毫米波雷达点云的密度和划分联合聚类方法  

Joint Clustering Method for Density and Division of Millimeter Wave Radar Point Clouds

作　　者：彭耀霖 李荣冰[1] 何梓君 PENG Yaolin;LI Rongbing;HE Zijun(Navigation Research Center,College of Automation,Nanjing University of Aeronautics and Astromautics,Nanjing 211106,China;The Fourth Military Representative Office of the Air Force Equipment Department in Nanjing,Nanjing 210000,China)

机构地区：[1]南京航空航天大学自动化学院导航研究中心,江苏南京211106 [2]空军驻南京地区第四军事代表室,江苏南京210000

出　　处：《测控技术》2023年第10期60-66,88,共8页Measurement & Control Technology

摘　　要：针对室内人员检测环境毫米波雷达点云数据特性,并考虑多目标点云密集复杂情况,提出一种毫米波雷达点云的密度和划分联合聚类方法。毫米波雷达点云数据具有稀疏、均匀性差的特征。首先采用基于DBSCAN(Density-Based Spatial Clustering of Applications with Noise)改进的参数自适应算法进行密度聚类,并对其存在的无限制密度扩张问题,通过决策树归类,对异常数据簇进行二次划分,保证了数据簇属性的单一性。试验结果表明,改进的密度聚类算法可自适应地寻找聚类过程中所需要的最佳参数并实现聚类,更适应毫米波雷达点云数据的特性,同时结合划分聚类对异常数据进行二次划分,使得聚类效果更加细腻和准确,实现了多目标密集情况下点云数据精准聚类划分的效果。According to the characteristics of millimeter wave radar point cloud data under indoor personnel detection environment,and considering the complex situation of multi-target point cloud density,a joint clustering method for density and division of millimeter wave radar point cloud is proposed.Millimeter wave radar point cloud data has the characteristics of sparsity and poor uniformity.Firstly,an improved parameter adaptive algorithm based on DBSCAN(Density-Based Spatial Clustering of Applications with Noise)is used for density clustering.To address the unrestricted density expansion problem in density clustering,decision tree classification is used to partition abnormal data clusters twice,ensuring the singularity of data cluster attributes.The experimental results show that the improved density clustering algorithm can adaptively find the best parameters required in the clustering process and achieve clustering,which is more suitable for the characteristics of millimeter wave radar point cloud data.At the same time,combined with the partition clustering,the abnormal data is divided twice,making the clustering effect more delicate and accurate,and achieving the effect of accurate clustering division of point cloud data in the case of multi-target density.

关 键 词：毫米波雷达 点云聚类 自适应密度聚类 划分聚类 决策树 

分 类 号：TN958[电子电信—信号与信息处理]