摘要
介绍了利用4波段卫星云图对8类云进行分类的方法,包括建立输入云图样本库,构造测量量进行特征提取与特征层融合,构造表述模型,建立特征分类树。由于所研究的对象云为非刚体,存在不稳定性,加大了正确分类的难度,其分类方法与刚体目标的分类方法也存在一定的差异;又由于输入为8类4波段云图,因此我们不但获取了各类云图单波段上的特征,而且组合了它们在不同单波段上的特征,进行特征层融合。
オ? An automatic classifiction method for multibands clouds images of satellites is presented,it consists of collecting clouds images samples,selecting measurements for features,constructing representing models and classification trees.In order to bring down the difficultities for classification and reconition of clouds images,the non rigid bodies,we not only use the single band features of clouds images,but also conducting features fusion to get more information.Using this method,we identify automatically eight classes of clouds images successfully.
出处
《遥感技术与应用》
CSCD
1998年第2期27-31,共5页
Remote Sensing Technology and Application
关键词
卫星云图
遥感
识别
模式识别
Measurements,Feature extraction,Classification tree,Fussy desicion,Represention models,Fetures fusion