一种高分辨率遥感影像最优分割尺度确定的方法  

A Method of Determining Optimal Segmentation Scale of High Resolution Remote Sensing Image

作　　者：张吉星[1] 程效军[1,2] 郭王[1] 

机构地区：[1]同济大学测绘与地理信息学院,上海200092 [2]现代工程测量国家测绘地理信息局重点实验室,上海200092

出　　处：《地矿测绘》2016年第2期12-14,31,共4页Surveying and Mapping of Geology and Mineral Resources

摘　　要：近年来,面向对象的影像分析方法成为提取高分辨率遥感影像中目标地物信息的新技术,其关键和核心的步骤是多尺度分割。而在多尺度分割中,最优分割尺度的确定成为了重点和难点。鉴于此,文章以某大学校区的高分辨率遥感影像为实验对象,利用e Cognition软件,结合分割对象的均值方差和最大面积来确定最优分割尺度,得出了几种主要地物的最优分割尺度:水体的最优分割尺度为30,建筑物的最优分割尺度为40,植被的最优分割尺度为75,道路的最优分割尺度为100。经实验验证,分割尺度准确。In recent years, the method of object-oriented image analysis has become a new technology to extract thetarget feature information of high resolution remote sensing images. The key and core step of this method is multi-scalesegmentation. The emphasis and difficulty of multi-scale segmentation is to determine the optimal segmentation scale.Therefore, this paper takes the high resolution remote sensing image of a university campus as the experimental objectand uses eCognition software,and using a combination of mean variance method and maximum area method to determinethe optimal segmentation scale indexes of the segmentation objects. The results showed that the optimal segmentation scaleof water was 30,the optimal segmentation scale of building was 40,the optimal segmentation scale of vegetation was 75,and the optimal segmentation scale of road was 100. The segmentation scales were proved to be correct by experiment.

关 键 词：高分辨率遥感影像 多尺度分割 最优分割尺度 均值方差 最大面积 

分 类 号：P237[天文地球—摄影测量与遥感]