摘要
利用单个数据源的数学形态学道路提取方法不能充分利用道路的特征,提取的道路信息不完整。针对这一缺陷,并考虑到机载LiDAR数据可以提供高程信息,提出了将机载LiDAR数据和高分辨率遥感影像数据结合起来的城市道路网的提取方法。以徐州市的机载LiDAR数据和高分辨率遥感影像数据作为实验数据,首先将两者进行精确配准,然后利用伪道路信息去除的方法分别将植被信息和建筑物信息等去除,得到基本的道路轮廓,再利用形态细化算法提取道路的中心线,最后,在ArcGIS和Matlab编程环境下实现了改进的道路修剪算法(IRT),利用该算法进行道路修剪,得到了平滑和连贯的城市道路网。经过精度评价可以看出:利用该方法可以较好地避免建筑物阴影、低矮植被群等对道路提取的影响,道路的识别精度达到84%以上。
The conventional mathematical morphology method using single data source to extract road net work which could not take full advantage of the road characteristics, the extracted road information was not complete. In view of this drawback, and base on the airborne LiDAR data can provide elevation informa tion,this paper proposes a method which combines the airborne LiDAR data with high resolution remote sensing images to extract city road network. The airborne LiDAR data and high resolution remote sensing QuickBird images of Xuzhou were taken as the experimental data, the precise registration between them were first done, then the FRIR （False Road Information Removing） method was used to remove the vegeta tion and buildings separately,so the basic road contour was displayed. Finally, this paper achieved an Im- proved Road Trimming （IRT） algorithm under the ArcGIS and Matlab programming environment, the road network was trimmed by the algorithm, then a smooth and continuous city road network was obtained. The result of the accuracy evaluation indicates that the method was proposed can be used to avoid the influence of the building shadow, city squares, parking lots and the vegetation groups on both sides of the road to the road centerlines extraction well, and the recognition accuracy of the road network is more than 84 %.
出处
《遥感技术与应用》
CSCD
北大核心
2013年第4期562-568,共7页
Remote Sensing Technology and Application
基金
国家863计划项目(2012AA12A305)
国家"十二五"科技支撑技术项目(2012BAJ15B04)
江苏省普通高校研究生科研创新计划资助项目(CX10B_143Z
CXLX12_0956)
江苏高校优势学科建设工程资助项目
关键词
LIDAR
高分辨率遥感影像
道路网提取
伪信息去除
数学形态学
影像配准
LiDAR
High resolution remote sensing images
Road network extraction
False road informa-tion removing
Mathematical morphology
Image registration
