针对激光点云数据中存在的大量冗余信息, 造成传输、存储等后续处理环节付出多余的硬件和时间成本的问题, 提出了一种基于特征信息的点云精简方法。利用自适应近邻点进行PCA计算点云法矢; 利用冯.米塞斯分布进行边缘点提取, 对非边缘点以点法矢为基础赋予其距离权重进行阈值判断, 提取特征点; 划分空间均匀网格, 以网格为单元计算法矢均值, 提取潜在特征点; 对网格非特征点进行单点提取。以标准的Bunny和工件模型为对象进行了MATLAB仿真实验, 所提算法与传统非均匀网格法、聚类法、三角面片消减法比较: 在精简比1∶5、1∶10、1∶15、1∶20情况下, 最大误差降低27%以上, 平均误差降低12%以上。实验结果表明所提算法在特征信息较多的模型处理上具有更好的精简能力。

A point cloud simplification method based on feature information was proposed to solve the problem that a large number of redundant information in the laser-point cloud data causeed redundant hardware and time costs in subsequent processing such as transmission and storage . The PCA with adaptive neighbor points was used to compute the normal vector of point cloud. The edge points were extracted by using vMF and the feature points in non-edge points were extracted by threshold based on point normal vector with distance weight, and divided the space to uniform grids, and calculated the normal vector mean in each cell to extract potential feature points. Single point was extracted from the non-feature points of a cell. The algorithm completed simplification of multiple point cloud models based on MATLAB, and was compared with Non-uniform 3D grids, clustering decimation, decimation of triangular meshes on the basis of the Bunny model and Workpiece model. In the simplification ratio of 1∶5, 1∶10, 1∶15, 1∶20, the maximum simplification error was reduced by more than 27%. And corresponding to the average error, it was reduced by more than 12%. Experimental results show that the proposed algorithm has better performance in simplification of model with more features.