摘要
高光谱遥感数据具有波段数目多、波段宽度窄、数据量庞大、波段间相关性高等特点,在一定程度上为图像的进一步处理和信息提取带来困难。为解决这一问题,在分析已有降维方法的基础上,提出了基于地物诊断性波谱吸收特征的高光谱遥感图像降维方法,将地物的诊断性吸收波谱特征区间作为一个独立的子空间进行处理,尽可能保留地物独有的吸收特征;在此基础上,进行子空间的特征提取和特征选择。为验证该方法的优越性,将其与传统的基于波谱区间的子空间划分方法进行分类对比,研究表明:基于该文方法降维后的图像分类精度更高,丰富了现有降维方法理论,具有一定的实用和推广价值。
Hyperspectral remote sensing data have some unique characteristics that with more channels, narrow-band width, the huge volume of data, the high correlation between bands, which have brought difficulties to a certain extent for the further image processing and information extraction. In order to solve this problem, on the basis of the analysis of the existing dimensional reduction methods, a new method of hyperspectral remote sensing image dimensional reduction based on diagnostic characteristic of spectral absorption was proposed. The method takes the diagnostic features of absorption spectrum range as a separate subspace, as far as possible to retain the unique features of the absorption characteristics. And then the information of sub-space can be extracted and the feature can be selected. To test the advantages of this method, a classification experiment was implemented, and the result showed that the classification accuracy of the dimensional reduced images based on this method is higher. The study enriches the existing dimensional reduced methods,and has a practical and promotional value.
出处
《地理与地理信息科学》
CSCD
北大核心
2009年第1期57-60,共4页
Geography and Geo-Information Science
基金
地球探测与信息技术教育部重点实验室基金项目(2004DTKF003)
成都理工大学青年基金项目(2006QJ17)
关键词
诊断性波谱特征
高光谱
遥感
降维
diagnostic characteristic of spectral
hyperspectral
remote sensing
dimension reduction
