中文摘要：

在研究曲面局部法向变化量与高斯曲率关系的基础上，提出了一种基于法向变化量的变分辨率曲面重建算法。该算法首先根据用户给定的法向变化量门限，自适应于曲面曲率对点云数据的最小立方体包围盒进行八叉树分割，并在其有效叶节点内进行局部等值面提取，然后在八叉树中不同级别且空间相邻的有效叶节点内，采用垂直投影法将它们间的缝隙进行拼接。对于给定不同的法向变化量门限，该算法可同时完成曲面重建和网格简化两种功能，从而直接得到点云数据的多分辨率模型。应用实例表明了该算法的有效性。

英文摘要：

Based on the study of relationship between local normal change of surface and the Gauss curvature, the a surface reconstruction algorithm with variable resolution. The algorithm splits the smallest cube box that contains all point cloud data sets and self-adapts surface curvature into octotree cells of different sizes according to the user-specified threshold value of a normal change and performs the local isosurface extraction from cubes which intersects adjacent surfaces at different levels. Then the cracks that may occur on adjacent surfaces at different levels were patched by using perpendicular projections, thus reconstructing surfaces with variable resolutions. As for the threshold values set at various normal changes, the algorithm can reconstruct surfaces and simplify meshes simultaneously, thus directly establishing the point cloud data model with multiple resolutions. Its effectiveness is demonstrated by two application instances.