Abstract
As a new machine learning approach, the extreme learning machine (ELM) has received much attention due to its good performance. However, when directly applied to hyperspectral image (HSI) classification, the recognition rate is low. This is because ELM does not use spatial information, which is very important for HSI classification. In view of this, this paper proposes a new framework for the spectral-spatial classification of HSI by combining ELM with loopy belief propagation (LBP). The original ELM is linear, and the nonlinear ELMs (or Kernel ELMs) are an improvement of linear ELM (LELM). However, based on lots of experiments and much analysis, it is found that the LELM is a better choice than nonlinear ELM for the spectral-spatial classification of HSI. Furthermore, we exploit the marginal probability distribution that uses the whole information in the HSI and learns such a distribution using the LBP. The proposed method not only maintains the fast speed of ELM, but also greatly improves the accuracy of classification. The experimental results in the well-known HSI data sets, Indian Pines, and Pavia University, demonstrate the good performance of the proposed method.
Highlights
The main goal of hyperspectral image (HSI) classification is to assign each pixel of the hypercube into a different class according to the spectral and spatial characteristics [1]
If we use the output of nonlinear ELM (NLELM) as the input for Markov random field (MRF) or loopy belief propagation (LBP), the structure of NLELM will seriously disturb the original information of HSI
All the experimental results are assessed by the overall accuracy (OA), average accuracy (AA), and kappa statistics (k) [35]
Summary
The main goal of HSI classification is to assign each pixel of the hypercube into a different class according to the spectral and spatial characteristics [1]. In [23], extended morphological profiles with ELM (EMP-ELM) were introduced for HSI classification These ELM-based spectral-spatial methods have produced reasonably good results, their performance can be further improved by using more effective spatial features, as discussed below. If we use the output of NLELM as the input for MRF or LBP, the structure of NLELM will seriously disturb the original information of HSI It cannot fully utilize the spectral information and spatial information of HSI and will degrade the classification accuracy. Since NLELM disturbs the features of the pixels in the same class, it causes the classification accuracy to be relatively low [29] To this end, LELM is used here with LBP for the spectral-spatial classification of HSI to achieve a higher classification accuracy.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
