城市场景中的异源点云配准方法

Registration method of heterogeneous point clouds in urban scene

为了实现不同传感器的优势互补,还原准确、完整的城市建筑三维模型,提出了一种将激光雷达点云和摄影测量点云进行配准融合的方法。该方法采用由粗到精的策略,并考虑尺度变换。首先,提取两个异源点云中的平面特征作为配准基元,然后构建基于平面/交线的特征描述符来匹配异源点云之间的对应平面,利用对应平面估计尺度、旋转和平移参数并通过优化得到粗配准下的转换参数。精配准使用融合尺度因子的迭代最近点算法对转换参数进行最后的优化。实验结果表明,提出的方法能够有效配准城市场景中的激光雷达点云和摄影测量点云。

In order to complement the advantages of different sensors and restore accurate and complete three-dimensional models of the urban buildings,this paper proposes a method of registration and fusion of LiDAR point cloud and photogrammetry point cloud.This method adopts the strategy from coarse to fine,and considers scale transformation.Firstly,the plane features of two heterogeneous point clouds are extracted as registration primitives,and then the feature descriptors based on plane/intersection line are constructed to match the corresponding planes between heterogeneous point clouds.The corresponding planes are used to estimate the scale,rotation and translation parameters,and the transformation parameters under rough registration are obtained through optimization.Fine registration uses iterative closest point algorithm with scale factors to optimize the transformation parameters.Experimental results show that the proposed method can effectively register LiDAR point clouds and photogrammetry point clouds in urban scenes.