基于多数投票法的扫描点云尖锐特征保留方法

Sharp Features-preserving for Scanned Point Clouds Based on Majority Voting Method

针对散乱点云预处理过程中尖锐特征难以保留的问题,提出一种基于多数投票法的尖锐特征保留方法。首先构建点云的k-d tree拓扑结构,采用主成分分析(PCA)方法来估计点云所有邻域中每个点的法向,结合曲面变化度对散乱点云进行区域划分;然后将多数投票法应用于特征区域,对模糊点进行识别;最后设定阈值,将判定出的噪声和离群点去除。由于点的种类识别精确,所以该方法可以避免出现去除不完全或者过度光顺等问题。实验结果表明:所提方法简单,耗时较短,能在去除噪声的同时保留点云尖锐特征。

This paper introduces a sharp features-preserving method based on the majority voting to solve the problem that’s difficult to preserve the sharp features in preprocessing of scattered point cloud.First,the k-d tree topology of the point cloud is set.Then the normals of each point in all the neighborhood is estimated with principal component analysis,and according surface variation the scattered point cloud is classified.Second,the majority voting method is applied to feature region to identify the fuzzy points.Finally,the threshold value is set to remove the identified noise and outliers.This method can avoid such problems as incomplete or excessive smoothing due to the accurate classification of points.Massive experiments showthat our method is simple,time is short,and the sharp features of the point cloud can be preserved while removing the noise.