采用IDL小波工具优化高光谱影像光谱特征匹配分类

An Optimized Method for Hyperspectral Imagery Spectral Feature Fitting Classification by Using IDL Wavelet Toolkit

光谱特征匹配分类是常用的高光谱影像分类、识别地物的方法,针对高光谱影像提取植被盖度存在的问题,文章根据高光谱遥感影像处理的方法,采用EO-1卫星在广州市过境的Hyperion高光谱影像,以"广州南肺"万亩果园作为试验区,经过大气纠正——最小噪声分离变换(MNF)——最纯净像元指数计算(PPI)——提取植被的端元,以此作为研究区识别植被的参考样本,进行光谱特征匹配提取植被盖度。其中提出利用连续小波变换对参考端元的波谱曲线降噪的方法,旨在优化光谱特征匹配,以提高识别植被的精度。实验结果表明,这种辅助匹配的方法能有效提高识别植被的精度。

Spectral feature fitting（SFF） classification is a method commonly used in hyperspctral image classification and feature identification.To tackle the shortage of vegetation cover extracting from hyperspctral image,according to the processing method for hyperspectral remote sensing image,this study used the Hyperion hyperspectral image in Guangzhou captured by the EO-1 satellite and took Wan Mu Fruit Garden as experimental area which is known as the ＂Guangzhou South Lung＂.The experiment goes through the following process： atmospheric correction——minimum noise fraction transformation（MNF）—— pixel purity index calculation（PPI）——vegetation end-member extraction,which is taken as the identification reference sample for the spectral feature fitting.In this study we propose a noise reduction method for reference endmember＇s spectral curve by using continual wavelet transformation with the purpose of optimizing the spectral feature fitting to enhance the precision of vegetation recognition.The experimental result indicates that this auxiliary fitting method can effectively enhance the precision for vegetation recognition.