基于虚拟格网的建筑物点云轮廓线自动提取

在分析现有轮廓线提取方法不足的基础上，提出基于虚拟格网的建筑物轮廓线自动提取方法。该方法利用建筑物点云生成虚拟格网并进行二值填充；采用邻域分析方法进行边界格网的标记与追踪；为了避免边界追踪错误，设计了基于方向的单边缘格网抑制方法及基于距离的连接关系调整方法以改善提取结果质量；根据格网追踪结果，从原始建筑物点云中提取真实轮廓点以保持原始建筑物轮廓形态；采用随机抽样一致性估计及最小二乘拟合方法进行轮廓线规则化处理，实现建筑物轮廓线的自动提取。实验结果表明，该方法能快速从建筑物点云中提取轮廓线，可为建筑物轮廓线的自动提取提供一种可行的解决方案。

Automatic Boundary Extraction from Building Point Clouds Based on Virtual Grids

Based on the analysis of the deficiency of building boundary extraction method, this paper proposes a new method for automatic boundary extraction from building point clouds based on virtual grids. Firstly, use building point clouds to generate virtual grids, and mark virtual grids by binary value based on points of the grid. Secondly, label and trace the candidate edge grids by a neighborhood analysis method. During edge tracing, the direction-based single-edge suppression and the distance-based connection adjustment method are developed to improve the tracing results. Then, extract true boundary points from original building point clouds based on grid tracing results. Finally, a random sample consensus estimation and least squares fitting method is used to regularize boundary lines. Experimental results show that the proposed method can quickly extract boundary directly from building point clouds and provide a feasible solution for the automatic extraction of building boundary.