高分辨率影像道路提取的整体矩形匹配方法  

Total rectangle matching approach to road extraction from high resolution remote sensing images

作　　者：朱长青[1] 杨云[2] 邹芳[2] 王奇胜[2] 

机构地区：[1]南京师范大学虚拟地理环境教育部重点实验室,江苏南京210054 [2]信息工程大学测绘学院

出　　处：《华中科技大学学报（自然科学版）》2008年第2期74-77,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家重点基础研究发展计划资助项目(2006CB701302);国家自然科学基金资助项目(40501058,40401052)

摘　　要：提出了一种整体矩形匹配方法,对高分辨率遥感影像中的城市直线道路进行提取.方法基于影像特征、道路知识和数学形态学的击中击不中思想,通过改变影像分割阈值、矩形宽度和矩形方向,从矩形的内部和外部进行整体匹配,使得矩形满足最佳匹配原则,从而提取出道路.利用提出的方法对高分辨率卫星影像进行了实验.结果表明:该方法能较好地消除树木、汽车等对道路提取的影响,有效地提取出直线道路的边缘.A total rectangle matching approach is proposed to extract straight roads in urban areas from high resolution remote sensing images. The characteristics of the image, the knowledge about the road and the hit-miss transformation of mathematical morphology were considered. Total rectangle matching is performed from the inside to the outside of the rectangle to satisfy the optimized criterion by changing the threshold of image segmentation, the width and the direction of rectangle, high-reso lution satellite images were tested. The results demonstrate that the proposed approach can eliminate the influence of trees, cars, etc. on road extraction, and extract the edges of straight roads effectively.

关 键 词：遥感影像 高分辨率遥感影像 道路提取 整体矩形匹配 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]