机载LiDAR点云数据建筑物检测和屋顶轮廓线提取算法研究

发布时间：2018-02-11 22:34

本文关键词： 机载激光雷达滤波处理建筑物检测屋顶轮廓线提取　出处：《辽宁工程技术大学》2014年硕士论文　论文类型：学位论文

【摘要】：建筑物是城市的重要组成部分,随着近年来“数字城市”的不断升温,对于城市三维模型,尤其是建筑物三维模型的需求越来越大。传统的城市三维模型获取手段以航空摄影测量为主,该技术已十分成熟,但在三维模型重建方面存在较大缺陷,比如地物几何信息的缺失、三维重建高程精度低等,这主要是由于航空影像数据是二维数据,地面控制点的选取、立体像对的匹配等都会造成一定的精度损失。机载激光雷达(Airborne LiDAR)作为近几年发展迅猛的新型主动传感器,集定姿定位系统(POS)和激光测距仪为一体,直接获取地表地物的高密度离散三维点云数据,为数字城市建设,特别是建筑物重建提供极大的便利。本文采用原始激光LiDAR点云数据,在分析建筑物检测和屋顶轮廓线提取现有算法的基础上,主要研究以下几个内容：(1)LiDAR点云数据的滤波处理。本文采用一种移动窗口的滤波算法,结合区域增长策略对地面点进行检测,针对陡坎等地物点,将其利用高差阈值归为地面点,实验证明,该方法对于地形起伏不大的城区有较好的处理效果,同时避免了与建筑物有相似阶跃特征的地下地物对建筑物提取可能存在的干扰。(2)建筑物点云提取算法。本文采用一种基于曲率变化、并结合共面判断、设定高差阈值和区域增长的建筑物点云提取算法,较为新颖地融合建筑物内部点信息和建筑物轮廓边缘信息,实验证明,可准确进行建筑物点云提取,并能有效提取曲面屋顶点云数据。(3)提出一种直接基于机载LiDAR点云数据的屋顶轮廓线提取算法,根据相邻边缘点间线段某侧不存在激光点的搜索策略,依次检测单个屋顶的边缘点,利用最小二乘方法拟合屋顶轮廓线,并对其进行规则化和扩展处理。实验表明该算法能快速准确地提取建筑物屋顶轮廓线。
[Abstract]:The building is an important part of the city, with the "digital city" heating up in recent years, for the three-dimensional model of the city, In particular, the demand for 3D models of buildings is increasing. The traditional methods of obtaining 3D models of cities are mainly aerial photogrammetry. The technology is very mature, but there are some defects in the reconstruction of 3D models. For example, the lack of geometric information of ground objects, the low accuracy of 3D reconstruction elevation, etc., which is mainly due to the fact that the aerial image data are two-dimensional data and the selection of ground control points. Airborne LiDAR (Airborne LiDAR), as a new type of active sensor with rapid development in recent years, integrates POS and laser rangefinder. Direct acquisition of high density discrete 3D point cloud data of surface features provides great convenience for digital city construction, especially for building reconstruction. The original laser LiDAR point cloud data are used in this paper. On the basis of analyzing the existing algorithms of building detection and roof contour extraction, this paper mainly studies the filtering processing of point cloud data from the following parts: 1 / 1 / 1 LiDAR. In this paper, a moving window filtering algorithm is used. Combined with the regional growth strategy, the ground points are detected, and the threshold of height difference is classified as the ground points. The experimental results show that the method has a better treatment effect for the urban areas with little topographic fluctuation. At the same time, it avoids the possible interference of underground ground objects with similar step characteristics to the extraction of buildings. The algorithm of point cloud extraction for buildings is based on curvature change and coplanar judgment. The building point cloud extraction algorithm, which sets the threshold of height difference and area growth, integrates the building interior point information with the building contour edge information. The experimental results show that the method can extract the building point cloud accurately. An algorithm for extracting roof contours directly based on airborne LiDAR point cloud data is proposed. According to the search strategy of laser points in a line segment between adjacent edge points, there is no laser point in the line segment between adjacent edge points. The edge points of a single roof are detected in turn, and the roof contour is fitted by the least square method, which is regularized and extended. The experimental results show that the algorithm can extract the building roof contour quickly and accurately.
【学位授予单位】：辽宁工程技术大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TU198;TN958.98

【参考文献】