基于出租车GPS数据的区域路网交通流状态演化识别方法研究

发布时间：2018-06-23 06:11

本文选题：交通工程 + 区域路网　；参考：《长安大学》2017年硕士论文

【摘要】：面对机动车保有量增长速度远高于交通基础设施建设速度所带来的交通问题,利用ITS技术解决交通症结是必然的发展趋势,而对于实时和未来交通流状态的准确把握是发挥ITS技术的基础。城市道路交通系统是一个复杂系统,具有很强的随机性和动态性,同时具有较强的规律性和联系性。据此,本文以西安市部分区域为例,以出租车GPS数据为基础,研究交通流状态识别方法,挖掘交通流状态演化特性,从而掌握实时、准确的交通流运行状态,判别常发性和偶发性交通拥堵,为区域交通管理与控制提供依据,确保城市区域交通安全、顺畅运行。基于此,论文对以下几个方面进行了研究:(1)交通参数估计方法研究本文以大量的出租车GPS数据为研究基础数据,由于无法通过原始数据提供的地点速度、GPS时间、经纬度坐标、方向、车辆状态和数据有效性等信息直接准确的判别交通流状态,因此提出了利用这些信息估计交通参数,从而间接获得更多的有效信息。在对采集到的出租车GPS数据进行了预处理、电子地图匹配等准备工作的前提下,构建了路段平均车速、路段平均延误和交叉口平均延误等交通参数估计模型,为后续的研究工作奠定了基础。(2)基于SVM的交通流状态识别方法研究对传统的SVM二分类算法进行改进,构建了基于SVM二叉树多分类算法的交通流状态识别模型,并以估计出的三个交通参数作为输入数据,通过多次训练标定模型参数,实验证明该模型能够解决非线性、高维数的问题,实现对城市区域路网的交通流状态的准确识别。(3)交通流状态演化特性分析方法研究以上述研究识别的交通流状态为基础,通过构建基于马尔可夫模型的交通流状态迁移网络模型和交通流状态演化特性分析模型,挖掘城市区域路网在一定时间段内交通流状态的迁移特性、稳定性、偏好性、活跃性、活跃时间、跳跃迁移、堵塞路段重叠率和时空分布等特性,从而更加深入地了解城市区域的整体运行规律,为城市区域交通管理与控制方案的制定提供依据。(4)交通流状态识别、演化特性的应用研究结合交通参数估计方法、交通流状态识别方法和交通流状态演化特性分析方法,以西安市高新区部分区域路网、出租车GPS数据和路段视频数据为实验基础,验证本研究构建的模型的准确性、可靠性和实用性,并分析该区域内路网交通流状态的演化过程及特性,判断常发性和偶发性交通拥堵。
[Abstract]:In the face of the traffic problems caused by the speed of vehicle ownership increasing much faster than the speed of traffic infrastructure construction, it is an inevitable development trend to use its technology to solve the traffic problem. The accurate understanding of the real-time and future traffic flow is the basis of its technology. Urban road traffic system is a complex system, with strong randomness and dynamic, at the same time has strong regularity and connection. On this basis, taking part of Xi'an as an example, based on GPS data of taxis, this paper studies the identification method of traffic flow state, excavates the evolution characteristics of traffic flow state, and grasps the real-time and accurate traffic flow running state. It can provide the basis for regional traffic management and control and ensure the safety and smooth operation of urban traffic. Based on this, this paper studies the following aspects: (1) Traffic parameter estimation method based on a large number of taxi GPS data as the basic data, because the location speed can not be provided through the original data GPS time, latitude and longitude coordinates, The information of direction, vehicle state and data availability can directly and accurately distinguish the traffic flow state. Therefore, it is proposed to use these information to estimate traffic parameters so as to obtain more effective information indirectly. Based on the preprocessing and electronic map matching of the collected taxi GPS data, a traffic parameter estimation model, such as the average speed of the road, the average delay and the average delay at the intersection, is constructed. It lays a foundation for further research work. (2) the traffic flow state recognition method based on SVM improves the traditional SVM two-classification algorithm and constructs a traffic flow state recognition model based on SVM binary tree multi-classification algorithm. The estimated three traffic parameters are used as input data, and the calibration model parameters are trained several times. The experiment shows that the model can solve the nonlinear and high-dimensional problems. To realize the accurate identification of the traffic flow state of the urban road network. (3) the analysis method of the evolution characteristics of the traffic flow state is based on the traffic flow state identified above. By constructing the network model of traffic flow state migration based on Markov model and the analysis model of traffic flow state evolution characteristic, the paper excavates the migration characteristics, stability, preference, activity of traffic flow state of urban regional road network in a certain period of time. The characteristics of active time, jump migration, overlap rate and space-time distribution of blocked road sections, so as to better understand the overall operation law of urban areas, and provide the basis for the formulation of traffic management and control schemes in urban areas. (4) recognition of traffic flow status, The application research of evolution characteristics combined with traffic parameter estimation method, traffic flow state identification method and traffic flow state evolution characteristic analysis method, based on Xi'an High-tech Zone regional road network, taxi GPS data and video data as experimental basis. The accuracy, reliability and practicability of the model are verified, and the evolution process and characteristics of the traffic flow state in the road network are analyzed to judge the frequent and accidental traffic congestion.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491

【参考文献】