一种高分辨率遥感影像最优分割尺度确定的方法.pdf

• 12 • 地矿测绘 2016,32(2) :12 ~14 CN 53 -1124/TD ISSN 1007 -9394 Surveying and Mapping of Geology and Mineral Resources 一种高分辨率遥感影像最优分割尺度确定的方法$ 张吉星\程效军 12 , 郭 王 1 (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) 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 eCogmrion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。 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 中图分类号:P 237 文献标识码 :A 文章编号:1007 -9394(2016)02 -0012 -03 A Method of Determining Optimal Segmentation Scale of High Resolution Remote Sensing Image ZHANG Ji-xing1 , CHENG Xiao-jun1 2 , GUO Wang1 (1. College of Surveying and Geo-Informatics , Tongji University, Shanghai 2QQQ92 , China ； 2. Key Laboratory of Advanced Engineering Survey of SBSM , Shanghai 200092, China) Abstract ： In recent years, the method of object-oriented image analysis has become a new technology to extract the target feature information of high resolution remote sensing images. The key and core step of this method is multi-scale segmentation. 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 object and uses eCognition software,and using a combination of mean variance method and maximum area method to determine the optimal segmentation scale indexes of the segmentation objects. The results showed that the optimal segmentation scale of water was 3 0 ,the optimal segmentation scale of building was 4 0 ,the opt