基于光谱曲线整体相似性测度的匹配分类

Matching and Classification Based on the Whole Comparability Measure of Spectral Curve

高光谱影像具有丰富的空间、辐射和光谱信息,每一个像元都可以提取出连续的光谱曲线。因此,可以通过高光谱数据与已知的参考光谱曲线波形或特征相似性对比分析的方法识别地物类型。在整体相似性测度约束下,综合考虑数值指数和形状指数,利用光谱特征向量间的差异和曲线信息熵提出了一种新的匹配分类的方法。实验结果表明,该方法具有分类精度高、适应性强的特点。

Hyperspectral images comprise abundant spatial, radiative and spectral information, a consecutive spectral curve could be picked up from every pixel. Therefore, the hyperspectral images could be used to distinguish different types by contrasting and analyzing the spectrum curve or feature comparability between hyperspeetral remote sensing data and the known reference spectrum. On the restriction of total comparability measure, after taking value index and figure index into account, a new method has been put forward to express the differences between spectral vectors and curve information entropy. The results show that this method is effective in classification, and has high adaptability.