基于车检器及收费数据融合的高速公路异常状态识别研究

发布时间：2018-07-15 10:53

【摘要】：异常状态识别是高速公路运管部门进行运营管控、路况信息发布和交通诱导的基础，对减少交通事故造成的人身伤亡、财产损失和避免二次交通事故等方面具有重要的作用。随着智能交通技术的发展和交通信息检测手段的增加，如何利用交通信息检测手段进行异常状态自动识别引起了广泛关注。因此，研究高速公路异常状态自动识别的关键技术对提升高速公路管理水平和服务水平具有重要的理论和实际意义。本文重点针对目前高速公路车辆检测器布设数量不足导致算法实际应用效果较差的问题，从现有高速交通信息检测手段出发，分析高速公路收费数据，提出了基于车检器和收费数据融合的异常状态识别方法。首先重点分析了基于车检器的交通状态参数对异常状态的灵敏度，同时考虑多种数据源进行异常状态识别，采用收费数据建立异常状态识别方法。最后利用信息融合，建立基于车检器及收费数据融合的异常状态识别方法，解决了单一数据源异常状态识别效果较差的问题。论文的主要研究工作如下： ①交通状态参数对异常状态的灵敏度分析。本文对异常状态下交通状态参数进行灵敏度分析，为基于车检器数据的ACI算法参数选择奠定基础。灵敏度分析主要考虑了流量和异常交通状态严重程度两个因素的影响，通过仿真表明，交通状态参数可灵敏的反映交通状态的变化。 ②针对现有高速公路车辆检测器布设数量不足，本文从多源交通信息采集方式出发，根据高速公路收费数据特征，建立基于收费数据的异常状态识别方法。由于算法性能受到样本车辆数影响，因此本文对算法进行改进，并选取实际数据进行验证，结果表明改进算法在低流量情况下的状态识别性能有所提高。 ③针对基于单一数据源的异常状态识别可能存在可信度低、实际应用效果较差等问题，本文考虑了基于不同数据源融合进行异常状态识别。本文将算法表决融合方法引入异常状态识别领域，建立基于算法表决融合的异常状态识别方法，，算法主要包括3部分：基于车检器数据的ACI算法模块、基于收费数据的ACI算法模块和算法表决融合模块。最后，本文采用实际数据进行验证，结果表明，相比基于单一数据源的状态识别算法，融合算法的在异常状态识别精度方面具有较好的性能。
[Abstract]:The recognition of abnormal state is the basis of highway transportation and management department for operation control, road condition information release and traffic guidance. It plays an important role in reducing casualties caused by traffic accidents, property losses and avoiding secondary traffic accidents. With the development of intelligent transportation technology and the increase of traffic information detection methods, how to use traffic information detection means to automatically identify abnormal state has attracted much attention. Therefore, it is of great theoretical and practical significance to study the key technology of automatic identification of abnormal state of expressway to improve the level of expressway management and service. In this paper, aiming at the problem that the number of vehicle detectors in freeway is not enough and the practical application effect of the algorithm is poor, this paper analyzes the toll data of freeway from the existing means of high-speed traffic information detection. An abnormal state recognition method based on vehicle detector and toll data fusion is proposed. Firstly, the sensitivity of traffic state parameters based on vehicle detector to abnormal state is analyzed. At the same time, several data sources are considered to identify abnormal state, and the method of abnormal state identification is established by using toll data. Finally, using information fusion, a method of abnormal state recognition based on vehicle detector and toll data fusion is established, which solves the problem of poor recognition effect of single data source. The main work of this paper is as follows: 1 sensitivity analysis of traffic state parameters to abnormal state. In this paper, the sensitivity analysis of traffic state parameters in abnormal state is carried out, which lays a foundation for the parameter selection of ACI algorithm based on vehicle detector data. The sensitivity analysis mainly considers the influence of two factors, the volume of traffic and the severity of abnormal traffic state, and the simulation results show that, Traffic state parameters can reflect the change of traffic state sensitively. (2) in view of the insufficient number of existing highway vehicle detectors, this paper starts from the multi-source traffic information collection method, according to the characteristics of highway toll data. A method of identifying abnormal state based on toll data is established. Because the performance of the algorithm is affected by the sample number of vehicles, this paper improves the algorithm, and selects the actual data to verify the algorithm. The results show that the improved algorithm can improve the performance of state recognition in the case of low traffic. 3 in view of the problems of low reliability and poor practical application of abnormal state recognition based on single data source, In this paper, anomaly state identification based on fusion of different data sources is considered. In this paper, the algorithm of voting fusion is introduced into the field of abnormal state recognition, and the algorithm of abnormal state recognition based on algorithm voting fusion is established. The algorithm mainly includes three parts: ACI algorithm module based on vehicle detector data. ACI algorithm module and algorithm voting fusion module based on charging data. Finally, the actual data is used to verify the proposed algorithm. The results show that the fusion algorithm has better performance than the state recognition algorithm based on a single data source in terms of the accuracy of abnormal state recognition.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U495;U491.116

【参考文献】