【摘要】 针对传统区域增长算法易受噪声影响且局部分割性能不稳定的问题,提出了一种结合超体素与区域增长的屋顶面片点云分割算法。利用八叉树组织初始点云数据,基于点云的欧氏距离和法向量信息两个约束分割点云获得超体素。结合超体素结构特征,改进种子点选取准则,在超体素的光滑性和表面几何特征约束下进行点云区域增长,提取屋顶面片点云。选取不同复杂程度的建筑物LiDAR点云进行实验,结果表明,结合超体素与区域增长算法能有效提取复杂建筑物屋顶面片点云,提取率高且具有较好的适应性,可以为基于机载LiDAR的建筑物三维模型重建提供可靠的屋顶面信息。更多还原

【Abstract】 Aiming at the problem that the traditional regional growth algorithm is susceptible to noise and the local segmentation performance is unstable, a point cloud segmentation algorithm based on super-voxel and region growth is proposed. The initial LiDAR point cloud data is organized by the octree, and the super-voxel is obtained by the two constrained segmentation LiDAR point cloud based on the point cloud euclidean distance and the normal vector information. Combining the characteristics of super-voxel structure, the seed point selection strategy is improved, and the regional growth is carried out under the constraints of smoothness and surface geometric features, and the roof patch point cloud is extracted. Experiments were carried out on buildings of different levels of complexity, the results show that the combination of super-voxel and regional growth algorithm can effectively extract the point cloud of complex building roof, with high extraction rate and good adaptability, which can provide reliable roof surface information for 3 D model reconstruction of buildings based on airborne LiDAR.更多还原