基于FMCW毫米波雷达的车辆轨迹跟踪和车型识别系统

VEHICLE TRACKING AND VEHICLE IDENTIFICATION SYSTEM BASED ON FMCW MILLIMETER WAVE RADAR

彭堉斌

11849196

硕士

计算机技术领域工程

张进

2020-05-30

2020-07-20

哈尔滨工业大学

深圳

中国国民经济的快速发展带来了城市交通量激增的现象，交通堵塞和交通事故频发等问题日益严重。针对交通路口的车辆实时监测管制可以从根源上预防大部分的交通事故，而基于交通基础设施的智能交通监测系统成为了替代人工监控的有效方案。智能交通检测系统能够实时监测道路上的车辆位置和速度，统计区段内的车辆数量，进一步改善交叉口的交通环境，提高交通安全保障。本文从技术层面分析了交通监测领域的研究现状，通过比较基于视频、基于激光雷达和基于毫米波雷达三种技术各自的优缺点，提出了在智能交通检测系统中使用FMCW毫米波雷达技术的优越性。本文实现了基于雷达原始数据获取车辆点云数据的方法，改进了基于椭圆的车辆点云聚类算法，设计了基于车辆点云特征的车辆轨迹判定算法和轨迹修正算法，提出了一种新的车型设定方法并实现可靠的车型判定算法。在车辆点云数据获取算法中，先对雷达原始数据做二维FFT算法获取距离和多普勒二维图谱数据，并通过CFAR峰值检测算法和DOA方向检测算法获取车辆点云的坐标信息。在对车辆点云聚类算法的实现中，统计分析了车辆长宽高和车辆点云分布特征，讨论了二维和三维数据层面上的聚类误差分析。在车辆轨迹跟踪算法中，重点研究了车辆点云分布和移动特征，通过设计针对性的轨迹修正算法提高了结果准确率。在车型分类算法中，对车型判定特征的选择进行可行性分析，基于轨迹跟踪算法实现简单高效的车型分类。本文的车辆轨迹跟踪和车型识别系统基于Python和Pyqt5开发界面化数据分析和显示工具，实现算法开发和应用结果展示，对实际的交通监测场景下的车辆流动进行多次检测，均取得较好的效果。

The rapid development of China's national economy rapidly increases urban traffic volume. The problem of traffic jams and frequent traffic accidents have become increasingly serious. Real-time vehicle monitoring and control for traffic intersections can prevent most traffic accidents from the root, and intelligent traffic monitoring system based on traffic infrastructure has become an effective alternative to artificial monitoring. The intelligent traffic detection system can monitor the vehicle position and speed on the road in real time, count the number of vehicles in the section, further improve the traffic environment of the intersection, and improve the traffic safety guarantee.In this paper, the research status of traffic monitoring is analyzed from the technical level. By comparing the advantages and disadvantages of video based, lidar based and millimeter wave radar based technologies, the advantages of using FMCW millimeter wave radar technology in intelligent traffic detection system are proposed. In this paper, the method of acquiring vehicle point cloud data based on radar raw data is realized, the vehicle point cloud clustering algorithm based on ellipse is improved, the vehicle track determination algorithm and track correction algorithm based on the characteristics of vehicle point cloud are designed, a new vehicle type setting method is proposed and a reliable vehicle type determination algorithm is realized. In the vehicle point cloud data acquisition algorithm, firstly, the radar raw data is used for two-dimensional FFT algorithm to obtain the range and Doppler two-dimensional map data, and the coordinate information of vehicle point cloud is obtained by CFAR peak detection algorithm and DOA direction detection algorithm. In the implementation of vehicle point cloud clustering algorithm, the length, width, height and distribution characteristics of vehicle point cloud are statistically analyzed, and the clustering error analysis on two-dimensional and three-dimensional data level is discussed. In the vehicle tracking algorithm, the focus is on the distribution and moving characteristics of the vehicle point cloud, and the accuracy of the results is improved by the design of targeted trajectory correction algorithm. In the algorithm of vehicle classification, the feasibility analysis of the selection of vehicle classification features is carried out, and the simple and efficient vehicle classification is realized based on the track tracking algorithm.Based on Python and pyqt5, this paper develops an interface data analysis and display tool to realize algorithm development and application result display. It can detect the vehicle flow in the actual traffic monitoring scene many times and achieve good results.

智能交通监测 毫米波雷达 车流量监测 轨迹跟踪 车型分类

intelligent traffic monitoring millimeter wave radar traffic flowmonitoring track tracking vehicle classification

中文

联合培养

南方科技大学

彭堉斌. 基于FMCW毫米波雷达的车辆轨迹跟踪和车型识别系统[D]. 深圳. 哈尔滨工业大学,2020.