基于PCA-TrICP的多源点云无缝三维重建

针对三维重建时使用单一数据源的局限性问题,融合异源的倾斜影像和三维激光点云进行三维实景重建。在前期工作中,利用倾斜摄影技术对校园图书馆建立初始模型,生成影像点云,并使用三维激光扫描仪获取其立面数据,完成去噪及拼接等预处理工作。针对预处理过后的影像点云和激光点云数据的配准,首先使用主成分分析方法对数据降低维度,对齐跨源点云的姿态方向,并利用数据的空间几何中心匹配得到初始位姿;然后使用TrICP算法完成低重叠率的跨源点云的精确配准。实验结果表明,本文粗配准方法的精度达到0.32 m,远高于快速点特征直方图算法精度;在精配准的精度分析中,建筑物立面的独立精度为0.03 m、0.08 m、0.22 m,整体精度相对于传统ICP算法提高了53.5%,达到了0.08m。

Seamless 3D reconstruction of multi-source point cloud based on PCA-TrICP

Aiming at the limitation of using a single data source for 3D reconstruction,this paper combines heterogeneous tilt images and 3D laser point cloud for 3D real scene reconstruction.In the preliminary work,the initial model of the campus library was established by using oblique photography technology to generate image point clouds,and its facade data was acquired by using a three-dimensional laser scanner to complete the preprocessing work such as denoising and splicing.For the registration of preprocessed image point cloud and laser point cloud data,firstly,the principal component analysis method is used to reduce the dimension of the data,align the pose direction of the cross-source point cloud,and obtain the initial pose by using the spatial geometric center matching of the data.Then,TrICP algorithm is used to complete the accurate registration of the cross-source point cloud with low overlap rate.The experimental results show that the accuracy of the coarse registration method in this paper is up to 0.32 m,which is much higher than that of the fast point feature histogram algorithm.In the precision analysis of precise registration,the independent precision of the building facade is 0.03 m,0.08 m,0.22 m,and the overall precision reaches 0.08 m,which is 53.5 % higher than the traditional ICP algorithm.