基于车载LiDAR点云数据的地面点滤波方法研究

Research on Ground Point Filtering Method Based onVehicle-borne LiDAR Point Cloud Data

为了提高基于车载LiDAR点云数据的地面点提取精度,最大限度发挥车载LiDAR点云数据的利用效率。本文提出了一种基于Otsu算法与最远点采样(Farthest Point Sampling,FPS)算法的地面点提取方法。首先,根据Otsu算法自适应计算高程分割阈值,滤除大部分地面点,实现地面点的粗提取;其次从地面点粗提取结果中随机选取一个种子点,并使用FPS算法选取剩余种子点进行最优平面模型拟合提取精确地面点。通过两组实验数据进行地面点提取实验,结果表明,两组实验数据地面点提取结果误差均在8%以内,验证了本文方法的有效性与适用性。

In order to improve the accuracy of ground point extraction based on vehicle-borne LIDAR point cloud data and maximize the utilization efficiency of vehicle-borne LIDAR point cloud data.This paper presents a ground point extraction method based on Otsu algorithm and farthest point sampling(FPS)algorithm.Firstly,the Otsu algorithm is used to adaptively calculate the elevation seg-mentation threshold,filter out most of the ground points,and realize the rough extraction of ground points;Secondly,a seed point is randomly selected from the rough extraction results of ground points,and FPS algorithm is used to select the remaining seed points for optimal plane model fitting to extract accurate ground points.Through two groups of experimental data,the results show that the error of two groups of experimental data is less than 8%,which verifies the effectiveness and applicability of this method.