一种遥感影像挖掘的相似性度量方法  

A Similarity Measurement Method of Remote Sensing Image Mining

作　　者：陈晓飞[1,2,3] 段洣毅[2,3] 张佳骥[1] 

机构地区：[1]石家庄通信测控技术研究所,河北石家庄050081 [2]北京航空航天大学,北京100083 [3]北京图形研究所,北京100029

出　　处：《无线电工程》2008年第12期29-31,共3页Radio Engineering

摘　　要：遥感影像数据挖掘是一个有着广阔应用前景的研究领域。对图像检索、图像分类、图像聚类、空间关联规则挖掘和图像变化检测等数据挖掘应用而言,相似性度量是基础和前提。采用了图像空间划分的策略,在此基础上计算颜色、纹理和形状等3方面的低层视觉特征来描述图像,采用多维特征空间的网格划分来降低数据维数并建立了影像的相似性度量。实验结果表明,该方法对影像具有一定的几何和光照不变性。Remote sensing image data mining is a research field with wide application prospect. For the data mining application such as image searching, image classing, image clustering, space associated rule mining, image change detecting, etc., the similarity measurement is the necessary condition and foundation. The paper introduces the strategy of image space partition, gives a image description by using the low-level vision feature of color, texture and shape etc, reduces the data dimension with the mesh partition of multidimensional feature space, and establishes the similarity measurement of image. The experimental results show that this method has a certain extent geometry and optics invariability.

关 键 词：遥感影像 挖掘 相似性度量 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]