SAM法在多光谱图像分类中的应用  

Application of SAM Algorithm in Multispectral Image Classification

作　　者：安斌[1] 陈书海[1] 严卫东[1] 

机构地区：[1]西北核技术研究所,西安710024

出　　处：《中国体视学与图像分析》2005年第1期55-60,共6页Chinese Journal of Stereology and Image Analysis

摘　　要：不同类型的地物具有不同的反射光谱, 在多维光谱空间中构成不同的特征向量, 这便是我们利用多光谱图像区分地物的物理依据。多光谱图像分类的精度受到多种因素的影响, 如模式样本集的数据结构、分类特征的提取、距离测度、分类准则、分类数目等。本文引入光谱匹配识别中的典型方法——光谱角度匹配法(SAM法), 通过与欧氏距离的比较分析与综合, 提出一种可以兼顾到光谱亮度与光谱向量方向(近似为光谱形状) 的距离测度, 并应用于K-均值动态聚类中, 通过与传统分类方法的比较, 证明这种方法的有效性。Different types of terrain would lie on the different site in the spectral space, which is the basic theory for remote analysis and clustering. The precision of clustering is due to the data structure of pattern sample、feature-selecting、distance measure、clustering rule、the number of clusters, and so on. In this Paper, a SAM(Spectral Angle Match) algorithm that is the classical algorithm for spectral match recognition is cited, By compared with Euclid distance, a distance measure based on spectral brightness and spectral vector direction (close to spectral shape) is presented. commonly, the spectral brightness is used as feature for clustering. However, the important and essential feature-Spectral shape is ignored in this clustering algorithm. And the algorithm is used to K-mean clustering. By compared with traditional clustering algorithm, the algorithm prove to be effective.

关 键 词：地物 反射光谱 分类 距离测度 SAM法 K-均值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]