摘要
基于城市典型地物反射波谱,采用相关系数聚类法寻找合适表达城市地物特征的波谱分辨率,得到25、50、100通道三种波谱分辨率;从信息量、可分性两个角度进行验证,表明25个通道划分方法保留了地物间的基本可分性,而后两种划分方法优于普通高光谱通道。研究为高光谱城市遥感数据降维提供了一种途径。
Based on the field reflectance spectra of urban typical objects, this study tries to find the fine spectral--resolution for detailed mapping of urban objects based on their spectral signal. Correlation coefficient matrix and clustering is applied to obtain potential spectral windows. Finally, the range from 400 to 900nm is divided to 25, 50 and 100 bands, three kinds of spectral resolution. Entropy evaluation and classification using neural network show that separation can be mostly guaranteed on new spectrum resampled by 25 bands. It is also find that the divisions of 50 and 100 bands are better than the usual high spectral bands. So it proves the bands--division method is rational. As a result, the division of 25 bands is suggested to reduce spectral dimension of hyperspectral urban remote sensing.
出处
《遥感信息》
CSCD
2008年第2期21-24,50,共5页
Remote Sensing Information
关键词
城市典型地物
反射波谱
波谱分辨率
urban typical targets
reflectance spectrum
spectral resolution