基于动态椭圆的LiDAR数据窗户边界提取研究

Study on Window Boundary Extraction Method of Introducing Dynamic Ellipse Based on Vehicle LiDAR

针对复杂建筑物立面中窗户精细提取的难题,该文基于车载LiDAR数据提出一套完整的不规则窗户边界提取方法:先通过RANSAC算法探测建筑物主墙面点云,借助语义特征分离墙面和非墙面点云;再通过坐标变换,将非墙面点云转换到水平面,采用格网对水平面内点集的邻域关系进行判断,使用平面八邻域连通性探测方法对二维平面内窗户点进行聚类,分别存储聚类后的每一个窗户点云;然后采用改进的动态椭圆凸壳算法,检测聚类后各窗户边界轮廓点;最后对窗户点云坐标逆变换,得到建筑物立面中的窗户边界。通过实验验证了该算法的准确性。

Focusing on the difficult problem of extracting window in the building facade,a complete set of irregular window boundary extraction method based on vehicle-borne LiDAR is presented in this paper:Firstly,the RANSAC algorithm is used to detect the main wall point cloud of the building,the wall surface and the non wall point clouds are separated with the help of semantic features.Then the non wall point cloud is converted to the horizontal plane by coordinate transformation,and the neighborhood of the point set in the horizontal plane is judged by grid.The plane eight neighborhood connectivity detection method is used to cluster window points inside the two-dimensional plane.Each window point cloud after clustering is stored respectively.Then the improved dynamic elliptical convex hull algorithm is used to detect the contour points of the window boundary.Finally,the window point cloud is converted by coordinate inversion,and the window boundary in the building facade is obtained.The experiment verifies the accuracy of the algorithm.