场景导向的kd-tree点云滤波算法  

Scene-orientedkd-tree filter algorithm for point cloud

作　　者：赵浚壹 马峻[1,2] 陈寿宏 郭玲[1,2] 徐翠锋 ZHAO Junyi;MA Jun;CHEN Shouhong;GUO Ling;XU Cuifeng(School of Electronic Engineering and Automation,Guilin University of Electronic Science and Technology,Guilin Guangxi 541004,China;Guangxi Key Laboratory of Automatic Testing Technology and Instrument,Guilin Guangxi 541004,China)

机构地区：[1]桂林电子科技大学电子工程与自动化学院,广西桂林541004 [2]广西自动检测技术与仪器重点实验室,广西桂林541004

出　　处：《激光杂志》2021年第11期74-78,共5页Laser Journal

基　　金：国家自然科学基金(No.61671008)、桂林电子科技大学研究生科研创新项目(No.2020YCXS097)。

摘　　要：采集点云是三维重建过程中的关键步骤,在采集点云的过程中,不可避免地产生一些噪声及离群点。针对噪声及离群点的传统滤波算法主要依赖于概率学模型假设,然而由于环境的复杂性导致噪声及离群点的分布并不完全服从于假设的模型,从而传统的滤波算法不能达到良好的滤波效果。另外,传统滤波算法通常需要对样本逐个遍历,因此耗时较高。针对这些问题,针对特定场景的结构特点,提出了一种场景导向的kdtree(k-dimensional tree)点云滤波算法。首先对点云下采样后计算其重心,再设定搜索半径阈值,最后依据所计算得到的重心及搜索半径结合kd-tree分割出场景结构并保留,从而达到滤波目的。实验结果表明,提出的算法不仅具有良好的滤波效果,而且在算法的处理速度方面,相较于传统的半径滤波算法、统计滤波算法分别提高了4.8倍、14.2倍。Obtaining point cloud data is a key step in the 3d reconstruction process.However,in the process of collecting point cloud,some outliers and noise are inevitably produced.Previous filtering algorithms for noise and outliers mainly rely on probabilistic model assumptions.However,the distribution of noise and outliers is not completely subject to models assumed previously due to complexity of environment,which then lead to traditional methods work invalid.Additionally,traditional algorithms usually need to iterate over the samples one by one;therefore,the traditional algorithms are hugely time-consuming.In order to solve these problems,this paper focuses on the structural characteristics of specific scenes,and proposes scene-oriented k-dimensional tree(kd-tree)filter algorithm for point cloud.Firstly,the center of gravity of the original point cloud are calculated after down sampling,and then the threshold of search radius is set.Finally,according to the calculated center of gravity and search radius,the scene structure is segmented and retained via kd-tree.Experimental results show that the proposed algorithm can not only filter effectively,but also improve the processing speed by 4.8 times and 14.2 times respectively,compared with the traditional Radius-Outlier-Removal and Statistical-Outlier-Removal filtering algorithm.

关 键 词：点云滤波 降噪 离群点 KD-TREE 三维重建 

分 类 号：TN249[电子电信—物理电子学]