纹理与几何特征在道路提取中的应用  

Application of Texture and Geometrical Characteristics in Road Extraction

作　　者：马欢[1,2] 楚恒[2] 王汝言[1] 

机构地区：[1]重庆邮电大学通信与信息工程学院,重庆400065 [2]重庆市勘测院,重庆400020

出　　处：《计算机与现代化》2012年第7期96-99,共4页Computer and Modernization

基　　金：国家自然科学基金资助项目(41001299)

摘　　要：提出一种利用纹理与几何特征的高分辨率遥感影像道路提取方法。首先分析高分辨率遥感影像的纹理特征,提出基于纹理特征的聚类方法,将影像大致分为道路区域和非道路区域,然后选取适当的几何特征指数,剔除道路区域中含有的非道路像素,得到初步道路信息。最后通过数学形态学处理,去除初步道路信息中由于车道线、树木影响而产生的孔洞,最后得到完整的道路信息。实验结果表明,与传统方法相比,该方法能够有效地从高分辨率遥感影像中提取道路。This paper proposes a new method that integrating texture and geometrical characteristics for the road extraction of high resolution remote sensing imagery. Firstly, texture characteristics of the high resolution remote sensing imagery is analyzed, and a clustering method based on texture characteristics is proposed. The clustering method approximately divides the imagery into road areas and non-road areas. Then the geometrical characteristics are utilized to reject the false road pixels which have been mingled in road areas, the preliminary road structure has been formed. Finally with the function of mathematical morphology, the holes that come from vehicles and trees are wiped off. The integrated road information is obtained. Compared with the other traditional methods, the result of the experiment shows that the proposed method can extract the road information in the high resolution re- mote sensing imagery effectively.

关 键 词：道路 纹理 聚类 几何特征 数学形态学 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]