Application of Adaboost based ensemble SVM on IKONOS image classification

Classification is one of the most important procedures in high-resolution remotely sensed image information extraction. This paper introduced Adaboost-SVM algorithm to IKONOS image classification. The classification performance of Adabost-SVM and single SVM were quantitatively analyzed and qualitatively evaluated. The results show that: In the case of small training samples, Adaboost-SVM outperforms single SVM in terms of classification accuracy greatly, and the training time of it is not too long. At the same time it can deal with the classes which are difficult for a single SVM to identify. In the case of big training samples, the generalization of Adaboost-SVM and single SVM are basically the same, but the training time of Adaboost-SVM is unbearable.