基于KNN聚类的三维点云去噪方法研究  

Research on 3D Point Cloud Denoising Method Based on KNN Clustering

作　　者：王子硕 郭育畅 赖天翔 高兴泉[1,2] WANG Zishuo;GUO Yuchang;LAI Tianxiang;GAO Xingquan(College of Information and Control Engineering,Jilin University of Chemical Technology,Jilin 132022,China;College of Information Engineering,Jilin Polytechnic University,Jilin 132022,China)

机构地区：[1]吉林化工学院信息与控制工程学院,吉林132022 [2]吉林工业职业技术学院信息工程学院,吉林132022

出　　处：《长江信息通信》2024年第12期78-80,98,共4页Changjiang Information & Communications

摘　　要：针对三维点云的随机高斯噪声问题,提出了一种基于层次的K近邻(K-Nearest Neighbor, KNN)聚类的三维点云去噪方法。该方法首先通过KNN寻找每个数据点周围最近的K个邻居,根据邻居的数值对三维点云数据的随机高斯噪声进行初步去噪处理,以减少孤立点的影响。然后,利用基于距离的聚类方法将三维点云数据按照距离进行分组,对点云数据进行进一步去噪,识别主要簇并剔除次要簇,以提高三维点云去噪精度。实验结果表明,KNN聚类算法的峰值信噪比提高了4~15dB,在去噪效果和保留点云细节方面均表现出优势,既能够较好的去除三维点云噪声,又能够保留点云的几何特征和细节信息,并且在三维点云去噪效果上表现出更高的性能和鲁棒性。In order to solve the problem of random Gaussian noise in three-dimensional point clouds,a three-dimensional point cloud denoising method based on hierarchical K-Nearest Neighbor(KNN)clustering is proposed.Firstly,KNN is used to find the nearest K neighbors around each data point,and the random Gaussian noise of 3D point cloud data is preliminarily denoised according to the neighbor values,so as to reduce the influence of isolated points.Then,the distance-based clustering method is used to group the 3D point cloud data according to the distance,further de-noising the point cloud data,identifying the main cluster and eliminating the secondary cluster to improve the 3D point cloud de-noising accuracy.Experimental results show that the peak signal-to-noise ratio of KNN clustering algorithm is increased by 4~15dB,showing advantages in denoising effect and preserving point cloud details.It can not only better remove three-dimensional point cloud noise,but also retain the geometric features and details of point cloud,and show higher performance and robustness in three-dimensional point cloud denoising effect.

关 键 词：KNN聚类算法 三维点云 随机噪声 峰值信噪比 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]