一种高效的二进制点云局部特征描述算法
An Efficient Local Feature Description Algorithm for Binary Point Cloud
刘旺 1陈燚涛 1刘芳1
作者信息
- 1. 武汉纺织大学机械工程与自动化学院,湖北 武汉 430200
- 折叠
摘要
针对现有点云局部特征描述算法时效性和紧凑型不能满足实际应用需求的问题,提出一种二进制多层切片距离特征描述算法(B-MSD).首先在特征点处建立了一种稳健的局部参考坐标系;然后设计了两种多层切片和子区域划分方式并计算平均距离,以提取点云局部三维形状特征;最后通过量化的方法将各子区域计算的平均距离转化为二进制码串,串联所有码串生成最终的二进制描述子.在多个公开点云数据集上与几个经典描述子进行仿真对比,结果表明,在鉴别力更强的基础上,上述算法的紧凑性、计算效率和匹配速度方面综合表现更好.3D目标检测和点云配准应用实验也验证了上述算法的有效性.
Abstract
Aiming at the problem that the timeliness and compactness of existing point cloud local feature descrip-tion algorithms cannot meet the practical application requirements,a binary multi-slice distance feature description algorithm(B-MSD)is proposed.Firstly,a robust local reference coordinate system was established at the feature points;Then,two multi-slice and sub-region division methods were designed and the average distance was calculated to extract the local three-dimensional shape features of point cloud.Finally,the average distance calculated by each sub-region was converted into binary strings by quantization,and all strings were concatenated to generate the final binary descriptor.Compared with several classical descriptors on several public point cloud datasets,the simulation results show that the proposed algorithm performs better in terms of compactness,computational efficiency and matc-hing speed on the basis of stronger discrimination.In addition,the application experiments of 3D target detection and point cloud registration also verify the effectiveness of the proposed algorithm.
关键词
点云局部特征/二进制描述子目标检测/点云配准Key words
Point cloud local feature/Binary descriptors object detection/Point cloud registration引用本文复制引用
出版年
2024
NETL
NSTL
万方数据