Abstract: Concerning the scattered and layered characteristic of point cloud acquired by laser scanners, a data reduction and ordering algorithm is proposed. Firstly the spatial index of point cloud is created based on known marked points using a method integrating Octree and 3D R-tree, ensuring fast and correct access to local data and high efficiency of data retrieval. Secondly one axis of the work coordinate system is selected as the projective direction for parameterizing the local data, which is determined by the normal vector of local reference plane. Then along the selected direction the local data is parameterized and the quadratic surface is approximated. Finally the ordered set of reference points is obtained by sampling the quadratic surface through the R-tree’s leaf nodes, making the scattered and layered point cloud be single layered. Application examples show that the algorithm can improve the overall accuracy of the data as well as maintain the details of point cloud, indicating good validity and practicability in the reduction of scattered and layered large-scale point cloud with complex geometric features.

Key words: data reduction, spatial ordering, scattered and layered point cloud, marked point, 3D R-tree