顾及几何形态与密度分布的点云压缩方法

Point Cloud Simplification Considering Geometric Morphology and Density Distribution

摘要: 针对传统点云压缩算法中主要应用法向量、曲率等测度来描述目标区域几何信息，易受噪点影响，无法保证压缩精度的问题，提出一种顾及几何形态与密度分布的点云迭代压缩算法。本文首先利用三维高斯核函数对原始点云进行平滑处理，依据点平滑前后平坦度变化分量与密度变化分量来衡量当前点在点云中的重要度，然后迭代去除重要度最低的点，并计算该点影响范围点集，更新其重要度，最终实现点云压缩。采用两份数据对本文算法进行验证，并与三种经典压缩方法进行对比，实验表明，相较于其他三种方法，本算法压缩结果在点分布均匀的同时，能够较好的保留特征点集。

Abstract: In this paper, a point cloud simplification considering geometric morphology and density distri-bution is proposed, in view of difficulty of ensure precision of simplified point cloud that generated by traditional method which mainly taking normal vector or curvature as the measurement of geometric information. First, we use the three-dimensional Gauss kernel function to smooth the original point cloud; the importance of current point is evaluated based on flatness variation component and density variation component. The simplification proceeds and finishes by removing the least important point and updating the importance values progressively. At last, the per-formance of the proposed method is illustrated with two sets of experiment where three classical simplification methods are employed for contrast. The results show that the proposed method can reserve feature points with a uniform distribution.