基于车牌识别数据的路网OD估计方法研究

发布时间：2018-06-09 01:47

本文选题：OD估计 + 转弯流量　；参考：《清华大学》2015年硕士论文

【摘要】：机动车保有量的迅速增加使得我国城市道路交通矛盾日益显现,道路拥堵时有发生,极大地影响了人们出行的效率和安全。科学合理的交通规划、交通管理措施以及智能交通系统的建设成为缓解城市交通问题的重要手段。机动车OD矩阵描述了交通小区之间机动车的出行特征,能够为交通规划提供必要的信息;另一方面,OD矩阵包含了道路网的车辆行驶情况,为交通管理的各种措施的制定提供准确的依据。本论文首先对卡口所提供的车牌识别数据的原始数据进行处理,得到路口的转弯流量、路段流量、实测旅行时间和部分机动车OD信息。结合交通规划中Fratar法的计算思路,使用部分机动车OD信息进行路网全样OD估计,并将其作为实际路网中基于转弯流量和路段流量的OD估计研究的基础数据。然后,分别计算仅基于转弯流量、基于转弯流量和实测旅行时间以及基于转弯流量和部分机动车OD信息的分配矩阵。其中在使用部分机动车OD信息进行分配矩阵的计算中对缺失数据进行了修复以得到较为完整的出行路径。第三,分别将转弯流量和路段流量作为输入数据,以最小化检测转弯流量与估计转弯流量的偏差和历史OD矩阵与估计OD矩阵偏差之和最小为目标函数,使用广义最小二乘模型进行OD估计。最后,在真实OD已知和真实OD未知的两种不同情况下,使用ND路网进行OD估计方法验证,对利用不同数据进行估计的结果进行评价分析。结果表明,使用转弯流量和部分机动车OD信息进行OD估计能够提高估计的准确性。在OD估计方法的研究之后,对城市实际部分路网进行案例研究,从实际情况方面验证基于车牌识别的OD估计方法的准确性。通过对评价指标MAPE和RMSE的分析,表明本文提出的OD估计方法能够有效地估计出真实OD。最后,在研究时间长度、实测旅行时间的计算和车牌识别数据的进一步挖掘方面提出了本论文以后的研究展望。
[Abstract]:With the rapid increase of the number of motor vehicles, the contradiction of urban road traffic becomes increasingly apparent in China, and road congestion occurs from time to time, which greatly affects the efficiency and safety of people's travel. Scientific and reasonable traffic planning, traffic management measures and the construction of intelligent transportation system (its) become important means to alleviate urban traffic problems. The vehicle OD matrix describes the travel characteristics of vehicles between traffic communities, which can provide necessary information for traffic planning. On the other hand, the OD matrix includes the vehicle driving situation of road network. To provide accurate basis for the establishment of various measures of traffic management. In this paper, the original data of the license plate recognition data provided by the bayonet are processed, and the turning flow, the section flow, the measured travel time and the OD information of part of the vehicle are obtained. Combined with the Fratar method in traffic planning, the whole OD estimation of road network is carried out by using the OD information of some motor vehicles, and it is used as the basic data of OD estimation based on turn flow and section flow in the actual road network. Then, the assignment matrix is calculated based on turning flow only, turning flow and measured travel time, turning flow and partial vehicle OD information. In order to get a complete travel path, the missing data is repaired in the calculation of allocation matrix using partial vehicle OD information. Thirdly, turning flow and section flow are taken as input data respectively. The objective function is to minimize the deviation between the measured turn flow and the estimated turn flow and the sum of the historical OD matrix and the estimated OD matrix deviation, and the objective function is to minimize the deviation between the historical OD matrix and the estimated OD matrix. A generalized least square model is used for OD estimation. Finally, in two different cases where real OD is known and real OD is unknown, the OD estimation method is verified by using ND network, and the results of estimation with different data are evaluated and analyzed. The results show that the accuracy of OD estimation can be improved by using turning flow and partial OD information. After the research of OD estimation method, a case study on the actual part of urban road network is carried out to verify the accuracy of the OD estimation method based on license plate recognition. The analysis of MAPE and RMSE shows that the OD estimation method proposed in this paper can effectively estimate the real ODs. Finally, the future research prospect of this paper is put forward in terms of the length of the research time, the calculation of the measured travel time and the further mining of the license plate recognition data.
【学位授予单位】：清华大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491.1

【相似文献】