基于距离阈值和密度区分改进的点云去噪算法  

Point Cloud Denoising Algorithm Based on Distance Threshold and Density Distinction

作　　者：贾朝阳 陈岳坪[1] 苏宇 Jia Zhaoyang;Chen Yueping;Su Yu

机构地区：[1]广西科技大学机械与汽车工程学院

出　　处：《工具技术》2024年第10期147-154,共8页Tool Engineering

基　　金：国家自然科学基金(52165054);广西自然科学基金(2020GXNSFAA1591);2016年广西高校高水平创新团队及卓越学者计划资助项目(桂教人[2016]42号);广西科技大学自然科学资金项目(197315);广西科技大学创新团队支持计划项目(科大科研发[2017]64号)。

摘　　要：为保证去除三维散乱点云数据噪声的同时能够保持模型的局部特征,提出了一种基于距离阈值和密度区分改进的点云去噪算法。通过引进比例因子ω来改进统计滤波的距离阈值表达式,调整ω的大小,在去噪的同时保持点云模型的局部特征。利用K-D Tree方法建立点云模型的拓扑结构,将点云数据分为大密度点和小密度点两类,对大密度点进行体素滤波。将大密度点滤波后的点集与小密度点集整合得到最后的点云模型。实验结果表明,本文算法对不同类型噪声的去噪效果稳定且良好,在去噪的同时能够有效保持点云模型的局部特征。经算法处理后点云数据的精度有较大提升,最高可达75.37%,且对不同强度的噪声模型具有较强的鲁棒性。In order to keep the local features of the model while removing the noise of three-dimensional scattered point cloud data,an improved point cloud denoising algorithm based on distance threshold and density discrimination is proposed.The scale factorωis introduced to improve the distance threshold expression of the statistical filter,and the size ofωis adjusted to maintain the local features of the point cloud model while denoising.The topology structure of the point cloud model is established by K-D Tree method,and the data of the point cloud is divided into two types.High-density points and low-density points,and the high-density points are filtered by voxel.The point set filtered by the high-density points is integrated with the low-density point set to get the final point cloud model.The experimental results show that the proposed algorithm has a stable and good denoising effect on different types of noise,and can effectively maintain the local features of the point cloud model while denoising.After the algorithm processing,the accuracy of point cloud data is greatly improved,up to 75.37%,and it has strong robustness to different intensity noise models.

关 键 词：点云去噪 距离阈值 密度区分 精度提升 

分 类 号：TG50[金属学及工艺—金属切削加工及机床] TH161.1[机械工程—机械制造及自动化]