基于散乱点云内部特征的网格重建

Mesh reconstruction of noisy point cloud based on intrinsic property

提出了一种对三维散乱数据点进行可靠重建的算法.通过组合二次误差势能函数和极值曲面,建立了描述点云内部分布特征的贝叶斯概率模型.在迭代收缩进行降噪处理的过程中,同时保持物体的形状特征.对于降噪后的点云,按照表面复杂程度进行自适应的筛选产生新的点集.将一种新的非Delaunay三角化方法应用到筛选点集中,通过空间圆球沿着物体表面不断增长来快速搜寻邻近点,并权衡Delaunay优化准则和尖锐特征度量来构造新的三角形.实验结果表明,该算法能够充分体现点云的网格化细节特征,具有快速、稳定可靠的优点.

A novel algorithm to reconstruct triangle meshes for 3D noisy point cloud was proposed. By integrating quadric error potential function with extremal surface, Bayesian probabilistic model was used to estimate the intrinsic property of point cloud. The algorithm uses an iterative clustering to improve the noise tolerance in geometric accuracy, while preserves the sharp features. After denoising, a new point set was generated by using surface splatting, which decimated adaptively the point cloud. The reconstruction mesh from the new point set adopts non-Delaunay triangulation method, which searches neighboring points by a spatial sphere progessively growing along the surface of object. A new triangle was generated accord- ing to the tradeoff between Delaunay optimum principle and sharp feature measurement. Experimental results show that the object’s detail features are preserved after meshing, and the algorithm is quick and robust.