基于高程统计的机载LiDAR点云三角网渐进滤波方法

A Filtering Method with Adaptive TIN Models for Airborne LiDAR Points Based on Elevation Statistics

机载LiDAR点云数据的滤波是DEM生成的关键,大部分点云滤波方法缺少先验知识,而点云数据在局部地表的高程分布具有一定的统计特征,可以为滤波算法提供一定的参考。本文提出一种基于高程统计先验知识的机载LiDAR三角网渐进滤波方法,该方法先按照点云数据的覆盖范围以及地形起伏情况,将点云数据进行区域划分,统计每个分区内的点云高程,进而进行多阈值分割,得到备选地面点集合。基于分割结果,自定义格网尺寸提取初始地面种子点,改进三角网渐进滤波方法。本文采用ISPRS公布的实验数据进行方法验证和误差评定,与原始三角网渐进滤波方法相比,该方法提高了分类精度。

Airborne LiDAR points' filtering is the key to generate DEM.Many filter algorithms of LiDAR point data lack prior knowledge.Elevation distributions of points cloud in local area show statistics characteristics,which provide certain reference for filter method.This paper proposes a filtering method with adaptive TIN models based on elevation statistics.In this method,based on the coverage area and terrain slopes,the data set is divided into several parts.Statistical analyses of raw points cloud elevation in each data partition are applied to carry out multi-threshold segmentation.Based on the segmentation results,initial seed ground points are extracted from user-defined grid size for adaptive TIN filtering.Data sites published by ISPRS are used to test method and evaluate both two type errors.Compared with original adaptive TIN filtering algorithm,this method improves both efficiency and classification accuracy.