基于机载LiDAR点云的道路提取方法研究

发布时间：2018-10-11 16:01

【摘要】：道路是关系国民经济的重要基础设施,道路信息的及时、准确、高效获取与更新对“智慧城市”的构建具有极其重要的意义。机载激光雷达(LiDAR)技术可以快速的获取地表高精度的三维点云数据,为道路信息的提取提供了一种准确、可靠的数据源。然而,现有基于LiDAR点云的提取方法无法准确提取道路的轮廓信息,提取结果存在道路碎片较多、完整性不高等问题。因此,本文综合利用机载LiDAR点云的高精度坐标信息和回波信息,以道路信息的快速、准确获取为目的,开展了以下研究工作： (1)综述了机载LiDAR点云数据的结构、特点及数据处理流程,概述了点云滤波的原理及主要方法。 (2)介绍了Clode算法以及基于强度信息的模糊C均值聚类的道路提取方法; (3)综合利用机载LiDAR点云的坐标和回波信息,在TIN约束方法的基础上引入DBSCAN聚类算法,实现了道路点云的提取,并通过与基于强度的模糊C均值聚类方法对比分析,验证了TIN约束和聚类方法的有效性。 (4)在道路点云提取的基础上,运用数学形态学和Hough变换对道路点栅格图像进行处理,实现了道路边界及中心线的提取,进而获得了规则化的道路网信息。 (5)在高分辨率遥感影像的辅助下,通过目视解译的方法对提取结果进行统计、分析,并引入准确度、完整度、总体质量三个指标进行精度评定,以此验证文中提取道路点云及道路网特征信息所采用方法的有效性和可靠性。研究表明：结合LiDAR点云的三维坐标和回波信息进行道路的提取是可行的。通过实验验证了结合TIN约束和DBSCAN聚类的方法能够去除大量的道路误分点,较好的保留道路点云,并在此基础上通过数学形态学及Hough变换处理道路栅格图像,可得到较为完整的道路网特征信息。本文的研究工作对“智慧城市”构建中道路信息的获取及更新具有一定的参考价值。
[Abstract]:Road is an important infrastructure related to the national economy. The timely, accurate and efficient acquisition and updating of road information is of great significance to the construction of "intelligent city". Airborne lidar (LiDAR) technology can quickly obtain high accuracy 3D point cloud data, which provides an accurate and reliable data source for road information extraction. However, the existing methods based on LiDAR point cloud can not accurately extract the contour information of the road. Therefore, in this paper, the high precision coordinate information and echo information of airborne LiDAR point cloud are synthetically used to obtain the road information quickly and accurately, and the following research works are carried out: (1) the structure of airborne LiDAR point cloud data is summarized. Characteristics and data processing process, The principle and main methods of point cloud filtering are summarized. (2) the Clode algorithm and the road extraction method based on fuzzy C-means clustering based on intensity information are introduced; (3) the coordinate and echo information of airborne LiDAR point cloud are synthetically used. Based on the TIN constraint method, the DBSCAN clustering algorithm is introduced to extract the road point cloud, and it is compared with the intensity based fuzzy C-means clustering method. The validity of TIN constraint and clustering method is verified. (4) on the basis of road point cloud extraction, road point raster images are processed by mathematical morphology and Hough transform, and road boundary and centerline extraction are realized. Then the regularized road network information is obtained. (5) with the help of high-resolution remote sensing images, the extraction results are analyzed and analyzed by visual interpretation, and the accuracy and completeness are introduced. In order to verify the validity and reliability of the method used to extract road point cloud and road network feature information, three indexes of total quality are evaluated. The research shows that it is feasible to extract the road by combining the 3D coordinates and echo information of LiDAR point cloud. The experimental results show that the method combining TIN constraint and DBSCAN clustering can remove a large number of road misdivision points and better preserve the road point cloud. On this basis, the road raster images are processed by mathematical morphology and Hough transform. More complete road network feature information can be obtained. The research work of this paper has certain reference value for obtaining and updating road information in the construction of "Wisdom City".
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P225.2

【参考文献】