基于NSCT与形状特征的遥感影像道路提取  

Road extraction from remote sensing images based on NSCT and shape features

作　　者：梁栋[1] 刘书丽[1] 颜普[1] 鲍文霞[1] 

机构地区：[1]安徽大学计算智能与信号处理教育部重点实验室,安徽合肥230039

出　　处：《华中科技大学学报（自然科学版）》2012年第3期9-12,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61172127;11071002);安徽大学博士科研启动经费资助项目;安徽大学青年骨干教师培养经费资助项目

摘　　要：提出一种基于非抽样Contourlet变换(NSCT)与形状特征的遥感影像道路提取算法.首先对图像进行非抽样Contourlet变换,得到不同尺度、不同方向上的变换系数,对变换系数进行增强处理并通过反变换得到增强图像;然后对增强图像进行分割处理,利用道路形状特征对分割结果进行道路提取;最后利用光谱特征对提取的道路进行判断,并利用形态学方法对道路网进行规整.对比实验结果表明:该算法取得了较好的高分辨率遥感影像道路提取效果.An algorithm for road extraction from remote sensing images based on the NSCT（nonsubsampled Contourlet transform） and shape features was proposed.Firstly,the coefficients in different scales and different directions were obtained by NSCT,which were enhanced and used to construct the enhanced image by the inverse transformation.Then the enhanced image was segmented,and the roads were extracted by using several object shape features.Finally,the extracted roads were judged by spectral feature and regulated by mathematical morphology.The experiments show the proposed method has a better effect on the high-resolution remote sensing images.

关 键 词：图像增强 图像分割 遥感影像 非抽样CONTOURLET变换 形状特征 道路提取 

分 类 号：TN911.73[电子电信—通信与信息系统]