摘要
毫米波雷达交通监测场景中待检测目标较多、各目标间点迹特征接近,导致点迹凝聚精度低,对此,文中提出一种改进的交通监测毫米波雷达数据预处理方法。首先通过短时多帧数据积累提高车辆目标点迹密度,随后利用加权欧式距离度量点间距离以提高密集间隔目标的类间距离,并对点间距离分布进行曲线拟合实现聚类算法参数的自适应求解,最后利用基于密度的噪声空间聚类(DBSCAN)算法对点迹进行凝聚处理。由雷达实测数据进行实验验证,相较于传统方法,原始点迹数据经凝聚后跟踪得到车流量统计精度提高10.97%,结果表明所提方法能够对车辆点迹信息进行较为精确的凝聚,改善了毫米波雷达在交通监测领域的应用效果。
Aiming at the problem of low accuracy of plots centroid caused by the multi-target and similar feature in the application of traffic monitoring millimeter-wave(MMW)radar,an improved radar data pre-processing method is proposed.The vehicle plot’s density is enhanced by merging short-time and multi-frame data.Calculating the weighted distance between each plot to improve close spacing targets’distance and the curve fitting of the distance distribution is carried out to realize the algorithm parameter configuration.Finally,Density-Based Spatial Clustering of Application with Noise(DBSCAN)algorithm is used to cluster the radar plots.Through the test experiment on radar measured data,the statistical accuracy of traffic flow is improved by 10.97%after the original plot is condensed and tracked,results show that the proposed method can effectively centroid vehicle plots,improve the application effect of MMW radar in traffic monitoring field.
作者
田丰
霍雨佳
符渭波
TIAN Feng;HUO Yu-jia;FU Wei-bo(College of Communication and Information Engineering,Xi'an University of Science and Technology,Xi'an 710054,China)
出处
《微波学报》
CSCD
北大核心
2022年第6期37-42,70,共7页
Journal of Microwaves
基金
陕西省科技计划项目(2020GY-029)。
关键词
交通监测
毫米波雷达
数据预处理
点迹凝聚
聚类算法
traffic monitoring
millimeter-wave radar
data pre-processing
plot centroid
clustering algorithm
