一种基于小波PCA的高光谱图像特征提取新方法

Wavelet-Based Principal Components Analysis Feature Extraction Method for Hyperspectral Images

针对高光谱图像数据量大、数据维数高、光谱信息丰富的特点,提出一种基于小波分解的主成分分析(PCA)降维的特征提取新方法.该方法充分利用小波变换的优势,在光谱域内针对每个像元进行降维,既减少了数据量,还能保留光谱信号的差别;PCA方法充分利用像元间的相关性,保留不同类在相邻像元间的局部空间信息,弥补了空间域内小波变换类间可分性较差的问题.实验结果表明,小波分解与PCA相结合的特征提取方法能够有效地提高高光谱数据分类效率及分类精度.

A new wavelet-based principal components analysis （PCA） feature extraction method is proposed for hyperspectral dimensionality reduction. Wavelet decomposition can reduce hyperspectral data in the spectral domain for each pixel. This may not only reduce the data volume, but also preserve the distinction among spectral signatures that is useful for most pixel-based classifiers. PCA can provide more local spatial information among neighborhood class pixels than wavelet decomposition. Experimental results show that the hybrid method can prove the effectiveness and accuracy in classifying the hyperspectral data.