改进的采样一致性点云配准算法  

Improved sampling consistent point cloud registration algorithm

作　　者：王月海[1] 庄志鹏 邢娜 WANG Yue-hai;ZHUANG Zhi-peng;XING Na(School of Information,North China University of Technology,Beijing 100144,China)

机构地区：[1]北方工业大学信息学院,北京100144

出　　处：《计算机工程与设计》2022年第5期1382-1388,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61571323)。

摘　　要：为解决传统迭代最近点(iterative closest point,ICP)算法存在配准效率低等问题,提出一种改进的采样一致性点云配准算法。通过自适应体素网格滤波法对点云进行处理,可以根据点云量级自动修改体素立方体大小,剔除偏差较大的噪点,降低点云数据量级;在快速点特征直方图(fast point features histogram,FPFH)中引入距离的二次函数,降低远距离邻域点的权值,提高近距离邻域点的权值。运用公开数据集Bunny点云数据进行实验的结果表明,该算法相对于传统点云配准算法的配准精度提升了54.65%,配准效率提升了39.39%。运用多组数据验证了该算法的有效性。To solve the problem of low registration efficiency in the traditional iterative closest point algorithm,an improved sampling consistency point cloud registration algorithm was proposed.The point cloud was processed using the adaptive voxel grid filtering method,which automatically modified the size of the voxel cube according to the magnitude of the point cloud,the noise was eliminated with large deviation,and the magnitude of the point cloud data was reduced.The quadratic function of distance was introduced into the fast point features histogram to reduce the weight of the points in the long-distance neighborhood and the weight of the points in the near-distance neighborhood was increased.The results of experiments using the public data set Bunny point cloud data show that compared with the traditional point cloud registration algorithm,the registration accuracy of the proposed algorithm is improved by 54.65%,and the registration efficiency is improved by 39.39%.Multiple sets of data were used to verify the effectiveness of the algorithm.

关 键 词：迭代最近点算法 采样一致性算法 体素网格滤波法 机器人视觉 三维重建 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]