基于逆向工程的三维激光扫描点云数据滤波方法

Filtering method of 3D laser scanning point cloud data based on reverse engineering

传统方法仅剔除原始点云边缘点,内部离散点数量过大,导致滤波处理后的点云保留点精确率较差。为此,提出基于逆向工程的三维激光扫描点云数据滤波方法。利用逆向工程技术,构建非均匀网格,精简原始点云数据,采用KD-Tree邻域搜索算法,选取点云数据最优滤波邻域,根据邻域特征信息计算局部收敛点,使噪声点漂移到局部模式点位置,通过滤波器迭代更新噪声点,直至局部模式点不会产生新的噪声。进行对比实验,三维激光扫描橄榄树点云数据,三组实验分别去除原始点云中残缺叶片和冠层边缘的噪声点,结果表明,此次方法提高了保留点精确率,点云数据的滤波品质要优于传统方法。

The traditional method only removes the edge points of the original point cloud,and the number of internal discrete points is too large,resulting in the poor accuracy of the point cloud preserved after filtering. Therefore,a filtering method of 3 D laser scanning point cloud data based on reverse engineering is proposed. By using reverse engineering technology,non-uniform grid is constructed to simplify the original point cloud data. KD tree neighborhood search algorithm is used to select the optimal filtering neighborhood of point cloud data. The local convergence points are calculated according to the characteristics of the neighborhood,so that the noise points drift to the local mode points,and the noise points are updated iteratively through the filter until the local mode points do not generate new noise. Three groups of experiments were carried out to remove the noise points in the original point cloud,the incomplete leaves and the edge of the canopy. The experimental results show that this method improves the accuracy of the point cloud,and the filtering quality is due to the traditional method.