利用层次聚类对移动曲面拟合滤波算法快速分类的研究  

Fast classification of moving surface fitting filtering algorithm based on hierarchical clustering

作　　者：唐菓 邢承滨 朱磊 邓兴升[1] 丁美青[1] TANG Guo;XING Chengbin;ZHU Lei;DENG Xingsheng;DING Meiqing(School of Traffic & Transportation Engineering, Changsha University of Science and Technology, Changsha 410014,China)

机构地区：[1]长沙理工大学交通运输工程学院,湖南长沙410004

出　　处：《测绘工程》2021年第3期32-40,共9页Engineering of Surveying and Mapping

基　　金：长沙理工大学公路地质灾变预警空间信息技术湖南省工程实验室开放基金资助项目(KFJ180604)。

摘　　要：经典移动曲面滤波算法由于算法简练,适用范围广泛且滤波效果较好,适用于多种地形。但是传统移动曲面滤波方法存在较多缺陷,如计算阈值参数难以确定、各个格网间阈值参数缺少相关性、分类主要依据高差阈值及水平距离相关性较小等缺点。文中提出层次聚类算法,将三维地形转换为二维平面,利用相邻点水平距离和高差构建数据集,进行聚类判断点云的属性,采用ISPRS提供的15组样本,定性和定量分析本算法的滤波精度。为验证本聚类算法的优越性和科学性,同时与改进型移动曲面和PTD滤波算法进行精度对比,充分说明本算法相较于其他算法的优越性和高效性。Because of its concise algorithm,wide range of applications and good filtering effects,the classic moving surface filtering algorithm is suitable for a variety of terrains.However,the traditional moving surface filtering method has many shortcomings:the threshold parameters are difficult to determine,lack of correlation between the threshold parameters across grids,the classification is mainly based on height difference threshold and the horizontal distance has slight correlation.This paper presents a hierarchical clustering algorithm,which converts three-dimensional terrain into two-dimensional plane,and builds a dataset using horizontal distance and height difference between adjacent points to cluster an judge the attributes of point cloud.This paper uses 15 sets of samples provided by ISPRS.The filtering accuracy of this algorithm is described qualitatively and quantitatively.In order to verify the superiority and scientific city of this clustering algorithm,it is compared with the precision of the improved moving surface algorithm and PTD filtering algorithm,which fully shows the superiority and efficiency of this algorithm compared with other filtering algorithms.

关 键 词：LIDAR 移动曲面 层次聚类 滤波 并项 

分 类 号：P237[天文地球—摄影测量与遥感]