利用多尺度几何特征向量的变化检测方法  

Change Detection Based on Multi-scale Geometric Feature Vector

作　　者：陆苗[1,2,3] 梅洋[4] 赵勇[4] 冷亮[5] 

机构地区：[1]中国农业科学院农业资源与农业区划研究所,北京100081 [2]农业部农业信息技术重点实验室,北京100081 [3]武汉大学遥感信息工程学院,湖北武汉430079 [4]国家基础地理信息中心,北京100830 [5]吉林大学应用技术学院,吉林长春130022

出　　处：《武汉大学学报（信息科学版）》2015年第5期623-627,共5页Geomatics and Information Science of Wuhan University

基　　金：国家科技支撑计划资助项目(2012BAK12B02);国家自然科学基金资助项目(41101442)~~

摘　　要：提出了一种利用多尺度几何特征向量的变化检测方法,其基本原理是基于多尺度影像分割将变化检测从传统的像素光谱空间转换到对象尺度空间,利用多尺度分割形成的几何特征向量进行变化检测。以陕西省渭南市为研究区域,使用本方法检测该区域2002~2009年的地表覆盖变化。从变化检测结果可以看出,本文方法的检测效果优于其他传统检测算法。We present a change detection method based on multi-scale FV）. The change analysis standard in this novel method is converted ment scale space. Context information beyond multi- change detection. Specifically, this approach specifi scales at first. Secondly, two-date imageries are ly, multi-scale geometric feature vector of a dete geometric feature vector from pixel spectral space scale imagery segmentation is applied to es number and values of multipl segmented respectively ction unit are constructed in g multiple different e segm scales dates. （MSG- to seg- perform entation Third- Finally,the change intensity between two-date multi-scale geometric teature vectors is calculatea as me change standard. The study area of Weinan in Shannxi province is tested to analyze the land cover change from 2000 to 2009. In this sample area, three geometric features are used in this approach. Then, the optimized geometric feature is compared to other existing methods （CVA, Correlation）. The MSGFV approach is proved to outperform other methods.

关 键 词：变化检测 基于对象的影像分析 影像分割 尺度关联对象 多尺度几何特征向量 

分 类 号：P237.4[天文地球—摄影测量与遥感] P237.9[天文地球—测绘科学与技术]