A novel method for multi-layer building three-dimensional( 3D) contour model reconstruction from air-borne LiDAR point clouds is presented. Based on the extraction results and regularization processing of building structure, the multi-layer building roof contour is projected onto the horizontal plane, and three types of adjacent edges is summarized after analyzing the neighboring relationship between roofs with same hierarchy level, then the consistency processing of roof boundaries is carried out with the building structure hierarchy information. Experi-mental results show that the proposed method can automatically reconstruct the building 3D contour model with high correctness and completeness, the regularization error can be reduced significantly, and the adjacent edges are coincident on horizontal plane, the average positional accuracy of building contour inflection points is better than the average interval of LiDAR point cloud.%以机载LiDAR点云数据为研究对象,提出一种新的基于点云数据的多层建筑物三维轮廓模型高精度自动重建方法.在已完成建筑物结构提取及轮廓规则化处理的基础上,利用多层屋顶轮廓在水平投影面内的相邻关系,将各层屋顶中同等级屋顶的相邻关系概括为平行边、不平行且不相交、相交3种相邻形式,结合多层屋顶的层级结构信息对相邻轮廓边界进行一致性处理.实验证明本文方法可以进一步消除多层建筑物各屋顶轮廓的规则化处理误差,使相邻轮廓边界在水平投影面内严格重合,同时重建后建筑物三维轮廓模型的正确性与完整性较高,拐点的定位精度优于激光点平均间距.
展开▼
