Classification of Hyperspectral Data Using An SVM-based AdaBoost Classifier System

Kurzfassung

Hyperspectral remote sensing data provide an efficient tool for land cover classifications in for complex information content.geographical. However, supervised classification of these data using conven-tional classification methods is difficult be-cause a sufficient number of training sam-ples is often not available for the wide range of spectral bands. In addition, spectral bands are usually highly correlated and contain data redundancies. To address these limitations, a multiple classification system based on Support Vector Machines (SVMs) is proposed to classify hyperspectral data. In this method, the data sets are first split into several band clusters based on the similarities between the contiguous bands. Next, the most useful bands in each cluster are determined using a band selection technique. Using an AdaBoost classification system, each band cluster is classified with an SVM, and a final decision is made by combining the outputs of each boosting step using the weights obtained from the AdaBoost processing. The obtained results are compared with a standard SVM applied on a feature set containing all the selected bands from different clusters and a standard SVM applied on all the spectral bands in terms of overall and single class accuracies and training times. In addition to this, the capability of AdaBoost to combine the decisions of component classifier is compared with majority voting classifier fusion method. The results demonstrate that the proposed method results leads to better classification performance, especially in for classes with greater complexity and fewer available training samples.

Verfügbare Versionen dieses Eintrags

• Classification of Hyperspectral Data Using An SVM-based AdaBoost Classifier System. (deposited 17 Okt 2013 07:29) [Gegenwärtig angezeigt]
  • An AdaBoost Ensemble Classifier System for Classifying Hyperspectral Data. (deposited 25 Feb 2014 16:17)