UKF和小波变换结合的车辆跟踪算法的研究

发布时间：2018-03-06 04:18

本文选题：车辆跟踪　切入点：无迹卡尔曼滤波　出处：《安徽理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：在经济高度发展的今天,汽车作为一种更便捷化的交通工具已经逐渐代替原有的电瓶车,成为人们出行的代步工具。但是有一个严峻的问题摆在面前,那就是交通问题,车辆繁多导致了交通路况的复杂度升高,并且在这样复杂的交通路况中极易发生交通事故。因此建立起智能交通系统成为当今社会迫在眉睫的事情,而车辆的目标跟踪又是智能交通系统的一个重要部分,对减少交通阻塞和交通事故的发生有着重大的意义。本文采用无迹卡尔曼滤波与小波变换结合的算法来实现车辆的跟踪,首先在理论上通过对比了几种卡尔曼跟踪算法的优缺点,选择了无迹卡尔曼滤波算法作为目标跟踪的算法,并且通过MATLAB进行仿真,验证了无迹卡尔曼滤波算法的优越性,为了提高车辆跟踪的精度,在无迹卡尔曼滤波算法的基础上,加入小波变换的多尺度分析,其原理是:对处理的信号在尺度N上进行多尺度分解,再对分解在不同尺度i上的平滑信号进行无迹卡尔曼滤波处理得到平滑信号的最优估计,然后把得到的平滑信号的最优估计和细节信号进行重构,得到尺度i上的最优估计,接着把不同尺度上的最优估计进行融合,从而得到整个系统的最优估计,并且通过MATLAB对其进行了仿真。最后对两种跟踪算法仿真的结果进行分析,从而验证了无迹卡尔曼滤波与小波变换结合的算法在车辆跟踪中的误差较小,估计的精度更高。在硬件方面为了适应雷达信号的高速采集,系统采用FPGA作为硬件控制芯片。整个系统的硬件设计部分包括三个模块:数据采集模块,数据处理模块,显示模块。每个模块的衔接流程为:首先通过激光雷达获取车辆的位置和速度信息,然后通过A/D转换电路实现从模拟信号到数字信号的转换,再将转换后的数字信号送入FPGA中进行数据的处理,最后对处理后的最优估计值进行显示。在系统硬件设计的基础上,使用Verilog语言对系统的每个硬件接口进行编程,同时对算法也进行软件的编程。最后对搭建的硬件平台,通过Quartus ii软件平台进行下载程序和调试。
[Abstract]:In today's highly developed economy, cars, as a more convenient means of transportation, have gradually replaced the original electric vehicles and become a means of transportation for people to travel. However, there is a serious problem in front of them, that is, the traffic problem. The complexity of traffic conditions is increased due to the large number of vehicles, and traffic accidents are easy to occur in such complex traffic conditions. Therefore, it is urgent to establish an intelligent transportation system in today's society. Vehicle target tracking is an important part of Intelligent Transportation system, which has great significance to reduce traffic congestion and traffic accidents. In this paper, the unscented Kalman filter and wavelet transform algorithm are used to achieve vehicle tracking. Firstly, by comparing the advantages and disadvantages of several Kalman tracking algorithms in theory, the unscented Kalman filter algorithm is chosen as the target tracking algorithm, and the superiority of the unscented Kalman filter algorithm is verified by MATLAB simulation. In order to improve the accuracy of vehicle tracking, the multi-scale analysis of wavelet transform is added on the basis of unscented Kalman filter algorithm. The principle is that the processed signal is decomposed on scale N. The smooth signal decomposed on different scales I is processed by unscented Kalman filter to obtain the optimal estimation of the smooth signal, and then the optimal estimation of the smooth signal and the detail signal are reconstructed to obtain the optimal estimation on the scale I. Then the optimal estimation on different scales is fused to get the optimal estimation of the whole system, and the simulation is carried out by MATLAB. Finally, the simulation results of the two tracking algorithms are analyzed. It is proved that the algorithm combined with unscented Kalman filter and wavelet transform has less error and higher estimation precision in vehicle tracking. In order to adapt to the high-speed acquisition of radar signal in hardware, The hardware design of the whole system includes three modules: data acquisition module, data processing module, data processing module. Display module. The link flow of each module is as follows: first, the position and speed information of vehicle is obtained by lidar, and then the conversion from analog signal to digital signal is realized by means of A / D conversion circuit. Then the converted digital signal is sent into FPGA for data processing, and the optimal estimated value after processing is displayed. On the basis of the hardware design of the system, every hardware interface of the system is programmed with Verilog language. At the same time, the algorithm is also programmed. Finally, the hardware platform is downloaded and debugged through Quartus II software platform.
【学位授予单位】：安徽理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN713

【相似文献】