论文部分内容阅读
针对单一分形维数不能表征高光谱数据光谱局部吸收特征的问题,提出了基于光谱概率测度的多重分形参数特征提取方法.基于光谱信息度量进行光谱概率测度的计算,基于配分函数法估计得到尺度函数;通过对尺度函数求导计算出Holder指数,并对尺度函数勒让德Legendre变换计算出多重分形谱;从多重分形谱和Holder指数之间的函数关系提取表征多重分形谱形态的4个多重分形谱参数作为光谱特征参数;并应用于基于最小距离准则的航空推扫式高光谱成像仪(PHI,Prush-broom Hyperspectral Imager)图像监督分类.结果证明:利用基于光谱概率测度的多重分形参数特征提取方法提取的光谱特征参数进行分类得到的总体分类正确率达94.789%,分类精度明显高于利用信息量维数和多重分形谱特征提取方法进行分类的结果,证明了基于光谱概率测度的多重分形参数特征提取方法提取的多重分形参数的有效性和可靠性.
Aiming at the problem that the single fractal dimension can not characterize the local absorption characteristics of hyperspectral data, a multi-fractal parameter feature extraction method based on spectral probability measure is proposed. The spectral probability measure is calculated based on the spectral information measure, and the scale function . The Holder index was calculated by deriving the scale function and the Multifractal spectrum was calculated by the Legendre transform of the scale function. The four multifractal features that characterize the multifractal spectrum were extracted from the functional relationship between the multifractal spectrum and the Holder index Spectral parameters as the spectral characteristic parameters, and applied to the image supervised classification of the PHS (Prush-broom Hyperspectral Imager) based on the minimum distance criterion.The results show that using the multi-fractal parameter feature extraction based on spectral probability measure The accuracy of the classification is 94.789%. The classification accuracy is obviously higher than that obtained by using the information dimension and multi-fractal spectral feature extraction methods. It is proved that the multifractal parameters based on spectral probability measure Feature extraction method to extract multiple Reliability and validity of shape parameters.