Skip to Main Content
A method is proposed for the classification of hyperspectral data with high spatial resolution by Support Vector Machine (SVM) with multiple kernels. The approach is an extension of previous sole-kernel classifiers by integrating spectral features with spatial or structural features for hyperspectral classification. Using Support Vector Machine (SVM) as the classifier, different multi-kernel SVM classifiers were constructed and tested using the Reflective Optics System Imaging Spectrometer (ROSIS) data with 115 bands to evaluate the performance and accuracy of the proposed multi-kernel classifier. The results show that integrating the spectral and morphological profile (MP) features, the multi-kernel SVM classifiers obtain more accurate classification results than sole-kernel SVM classifier. Moreover, when the multi-kernel SVM classifier is used, the combination the first seven principal components derived from Principal Components Analysis (PCA) and MP provided the highest accuracy (91.05%).