基于频率直方图的K邻域稀疏离群点移除算法

K-NEAREST NEIGHBORS SPARSE OUTLIER REMOVAL ALGORITHM BASED ON FREQUENCY HISTOGRAM

在点云预处理阶段,传统的基于k邻域的稀疏离群点移除算法尚存在一些不足。在点云的处理过程中,关于k邻域的大小以及所要滤去的稀疏离群点的噪声阈值方面,没有给出合理的选取方案。通过对散乱点云传统k近邻稀疏离群点移除算法的分析与研究,提出一种基于k邻域平均距离的频率直方图的分析方法,对传统基于k邻域的离群点移除算法进行了改进。通过该方法可以有效选取合理的k值与噪声阈值。该方法通过对散乱点云设置依次增大的k值,生成k邻域平均距离的统计直方图,分析统计直方图来确定k邻域值的适当大小。针对适当的k值,选取合理的噪声阈值对其进行去噪处理。通过这种方法,为稀疏离群点移除算法中k值和噪声阈值的选取提供了理论依据,提高了点云搜索效率的同时有效防止了离群点的过度删除。

At the point cloud preprocessing stage, there are still some deficiencies existing in the traditional sparse outlier removal algorithm based on k-nearest neighbors. In the point cloud processing, there is no proper selection scheme about the size of the k-nearest neighbors and the noise threshold of the sparse outliers which will be eliminated. According to the analysis and research of the traditional k-nearest neighbors sparse outlier removal algorithm of scattered point cloud, an analysis method of statistical histogram based on the k-nearest neighbors average distance is proposed, improving the distribution of the traditional sparse outlier removal based on k-nearest neighbors. This method is able to select the reasonable k value and the noise threshold effectively. This method generates the frequency statistical histogram of k-nearest neighbor＇s average distance and analyzes the statistical histogram so as to determine the appropriate value of k-nearest neighbors by setting the k values of the scattered point cloud in successive increase. According to the proper k value,it selects the reasonable noise threshold and carries out de-noising. Through this method,it provides the theoretical basis for the selection of k value and noise threshold in the sparse outlier removal algorithm, and improves the efficiency of point cloud search and prevents the excessive deletion of outliers at the same time.