机载激光雷达点云数据分类方法研究

发布时间：2018-02-16 05:45

本文关键词： 坡度算法区域增长阶层式分类 ObjectARX　出处：《西南交通大学》2013年硕士论文　论文类型：学位论文

【摘要】：机载激光雷达技术是新兴的一门技术,在地形勘测、城市建筑物三维建模等方面都有很好的应用。本文从滤波分类和建筑物点云提取两方面进行了探讨。首先介绍了常用的滤波分类算法和建筑物提取算法。实验验证了传统数学形态法、改进的数学形态法和距离期望滤波算法的滤波效果。传统的坡度滤波算法,利用经验知识或样本区间确定坡度阈值,不能完全得到真实的坡度阈值。在山区地形中,地形自身坡度与地物坡度之间的差别不是很大。仅仅应用坡度阂值很难判定,容易造成过度滤波或真实地形的丢失。针对以上问题,本文对传统的坡度滤波算法进行了改进,增加了双向坡度阈值、高程阈值和二次滤波。首次滤波时以较小的坡度阈值得到初始DEM,计算其最大坡度值,以最大坡度值作为第二次滤波的坡度阈值,一定程度上解决了坡度阈值选择的自适应问题。针对单向阈值在地形突变地区容易过度滤波的情况,增加了双向阈值,在地形变化较大地区有一定的普遍适用性。另外融合了区域增长的思想,增加了高程阈值,加强了细节保护。在研究建筑物区域增长算法的基础上,对现有区域增长算法进行了一些改进。按照高程阈值分割、梯度分割、邻域半径密度分类的步骤得到建筑物的脚点,该算法较好地剔除了植被点和地面低矮附属物的干扰,解决了建筑物和植被点分离过度和分离欠佳的问题,利用邻域半径密度法可以准确的得到建筑物种子点,较快的实现建筑物的分割分类。按照阶层式分类分割的策略,利用C#.net和ObjectARX技术在AutoCAD2010中开发了学习型的机载激光雷达点云分类分割软件LIDAR-Scane。该软件可以实现点云分类显示、不规则三角网构建、点云分类分割等功能。
[Abstract]:Airborne lidar technology is a new technology in topographic survey, In this paper, we discuss the filtering classification and building point cloud extraction. Firstly, the common filtering classification algorithms and building extraction algorithms are introduced. Proved the traditional mathematical form method, The filtering effect of the improved mathematical morphology method and the distance expectation filtering algorithm. In mountain terrain, the traditional slope filtering algorithm can not get the true gradient threshold completely by using empirical knowledge or sample interval to determine the slope threshold. The difference between the slope of the terrain and the slope of the object is not very big. It is difficult to judge only the threshold value of the slope, which is easy to cause excessive filtering or the loss of the real terrain. In order to solve the above problems, the traditional slope filtering algorithm is improved to increase the bidirectional gradient threshold, elevation threshold and secondary filtering. At the first time, the initial DEM is obtained with a smaller gradient threshold, and the maximum slope value is calculated. Taking the maximum slope value as the slope threshold of the second filtering, the adaptive problem of slope threshold selection is solved to some extent. In view of the situation that one-way threshold is easy to be over-filtered in the abrupt terrain area, the bidirectional threshold is increased. In addition, the idea of regional growth is integrated, the elevation threshold is increased, and the detail protection is strengthened. Based on the study of the building region growth algorithm, some improvements have been made to the existing area growth algorithm. According to the steps of elevation threshold segmentation, gradient segmentation and neighborhood radius density classification, the foot points of the building are obtained. The algorithm can eliminate the interference of vegetation points and low appendages on the ground, solve the problem of excessive separation and poor separation of buildings and vegetation points, and use the neighborhood radius density method to accurately obtain the building seed points. The faster implementation of building segmentation and classification. According to the strategy of hierarchical classification and segmentation, the LIDAR-Scane-based airborne lidar point cloud classification software LIDAR-Scaneis developed in AutoCAD2010 by using C#.net and ObjectARX technology. The software can realize point cloud classification display and irregular triangulation network construction. Point cloud classification and segmentation functions.
【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】