城市快速路交通流多源数据修正方法研究

发布时间：2018-04-24 20:14

本文选题：交通流数据 + 异常数据识别　；参考：《北京交通大学》2014年硕士论文

【摘要】：城市道路基础设施的供给与交通需求之间的矛盾日益突出,交通拥堵问题日益严重。目前,仅仅依靠拓宽道路、增加路网密度等基础设施已经难以解决城市的交通拥堵问题,而智能交通系统的研发是一条可行的途径。先进的交通管理系统是智能交通系统的重要组成部分,其中交通流数据的质量识别与修正是先进的交通管理系统中最基础和关键的组成部分,如何提高城市道路交通流检测数据的精确度和实时性就变得非常重要。论文提出单源数据修正和多源数据修正相结合的方法来提高道路交通流数据的精确度。首先对不同检测数据进行单源数据的质量识别与修正研究。文中基于交通流理论和阈值理论对原始数据进行异常数据识别与分类,之后应用历史趋势法、时间序列法和交通流局部稳定的特性对缺失数据进行修补,并提出改进的埃特金插值算法对错误数据进行修正；其次,论文提出基于数值优化的BP神经网络对多源数据进行融合修正,分别选用BFGS拟牛顿法、正切拟牛顿法、Fletcher-Reeves共轭梯度法、Polak-Ribiere共轭梯度法和Levenberg-Marquardt算法等五种数值优化的方法对BP神经网络进行改进,并分别对其相对误差、运行时间、迭代次数进行了对比分析；最后,论文选取北京市二环路多源检测数据进行实例验证。
[Abstract]:The contradiction between the supply of urban road infrastructure and traffic demand is increasingly prominent, and the traffic congestion problem is becoming more and more serious. At present, it is difficult to solve the problem of urban traffic congestion only by widening roads and increasing road network density, and the research and development of intelligent transportation system is a feasible way. Advanced traffic management system is an important part of intelligent transportation system, in which the quality identification and correction of traffic flow data is the most basic and key part of advanced traffic management system. How to improve the accuracy and real-time of urban traffic flow detection data becomes very important. In this paper, the method of single source data correction and multi-source data correction is proposed to improve the accuracy of road traffic flow data. First, the quality identification and correction of single source data for different detection data are studied. Based on the traffic flow theory and the threshold theory, the original data are identified and classified, and then the missing data are repaired by using the historical trend method, time series method and the local stability of traffic flow. An improved Etkin interpolation algorithm is proposed to correct the error data. Secondly, a BP neural network based on numerical optimization is proposed to modify the multi-source data, respectively, using BFGS quasi-Newton method. The tangent quasi Newton method, Fletcher-Reeves conjugate gradient method, Polak-Ribiere conjugate gradient method and Levenberg-Marquardt algorithm are used to improve the BP neural network, and the relative error, running time and iteration times are compared and analyzed respectively. The paper selects the Beijing second Ring Road multi-source detection data to carry on the example verification.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491

【参考文献】