A Fast and Accurate Segmentation Method for Ordered Point Cloud of Large-Scale Scenes

A fast and accurate segmentation method for point cloud of large-scale scenes is proposed. A scan-line-based ground filter algorithm is designed based on the ordering of point cloud and the geometrical characteristic of the ground, complex ground conditions such as slopes can be handled. Non-ground points are fast segmented point-by-point based on the initial threshold which takes the performance of the scanning system into consideration. Then over-segmented points are merged through the volume-based adaptive algorithm. The accuracy rate of the proposed method is over 90% and the point-by-point processing speed is 14.5 μs per point, real-time processing can be achieved.