局部保护降维与高斯混合模型的高光谱图像分类

高光谱图像具有高谱间分辨率和低空间分辨率的特点,传统的分类方法难以得到较高的分类精度。针对该问题,该文研究了两种局部保护降维法——局部保护投影(LPP)和局部保护非负矩阵分离(LPNMF)对高光谱图像降维,这两种方法能很好地保护输入空间相邻像素间的局部特征。由于高光谱图像各类间的统计分布多为复杂的多模型结构,文中采用高斯混合模型(GMM)分类器对降维后的数据进行分类。实验结果表明,将局部保护降维与高斯混合模型相结合的高光谱图像分类算法不但在小样本情况下能有效地提高分类精度,而且在背景像素混合的情况下和高斯白噪声环境中具有一定的鲁棒性。

Hyperspectral image classification based on locality preserving dimension reduction and Gaussian mixture model

Hyperspectral images has the characteristic of high spectral resolution and low spatial resolution,the classification accuracy of traditional classification methods is unsatisfactory.For this problem,two locality preserving dimension reduction methods-locality preserving projection (LPP) and locality preserving nonnegative matrix factorization (LPNMF) are studied in this paper for preserving the local structure of neighboring samples.The statistical distribution of hyperspectral image classes is often a complicated multimodal structure.Therefore,classifiers based on Gaussian mixture model(GMM) are hence a natural fit for hyperspectral data.Experimental results show that the algorithms combined locality preserving dimension reduction methods with Gaussian mixture model can be effective in small training-samplessize situation,as well as provides outstanding robustness under background pixel-mixing conditions and strong Gaussian noise environments.