|
|
|
Abstract
Using a Weighted Random Subspace Method to Improve the Classification Accuracy of Hyperspectral Images
Yasser Maghsoudi
M. Sc. Student
Faculty of Geodesy and Geomatics Eng., KN Toosi University of Technology, Iran
ymaghsoudi@yahoo.com
B. Mojaradi
Phd Student
Faculty of Geodesy and Geomatics Eng.
KN Toosi University of Technology, Iran
Mojaradi@alborz.kntu.ac.ir
M. J. Valadan Zoej
Assistant Professor
Faculty of Geodesy and Geomatics Eng.
KN Toosi University of Technology, Iran
valadanzouj@kntu.ac.ir
A. Alimohammadi
Assistant Professor
Faculty of Geodesy and Geomatics Eng.
KN Toosi University of Technology, Iran
alimoh_abb@yahoo.com
Abstract :
Classification of hyperspectral images is challenging. A very high dimensional input space requires an exponentially large amount of data to adequately and reliably represent the classes in that space. On the other hand with increase in the input dimensionality the hypothesis space grows exponentially which makes the classification performance highly unreliable. In this paper a new approach that is based on the concept of Random Subspace Method is proposed. In this method different feature subsets are taken randomly then they are passed on to the maximum likelihood classifiers. The weights are chosen in terms of the classifiers accuracies .The better the classification accuracy the higher the weight that is assigned to that classifier. We finally used a weighted majority scheme to combine the decisions of the classifiers. The experimental results show that the proposed method offers significant performance improvements with respect to the individual classifiers.
|
|
|