基于RANSAC平面检测种子点的机载屋顶点云提取方法

An airborne rooftop point cloud extraction method based on RANSAC plane detection for seed points

机载激光雷达技术已经成为快速获取城市建筑三维数字模型的有效手段,而建筑物屋顶点云提取则是建筑物三维数字模型重建的关键。针对该问题,本文提出了一种基于随机抽样一致算法(RANSAC)平面检测种子点的机载屋顶点云提取方法。首先,通过近地点分离将地面点从点云中分离出来,保留建筑物点云和少量树冠点云。然后,利用RANSAC平面检测选取种子点,这些种子点几乎全部为屋顶点,且非屋顶点比例极低。接下来,将种子点作为初始增长点,利用种子点与其邻域点法线夹角和Z方向上的距离差值作为聚类特征,进行屋顶点提取。实验结果表明,该方法在不同数据集上取得了良好的屋顶点提取效果。点云密度对屋顶点提取结果有一定影响,较高的点云密度有利于提取几何特征明显的屋顶点。此外,通过该方法选取的种子点准确性较高,非屋顶点的影响非常有限。综上,该方法能够有效地提取机载点云中的屋顶点,为建筑物三维重建和城市规划等应用提供了重要的数据支持。

Airborne LiDAR technology has become an effective means to quickly acquire three-dimensional(3D)digital models of urban buildings,and the extraction of rooftop point clouds is the key to the reconstruction of 3D digital models of buildings.To solve this problem,an airborne rooftop point cloud extraction method based on random sampling consensus algorithm(RANSAC)plane detection for seed points was proposed.Firstly,the ground point was separated from the point cloud by perigee separation,retaining the point cloud of buildings and a small amount of canopy point cloud.Secondly,the RANSAC plane detection was used to select seed points,which were almost all rooftop points,and the proportion of nonrooftop points was very low.Finally,the seed point was taken as the initial growth point,and the normal angle between the seed point and its neighborhood point and the distance difference in the Z direction were used as the clustering features to extract the rooftop point.Experimental results show that the proposed method achieves good results in rooftop point extraction on different data sets.Point cloud density has a certain influence on the extraction results of rooftop points,and higher point cloud density is conducive to the extraction of rooftop points with obvious geometric features.In addition,the seed points selected by this method have high accuracy,and the influence of non-rooftop points is very limited.In summary,this method can effectively extract rooftop points in airborne point clouds and provide important data support for its applications in the 3D reconstruction of buildings and urban planning.