基于改进的扩散平滑和RBM的高光谱图像分类

为了改善传统分类方法在高光谱遥感图像去噪和特征提取方面的不足,提出了一种基于改进的扩散平滑算法和RBM的方法.该方法使用自适应扩散系数,对相应的区域进行不同程度的扩散平滑,实现了对高光谱遥感图像的快速去噪;然后利用多层限制玻尔兹曼机构建DBN网络,实现对高光谱遥感图像的分类.实验表明,与传统的分类方法和DBN相比,该方法在高光谱图像地物分类精度上有所改善.

Feature extraction and classification of hyperspectral image based on improved diffusion smoothing and RBM

In order to improve the shortcomings of traditional classification methods in hyperspectral remote sensing image denoising and feature extraction, a new method based on improved diffusion smoothing algorithm and RBM model is proposed. The method uses the adaptive diffusion coefficient to the high spectral image denoising. To the corresponding regions, different degree of diffusion smoothing is adopted to realize the fast denoising of hyperspectral remote sensing image.Then, restricted Boltzmann machine is used to build DBN network to classify hyperspectral remote sensing images. The experimental results show that, compared with the traditional classification method and DBN, the proposed method obviously improved the classification accuracy of the remote sensing image.