一种基于高分辨率遥感影像的道路提取方法  

A New Approach of Extracting Road from High-Resolution Remote Sensing Images

作　　者：谢谦礼[1] 程承旗[1] 马廷[2] 

机构地区：[1]北京大学遥感所,北京100871 [2]中国科学院地理科学与资源研究所资源与环境信息国家重点实验室,北京100101

出　　处：《计算机工程与应用》2006年第17期188-190,共3页Computer Engineering and Applications

基　　金：国家863高技术研究发展计划资助项目(编号:2003AA783060)

摘　　要：道路等线性地物的自动提取一直是高分辨率遥感影像研究所关注的内容。论文在分析现有的各种提取方法的基础上,结合城市道路在高分辨率遥感影像上的特点,提出一种半自动的道路提取法。该方法先对遥感影像中的道路点进行采样,统计其光谱特征,然后再在道路上设置道路生长原点,从这些生长原点开始,根据统计得出的道路光谱特征对邻域点进行判断生长,最后对生长出的道路图利用数字形态学进行内部腐蚀和边界平滑处理。试验结果表明,该方法提取道路具有较高的精度和实用性。Automatic extraction of linear objects such as roads has been the hotspots in the research of high-resolution remote sensing images.In this paper,according the characteristics of urban road in high-resolution remote images,a new semiautomatic extraction road approach is proposed based on the analysis of various existent ones.In the approach,we first have sampled some pixels and maken statistical calculation of their spectrum.Then we have set an original pixel in the road area,and according to the statistical result,the neighbor pixel of the original one has been judged whether it is road pixel or not,and then the neighbor pixels of pixels that has been judged to be road pixel has been judged and go on.Finally,mathematical morphology has been used to eliminate disturbances inside and smooth the edges of the road in the grown image.The experiments indicate that the proposed approach is efficient and practical for extracting roads.

关 键 词：高分辨率遥感影像 道路提取 数学形态学 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]