基于布模拟算法在LiDAR点云中生成数字高程模型

Digital Elevation Model Generation in LiDAR Point Cloud Based on Cloth Simulation Algorithm

机载激光雷达(LiDAR)能获得高精度的三维点云数据,可透过林叶到达地表面,快速准确反映出研究区域连续起伏的地形特征,有利于建立高分辨率数字高程模型(DEM)。用布模拟滤波(CSF)算法对机载LiDAR点云数据进行滤波,通过设置算法的布粒子生成个数和地面点分类阈值,从6组不同地势条件的点云数据中提取地面点云,用交叉表法计算出分类的Kappa系数在0.851~0.954之间。用CSF算法提取的地面点生成1 m×1 m的DEM,并与研究区域提供的DEM进行线性拟合,实验结果表明,两者的回归直线拟合优度系数R^2大于0.99,均方根误差在0.10451~0.30387之间。布模拟算法在提取点云地面点方面设置参数少、适用地形广,且基于该算法生成的高分辨率DEM能很好地表达该地区连续起伏的地表变化和地形特征。

Airborne light laser detection and ranging(LiDAR)can obtain three-dimensional point cloud data with high accuracy,can reach the ground surface through forest leaves,and reflect the continuous terrain features of the study area quickly and accurately,which is conducive to the establishment of a high-resolution digital elevation model(DEM).In this paper,cloth simulation filtering(CSF)algorithm is applied to filter the airborne LiDAR data.By setting the number of particles generated in the algorithm and the threshold of ground point classification,the ground point cloud is extracted from six groups of point cloud data under different terrain conditions,and the Kappa coefficient of classification is between 0.851 and 0.954.The DEM of 1 m×1 m is generated from the ground points extracted by the CSF algorithm,and the DEM provided by the research area is linearly fitted.Experimental results show that the regression line fitting goodness factor R2 is larger than 0.99 and the root mean square error is between 0.10451 and 0.30387.The cloth simulation algorithm has few parameters for extracting ground points of the point cloud and is suitable for a wide range of terrain.The high-resolution DEM generated by the proposed algorithm can well express the continuous undulating surface changes and terrain features of the region.