Abstract: Aiming at the problem that the existing building extraction methods of airborne laser scanning data are complicated and easily disturbed by vegetation,a building extraction method using a two-way cloth simulation strategy is proposed.First,a regular grid is established for the original point cloud,and on the basis of forward cloth simulation filtering,the over-high buildings are first extracted through the grid elevation difference statistics,and then the reverse cloth simulation is used to roughly extract the top of the building from the remaining feature points.Surface point cloud;then conduct penetration analysis and combine the morphological opening operation to further eliminate the wrongly mentioned vegetation points;finally,the three-dimensional grid containing the top surface point cloud is used as the seed grid,according to the adjacency relationship between the grids and The geometric features of the internal point cloud are used for regional growth to obtain a complete building point cloud.Experimental results show that in complex scenes,this method can effectively avoid the interference of vegetation,quickly extract building point clouds,and has the advantages of high extraction accuracy and less calculation time.

Key words: airborne laser point cloud;bidirectional cloth simulation;building extraction;morphological open operation;regional growth