基于AMR的交通信息检测算法研究

发布时间：2018-03-27 05:00

本文选题：交通信息检测　切入点：AMR　出处：《长安大学》2015年硕士论文

【摘要】：目前,以车辆拥堵为代表的交通问题已成为制约城市发展的重大因素之一,而智能交通系统作为解决交通问题的有效途径,其重要性日益凸显。如何获取实时、准确的交通路网信息,为交通系统提供有力的数据支持,将是建设智能交通系统的重要前提。因此,本文设计了基于各向异性磁阻传感器(AMR,Anisotropic Magneto Resistance)的交通信息检测系统,并提出了实时基线跟踪的车辆检测算法与改进的有向无环图支持向量机(DAG-SVM,Directed-Acyclic-Graph SVM)车型分类算法,可以有效提高道路车流量、车速、车道占有率、车辆类型检测的准确性。本文主要工作包括:(1)研究AMR地磁感应车辆检测原理,结合交通信息获取需求,设计了基于AMR的交通信息检测系统。系统分为前端信息采集子系统与信息监控处理子系统,前者由传感器节点与信号接收机组成,负责地磁车辆特征信息的检测,后者利用检测信号特征计算获取交通信息。(2)分析AMR采集的原始信号特点,对地磁车辆特征信息检测进行研究,提出了实时基线跟踪的车辆检测算法。该算法以多状态机判断为主体,设计加权函数实时基线修正,能够实时有效检测车辆信号,提高了车流量、车道占有率以及车速计算的准确率。并采用Filter-Filter-Wrapper特征选择与优化模型,通过实验,获取有利于车型分类的特征集,作为有效地磁车型特征。(3)根据提取的有效地磁车型特征,选择DAG-SVM方法进行车型分类,论文分析了DAG-SVM模型结构的“误差累积”,提出了改进的DAG-SVM车型分类方法,该方法结合特征权值计算类间距离,利用类间距离搭建有向无环图。(4)搭建系统测试平台,现场测试车辆检测算法,并获取样本集,离线训练并测试车型分类算法,测试结果表明,车辆检测准确率达到98%以上,车型分类准确率为75%以上,改进的DAG-SVM具有更好的分类效果。本文设计的基于AMR交通信息检测算法,能够有效提高交通信息获取的准确率,具有较高的推广价值与应用前景。
[Abstract]:At present, traffic problem represented by vehicle congestion has become one of the major factors restricting urban development. As an effective way to solve traffic problems, Intelligent Transportation system (its) is becoming more and more important. Accurate traffic network information, providing strong data support for traffic system, will be an important prerequisite for the construction of intelligent transportation system. Therefore, a traffic information detection system based on anisotropic magnetoresistive sensor (AMRN) is designed in this paper. The vehicle detection algorithm based on real-time baseline tracking and the improved DAG-SVM- Directed-Acyclic-Graph SVM-based vehicle classification algorithm based on directed acyclic graph support vector machine are proposed, which can effectively improve the traffic flow, speed and lane occupancy. The main work of this paper includes: 1) studying the principle of AMR geomagnetic induction vehicle detection, combining with traffic information to obtain the demand. The traffic information detection system based on AMR is designed. The system is divided into front-end information acquisition subsystem and information monitoring and processing subsystem. The former is composed of sensor node and signal receiver, which is responsible for the detection of geomagnetic vehicle characteristic information. The latter uses the characteristic calculation of detection signal to obtain traffic information. (2) analyzing the characteristics of the original signal collected by AMR, and studying the detection of the characteristic information of geomagnetic vehicle. A vehicle detection algorithm based on real-time baseline tracking is proposed, in which the multi-state machine judgment is the main body, and the real-time baseline correction of the weighting function is designed, which can effectively detect the vehicle signals in real time and improve the traffic flow. Using the Filter-Filter-Wrapper feature selection and optimization model, through experiments, the feature set is obtained as the effective geomagnetic vehicle type feature, which can be used as the effective geomagnetic vehicle type feature. This paper analyzes the "error accumulation" of the DAG-SVM model structure, and proposes an improved DAG-SVM vehicle classification method, which combines the characteristic weights to calculate the distance between classes. Using distance between classes to build the directed acyclic graph. 4) to build the system test platform, to test the vehicle detection algorithm on the spot, to obtain the sample set, to train off line and to test the vehicle classification algorithm. The test results show that the accuracy of vehicle detection is over 98%. The accuracy of vehicle classification is more than 75%, and the improved DAG-SVM has better classification effect. The traffic information detection algorithm based on AMR designed in this paper can effectively improve the accuracy of traffic information acquisition, and has high promotion value and application prospect.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491.116

【相似文献】