HYPERSPECTRAL SUPERVISED CLASSIFICATION USING MEAN FILTERING BASED KERNEL EXTREME LEARNING MACHINE

Abstract

Extreme learning machine (ELM) is a single hidden layer neural network, and has a faster learning speed than the traditional neural networks. However, the kernel ELM (KELM) is increasingly gaining attentions in hyperspectral image(HSI) classification, due to its robustness. The widely used RBF kernel in KELM has achieved promising classification performance in hyperspectral image(HSI) classification, but the underlying data structure of the hyperspectral image is not taken into consideration. In this paper, we propose a novel spectral-spatial KELM method by incorporating the mean filtering (MF) kernel into the KELM model, which can properly compute the average of the spatial neighboring pixels in the kernel space. Experimental results on real hyperspectral datasets show that, the proposed method outperforms others kernel based KELM method, in terms of both computational efficiency and classification accuracy in HSI supervised classification.