盾构隧道点云典型要素分割

Typical Elements Segmentation of Shield Tunnel Point Cloud

针对盾构隧道点云及其几何特征构建八叉树索引,使用体素化网格降采样进行点云降采样,利用统计特征滤波器达到精简点云,继而实现典型要素自动分割。结合点云精简算法,提高了随机抽样一致性(RANSAC)算法效率,通过拟合模型的几何特征,设置合理阈值,自动分割隧道典型要素。实验结果表明,该方法可以精确地分割出相邻距离阈值较小的盾构隧道典型要素,拟合的隧道半径与设计半径误差仅为3 mm;相比传统RANSAC算法,该算法运行速度提高了17倍,实现了相邻距离阈值达1.0 cm精度的目标分割。

Based on the shield tunnel point cloud and its geometric characteristics, the octree index is constructed.The voxelization grid downsampling is used to perform point cloud downsampling, the statistical feature filter is used for filtering, and the automatic segmentation of typical elements is implemented. In combination with the point cloud simplification algorithm, the efficiency of random sample consensus( RANSAC) algorithm is improved. By fitting the geometric features of the model, the reasonable thresholds are set and the typical elements of the tunnel are automatically segmented. The experimental results show that this method can accurately segment typical shield tunnel elements with small adjacent distance thresholds. The error between the radius of the tunnel and the design radius is only 3 mm. Compared with the traditional RANSAC algorithm, the rate of the algorithm reaches 17 times faster, and the method achieves a target distance of 1.0 cm accuracy of the adjacent distance threshold segmentation.