一种高分辨率遥感影像道路信息提取方法  

A Methodology for Road Extraction from High Resolution Remote Sensing Images

作　　者：周安发[1] 周家香[1] 

机构地区：[1]中南大学地球科学与信息物理学院,湖南长沙410083

出　　处：《遥感技术与应用》2012年第1期94-99,共6页Remote Sensing Technology and Application

基　　金：国家自然科学基金青年科学基金项目(40901171);国家863计划项目(2009AA122004);武汉大学测绘遥感信息工程国家重点实验室基金项目(09R03)

摘　　要：道路在国民经济建设和国防建设中发挥着重大的作用,是非常重要的基础地理信息,一直以来研究道路的自动或半自动提取都被视为热点和难点,提出的理论和方法层出不穷,但没有一种方法能推广应用。针对道路难以从高分辨率遥感影像中提取的问题,提出了一种道路提取方法,首先影像将RGB颜色空间转换成HSV颜色空间,运用改进的区域分隔算法,实现道路区域的分隔。然后利用数学形态学腐蚀、膨胀等基本运算,结合区域的形态特征(如面积、紧凑度等)实现道路与非道路信息的分离。对多幅复杂道路图像进行试验,结果表明本方法能够很好地实现从复杂环境中提取道路信息。The road is a very important geographic dataset and it plays an important role in national economy and national defense.Research on the automatic or semi-automatic road extraction has always been regarded as a hot and difficult topic.Although lots of theories and methods have been proposed,there is no one that is widely used in production environment.Since road extraction from high resolution remote sensing images is difficult,in this paper,we propose a new method for road extraction.Firstly,a RGB color space image is converted to HSV color space image and an improved regional segmentation algorithm is used to separate the road regions.Then,a combined mathematical morphology operation（such as erode,dilate etc.） with regional shape indices（such as area,compactness etc.） is applied to separate the road objects and non-road objects.The experiment results over several complex road images show that this method can be used to extract roads under more complex conditions.

关 键 词：高分辨率遥感影像 区域分隔 数学形态学 形状特征 道路提取 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]