NURBS曲面重构与点云-曲面误差分析

通过对某汽车零件进行光栅扫描、解相和去包裹处理，获取物体三维点云数据，对点云数据进行降噪、精简和网格化处理。然后根据点云曲率分布云图将点云数据分割成11块区域，首先对A区域点云数据进行曲面拟合，生成4×4阶均匀曲面，然后对其余分块点云数据分别进行曲面拟合，最后通过曲面延伸、拼接、倒角、修剪等处理，获取物体NURBS自由曲面，总体点云-曲面误差为0.2645 mm，并且曲面间符合G1相切连续和法向曲率连续，解决了在曲率较大的地方拟合误差较大的问题，提高了曲面的重构精度。

NURBS surface reconstruction and point cloud-surface error analysis

An automotive part was processed by fringe scanning,phase demodulation and unwrapping to obtain 3D points cloud data,and they are processed by noise reduction,data reduction and meshing.Then,the cloud data were divided into 11 regions according to the point cloud curvature distributing nephogram.Firstly,point cloud data of range A were fitted to form a 4×4-order uniform surface.Secondly,the other data left were fitted to surface separately.Finally,the free-form NURBS surface was obtained via surface extension,splicing,chamfering and trimming.The total error of point-surface was 0.2645 mm,and the surfaces have met the criterion of G1 tangent continuation and normal curvature continuation regulations.As a result,the problem of larger fitting error at high curvature has been solved as well as the surface reconstruction accuracy has been improved by using the methods proposed in the article.