|
Classification of Hyperspectral Data Using a Band Grouping-based SVM Ensemble System
|
| B. Bigdeli *, F. Samadzadegan |
|
|
|
Abstract: (8236 Views) |
| With recent technological advances in remote sensing sensors and systems, very high-dimensional hyper spectral data are available for a better discrimination among different complex land-cover classes. However, the large number of spectral bands, but limited availability of training samples makes the problem of Hughes phenomenon or ‘curse of dimensionality’ in these data. Moreover, these high numbers of bands are usually highly correlated and the information provided can contain several data redundancies. Because of these complexities of hyperspectral data, traditional classification strategies have often limited performance in classification of hyperspectral imagery. Referring to the limitation of single classifier in these situations, classifier ensemble systems may have better performance than single classifiers especially on hyperspectral data with this high level of complexities. This paper presents a new method for classification of hyperspectral data based on a band grouping strategy through a SVM ensemble system. Proposed method used a band grouping process based on a mutual information (MI) strategy to split data into few band groups. After band grouping step, the proposed algorithm aims at benefiting from the capabilities of SVM as classification method. So, proposed method applied SVM on each band groups that produced in previous step. Finally, this paper applied Naive Bayes (NB) as a novel and robust classifier fusion method for combining classifiers in classifier ensemble system. NB is a precise classifier fusion based on the concepts of Bayesian theory. Experiments are applied on two common hyperspectral data. Obtained results show that the classification accuracy is significantly improved by the proposed method in comparison with standard SVM on all bands of hyperspectral data. Also, these results confirm the high performance of band grouping strategy in contrast to using of standard SVM on all feature space. |
|
| Keywords: Hyperspectral data, Support Vector Machine, Multiple Classifier System, Band Grouping |
|
|
Full-Text [PDF 383 kb]
(1801 Downloads)
|
|
Type of Study: Research |
Subject:
Photo&RS
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
