SAM法在多光谱图像分类中应用.pdfVIP

2005 10 1 CH INESE JO URNA L O F STEREO LOGY AND MI AG E ANA LYS IS V ol 10 No1 M a 2005 55 : 1007- 1482 ( 2005) 01- 0055- 06 SAM , , (, 7 10024) : 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 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 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 : 地物; 反射光谱; 分类; 距离测度; SAM 法; K- 均值 : TP 391 : A App lication of SAM A lgorithm in M u ltispectral Im age C lassification AN Bin, CHEN Shuha,i YAN W eid ng (N rthw est Institute f NuclearT echn l gy, X ian 710024) Abs tract: D ifferent types f terrain w uld lie n the d ifferent site in the spectral space, w hich is the basic the ry f r rem te analysis and clustering. The precisi n f clustering is due t the data structure f pattern samp lefeatureselectingd istancemeasureclustering ru lethe number f clusters, and s n. In this Pa per, a SAM (SpectralAngleM atch) alg rithm that is the classical alg rithm f r spectralmatch rec gniti n is cited, By c mpared w ith Euclid distance, a d istance measure based n spectral brightness and spectral vect r directi n ( cl se t spectral shape) is presented. c mm nly, the spectral brightness is used as fea ture f r clustering. H w ever, the mi p rtant and essential featureSpectral shape is ign red in this cluste ring alg rithm. And the alg rithm is used t Kmean clustering. By c mpared w ith trad iti nal clustering alg rithm, the alg rithm pr ve t be effective. K ey w ord s: terrain; spectral; clustering; d istancemeasure; SAM alg rithm; Km ean ( 1