中文摘要：

提出了针对机载激光雷达点云的建筑物三维模型提取方法。首先通过滤波从点云中提取数字地形模型（DTM），再从数字表面模型（DSM）中剔除DTM影响，得到正规化数字表面模型（nDSM）；然后通过高程滤波和双边滤波从nDSM中得到建筑物区域；在建筑物几何形状约定下，将建筑物分“层”处理，通过边缘探测与规格化算法得到每“层”建筑物的边缘；边缘点坐标构成了建筑物三维模型的基础。该方法借助数学形态学易于计算机处理的优点，能较为快速、准确地提取建筑物三维模型，适合大规模、快速、对精度要求不高的lidar数据提取。最后通过数据验证了该方法的可行性及有效性。

英文摘要：

A method of extracting 3D building model from airborne lidar point cloud was presented. Digital Terrain Model （DTM） was generated by filtering to lidar point cloud. Then normalized Digital Surface Model （nDSM） was obtained by subtracting DTM from DSM. The building regions were detected from nDSM after elevation filtering and bilateral filtering. The building was arranged into several layers, and the building edges of every layer were extracted by edge detecting and edge normalizing technique on the assumption of building shape. The building model was vectorized by recording 3D coordinates of building edges. The method utilized the mathematical morphology, which was easy to process on computer, extracted 3D building model quickly and accurately. It is suitable for large-scale lidar point cloud application which has lower requirement in accuracy and higher requirement in processing speed. At last, the feasibility and validity of method has been demonstrated by example data in limited experiment conditions.