Kernel Extreme Learning Machine joint Spatial-spectral Information for Hyperspectral Image Classification

Abstract

Due to its fast learning speed and good generalization ability, extreme learning machine(ELM) has gained significant attention in machine learning and pattern recognition fields. However, when directly applied ELM to hyperspectral image classification (HSI), the accuracy is not high. In this paper, we propose a novel kernel ELM method, which joint spatial-spectral information together to investigate the performance of kernel ELM for HSI classification. In the proposed method, the spatial information are employed by extended morphological profiles. Experiments carried on two widely used hyperspectral datasets demonstrate that the proposed method outperform the SVM and kernel SVM methods. At the same time the cost of computation is much less than traditional methods.