| 基于小波特征融合的Hyperion影像降维方法 |
| |
| 引用本文： | 孙小芳.基于小波特征融合的Hyperion影像降维方法[J].测绘工程,2015(8):11-15. |
| |
| 作者姓名： | 孙小芳 |
| |
| 作者单位： | 闽江学院 地理科学系,福建 福州,350121 |
| |
| 基金项目： | 福建省科技厅重点项目，福建省教育厅资助项目，福建省测绘局资助项目 |
| |
| 摘 要： | 根据影像中地物光谱曲线的小波特征点确定地物识别的合适光谱分辨率,通过融合原先若干窄波段生成具有适合地物识别光谱分辨率的宽波段数据,达到降维高光谱数据的目的。文中对hyperion影像进行坏线和Smile效应去除,经过FLAASH大气校正,得到155个波段。对提取的八类地物的样本平均光谱进行DB4小波分解,计算小波细节系数方差;以小波细节系数信息熵作为特征点,得出不可渗透表面、居民地、水田、裸土4类地物识别适宜光谱分辨率为80nm,其余地物识别适宜光谱分辨率为160nm。以窄波段间的活跃度为指标进行融合,生成降维后的宽波段分别是21个波段和11个波段。8类地物在3尺度和4尺度下的分类结果说明降维影像能满足应用需求,提出的降维方法可行。
|
| 关 键 词： | Hyperion 小波特征 融合 降维 地物识别 |
The Hyperion image dimensional reduction method based on wavelet feature fusion |
| |
| Abstract: | According to spectrum wavelet features of image objects that determine a suitable spectral resolution about land use identification ,it achieves hyperspectral dimensional reduction ,through fusion a number of narrow-band to generate a wide-band spectral resolution with suitable for object recognition . After bad lines and smile effect are removed ,FLAASH atmospheric correction and 155 hyperion bands are used in the study .The average spectrums of eight feature sample extracted are decomposed by DB4 wavelet to calculate the wavelet detail coefficient variance and wavelet detail coefficient entropy as feature points .It is concluded that the suitable spectral resolution is 80nm to identify four kinds of objects such as impermeable surfaces ,residential areas ,paddy fields ,bare soil ,and the other four kinds of objects recognition suitable spectral resolution of 160 nm .The active degree taken as fusion indicator ,it is generated with 21 bands and 11 bands of wide-band spectrums after dimensional reduction by image fusion . Eight land classifications of dimensional reduction image can meet the application requirements on the 3-scale and 4-scale ,and the reduction method proves to be capable . |
| |
| Keywords: | Hyperion wavelet feature fusion dimensional reduction land use identification |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
