机载LiDAR点云体元化及其在3D滤波中的应用

Airborne LiDAR point cloud voxelization and its 3D filtering application

相较于规则格网、不规则三角网和点云等数据结构,体元的优势在于其既是真3D结构同时又隐含有邻域关系。提出了一种二值（1和0）3D体元数据结构模型以重构机载Li DAR点云,在此模型基础上提出了一种基于体元的3D滤波算法。该算法首先选取高程最低目标体元作为地面种子体元,然后标记与其3D连通的＂1值＂体元集合为地面体元。采用ISPRS标准测试数据验证提出算法。结果的定量分析表明算法平均Kappa系数在相对平坦、不连续及陡坡地形分别为92.94%、60.43%和73.69%。

Compared to raster grid,triangulated irregular network and point cloud,the voxel is the true 3 D structure and can represent the implicity of neighborhood relationships. A binary（ ＂1 ＂and ＂0 ＂） 3 D voxel-based data（ B3 VD） structure model is proposed to reconstruct airborne Li DAR point cloud,and a voxel-based 3 D filtering（ V3 F） is developed for separating ground voxels from nonground ones. The proposed V3 F algorithm selects the lowest voxels with value 1 as ground seeds and labels their 3 D connected voxel set as ground voxels. ISPRS benchmark dataset is utilized to evaluate the performance of V3 F and compared with other eight filtering methods. Results indicat that the average Kappa coefficients for sites with relatively flat urban areas,discontinuous and rough slope terrains are 92. 94%,60. 43% and 73. 69%,respectively.