Abstract: In order to improve the accuracy and the stability of surface reconstruction with the multi-level partition of unity, an algorithm of multi-level partition of unity implicits based on parameter optimization was proposed. Firstly, analyzed the factors which the radius coefficient α of bounding spheres and the minimum Nmin of the bounding spheres influencing surface reconstruction. And then, the Nmin was adjusted by the condition of approximation implicit function and the α was adjusted by the requests of coverage density in different models. Finally, the method of surface reconstruction was more stable can be achieved by optimal combination selection of Nmin and α.The experiments proved that the method proposed was more stable and accurate in different models.

Key words: coverage density, Multi-level Partition of Unity (MPU), surface reconstruction, parameter optimization

摘要： 为了提高现有MPU曲面重建的稳定性，提出了一种基于参数优化的MPU曲面重建算法。通过分析包围球半径的比例系数α和包围球内点云最小数目Nmin对曲面重建的影响，根据局部隐含数逼近的条件，对Nmin进行调整。同时，结合不同形态特征的点云模型对覆盖密度的要求，对α进行调整，使得α和Nmin达到一个最优组合，进而使得曲面重建算法更稳定。实验结果表明，该算法能快速准确的对不同点云模型进行参数选取，从而得到更理想的曲面重建效果。

关键词: 覆盖密度, 多层次单元划分, 曲面重建, 参数优化