一种优化的高光谱图像特征提取方法

An Optimization Feature Extraction Method of Hyperspectral Images

众所周知,特征提取方法在降低计算复杂度和增加高光谱图像分类精确度方面十分有效,针对此,提出了一种优化的图像融合和迭代率波的特征提取方法。首先,将高光谱图像分割成多个相邻波段的子集;然后,通过均值法把每个子集中的各个波段融合在一起;最后,对融合后的波段进行递归滤波处理,获得分类的特征信息。结果表明,该方法展示出较高的分类精确度。

As is well known, feature extraction methods to reduce the computational complexity and in-crease the hyperspectral image classification accuracy is very effective, based on this, this paper put for-ward a kind of image fusion and iterative optimization rate wave feature extraction method. First, hyperspec-tral image was divided into a plurality of adjacent band subset. Then, each sub band was fused togetherby means of the mean method; finally, after the fusion of the band for recursive filtering processing, thecharacteristics of the classification of information was obtained. The results showed that the proposed meth-od showed higher classification accuracy.