Based on
Wavelet Kernel Non-negative Matrix Fractorization
Lin Bai
1
, Defa Hu*
2
, Meng Hui
3
, Yanbo Li
4
1,3,4
School of Electronics and Control Engineering, Chang'An University,
Xi’An 710064, Shaanxi, China
2
School of Computer and Information Engineering, Hunan University of Commerce,
Changsha 410205, Hunan, China
*Corresponding author, e-mail: hdf666@163.com
Abstract
A novel kernel framework for hyperspectral image classification based on relevance vector
machine (RVM) is presented in this paper. The new feature extraction algorithm based on Mexican hat
wavelet kernel non-negative matrix factorization (WKNMF) for hyperspectral remote sensing images is
proposed. By using the feature of multi-resolution analysis, the new method of nonlinear mapping
capability based on kernel NMF can be improved. The new classification framework of hyperspectral
image data combined with the novel WKNMF and RVM. The simulation experimental results on HYDICE
and AVIRIS data sets are both show that the classification accuracy of proposed method compared with
other experiment methods even can be improved over 10% in some cases and the classification precision
of small sample data area can be improved effectively.
Keywords: hyperspectral classification, non-negative matrix factorization, relevance vector machine,
kernel method
Copyright © 2015 Universitas Ahmad Dahlan. All rights reserved.
1. Introduction
It is well know that each material has its own specific electromagnetic radiation
spectrum characteristic. Using hyperspectral imagery (HSI) sensors, it is possible to recognize
materials and their physical states by measuring the spectrum of the electromagnetic energy
they reflect or emit. The spectral data which consist of hundreds of bands are usually acquired
by a remote platform, such as a satellite or an aircraft, and all bands are available at increasing
spatial and spectral resolutions. After 30 years of development, HSI technology has not only
been widely used in military, but also has been successfully applied in ocean remote sensing,
vegetation surveys, geological mapping, environmental monitoring and other civilian areas [1,
2].
Due to the state of art of sensor technology developed recently, an increasing number
of spectral bands have become available. Huge volumes of remote sensing images are
continuously being acquired and archived. This tremendous amount of high spectral resolution
imagery has dramatically increased the information source and increased the volume of imagery
stored [2, 3].
However, the excessive HSI data increase the difficulty of image processing and
analysis. Such as supervised classification of HSI images is a very challenging task due to the
generally unfavorable ratio between the large number of spectral bands and the limited number
of training samples available a priori, which results in the ‘Hughes phenomenon’. Without the
supports of new scientific concepts and novel technological methods, the existing large volumes
of data prohibit any systematic exploitation. This has led to great demands to develop new
concepts and methods to deal with large data sets [2-4].
Over the last years, many feature extraction techniques have been integrated in
processing chains intended for reduce the dimensionality of the data, thus mitigating the
Hughes phenomenon. These methods can be unsupervised or supervised. Classic
unsupervised techniques include principal component analysis (PCA), or independent
component analysis (ICA). Supervised approaches comprise discriminate analysis for feature