24th Signal Processing and Communication Application Conference (SIU), Zonguldak, Turkey, 16 - 19 May 2016, pp.2065-2068
In this study, Support Vector Data Description (SVDD) method is used for the classification of hyperspectral data, and a new outlier elimination method is proposed to increase the classification rates of the SVDD. The method is applied to the Pavia University (Italy) hyperspectral data which was acquired by flights over the university campus and has ground truth. Well-known and commonly applied detection algorithms "Spectral Angle Mapper" (SAM), "Spectral Matched Filter" (SMF), "Constrained Energy Minimization" (CEM) and "Adaptive Coherence/Cosine Estimator" (ACE) were prefered as performance referance. Performance of the SVDD classifier and these four algorithms was compared. Although the SVDD could not perform as well as the others, its performance was significantly improved by the proposed outlier removal method.