系统仿真学报 ›› 2016, Vol. 28 ›› Issue (10): 2329-2335.

• 仿真建模理论与方法 • 上一篇    下一篇

基于K-means聚类的RGBD点云去噪和精简算法

苏本跃1,2, 马金宇1,2, 彭玉升2,3, 盛敏2,3   

  1. 1.安庆师范大学计算机与信息学院,安徽 安庆 246133;
    2.安徽省智能感知与计算重点实验室,安徽 安庆 246133;
    3.安庆师范大学数学与计算科学学院,安徽 安庆 246133
  • 收稿日期:2016-06-10 修回日期:2016-07-14 出版日期:2016-10-08 发布日期:2020-08-13
  • 作者简介:苏本跃(1971-),男,安徽芜湖,博士,教授,研究方向为图形图像处理、虚拟现实、行为识别。
  • 基金资助:
    国家自然科学基金(11471093); 安徽省教育厅自然科学研究项目(KJ2014A142); 安徽省重点实验室开放课题(ACAIM160102)

Algorithm for RGBD Point Cloud Denoising and Simplification Based on K-means Clustering

Su Benyue1,2, Ma Jinyu1,2, Peng Yusheng2,3, Sheng Min2,3   

  1. 1. School of Computer and Information, Anqing Normal University, Anqing 246133, China;
    2. The University Key Laboratory of Intelligent Perception and Computing of Anhui Province, Anqing 246133, China;
    3. School of Mathematics and Computational Science, Anqing Normal University, Anqing 246133, China
  • Received:2016-06-10 Revised:2016-07-14 Online:2016-10-08 Published:2020-08-13

摘要: 针对Kinect等深度相机扫描获取的点云数据数量庞大、噪声较多的问题,提出一种特征保持的点云去噪和精简算法。使用K-D树快速分类点云;通过曲率估计算法得到局部曲面的曲率值;使用K-means聚类算法对点云进行聚类,对每个类中的点,根据点到聚类中心的欧式距离和邻近点曲率变化判断是否为噪声点;通过保持特征的点云精简算法实现对点云数据的简化。实验结果显示,算法快速有效,对于去除大量外部噪声有良好效果,且精简后的点云数据保持了原始点云特征

关键词: K-means聚类, 点云去噪, 点云精简, RGBD数据

Abstract: Aiming at the problem that the point cloud data scanned by Kinect or other depth camera have a huge number and more noise, a feature preserving method for point cloud denoising and simplification was proposed. This algorithm classified the point cloud rapidly by K-D tree; find The corresponding surface curvature values were obtained using curvature estimation algorithm; The K-means clustering algorithm for point cloud clustering was used. For each point in the cluster, the Euclidean distance was depended on between the point and center of the cluster and the change of the near points curvature to determine whether the noise points. The point cloud data was simplified by the feature preserving method. The experimental results show that the denoising and feature preserving point cloud simplification method is quickly and efficiently, for the removal of a large number of external noise has a positive effect, and the streamline point cloud data have the retention of original point cloud features.

Key words: K-means clustering, point cloud denoising, point cloud simplification, RGBD data

中图分类号: