基于BP神经网络预测的城区占道停车智能管理系统的设计与实现

发布时间：2018-04-02 17:52

本文选题：占道停车　切入点：智能管理　出处：《北京邮电大学》2015年硕士论文

【摘要】：随着经济的发展和人们消费能力的提高,家用小汽车的数量日渐增长,城市中“停车难”问题越来越严重。现有的大型停车场难以满足停车需求,增加占道停车的方式可以大大缓解这一难题。然而因为占道停车场分布的不规律性及管理体系的不健全,存在个人私行收取费用的状况,且无严格统一的收费准则,管理部门没法及时掌握泊车信息及停车费用,驾驶员无法及时获取所在位置附近的停车场及车位占用情况信息,所以需要构建统一的智能管理系统。此外,国内的大多数停车管理系统中的信息发布模块,仅能显示车位的实时信息,未能提供对短时间内车位变化情况的预测,导致驾驶员到达停车场后的实际车位占用情况可能与在停车场外看到或者查询到的信息差别很大,甚至只能到其他停车场寻找车位,带来诸多不便。首先,本文针对当前国内外普遍使用的停车场系统的情况进行了调研,指出了当前工作中存在的问题并予以剖析。进而讨论了时间序列预测的相关方法,重点讨论了基于神经网络预测的方法,针对BP神经网络预测算法进行了深入的研究。然后,本文对城区占道停车智能管理系统的功能需求进行了分析,并完成了整体的设计工作。第一,为了使管理系统智能化,将系统划分为四个子系统：分别是车位信息采集子系统,手持终端收费子系统,中心管理子系统以及停车诱导及车位预测子系统。在各子系统之间定义了通信格式及交互协议,实现了数据的采集、传输、处理及应用。第二,在停车诱导子系统中引入了车位信息的预测功能,目的是对未来车位的变化情况实现短时预测。通过对时间序列预测的传统方法、时间序列的非线性预测方法以及神经网络预测方法的比较,最终创建了基于BP神经网络的车位信息预测模型,并用实际数据对该模型进行了验证。最后,结合实际项目需求,完成了对本系统的开发和实现。
[Abstract]:With the development of economy and the improvement of people's consumption power, the number of household cars is increasing day by day, and the problem of "parking difficulty" is becoming more and more serious.The existing large parking lot is difficult to meet the parking demand, increasing the parking on the road can greatly alleviate this problem.However, due to the irregular distribution of parking lots and the unsound management system, there is a situation in which private individuals collect fees, and there are no strict and uniform charging criteria, so the management can not grasp parking information and parking fees in a timely manner.The driver can not get the information of parking lot and parking space in time, so it is necessary to construct a unified intelligent management system.In addition, the information release modules in most parking management systems in China can only display the real-time information of parking spaces, and fail to predict the changes of parking spaces in a short period of time.As a result, the actual parking space occupation after the driver arrives in the parking lot may be very different from the information seen or inquired outside the parking lot, even can only look for the parking space in other parking lot, which brings a lot of inconvenience.Firstly, this paper investigates the situation of parking lot system which is widely used at home and abroad, points out the problems existing in the current work and analyzes it.Then, the related methods of time series prediction are discussed, and the methods based on neural network prediction are discussed, and the BP neural network prediction algorithm is studied deeply.Then, this paper analyzes the functional requirements of urban parking intelligent management system, and completes the overall design work.First, in order to make the management system intelligent, the system is divided into four subsystems: parking information collection subsystem, handheld terminal charge subsystem, central management subsystem and parking guidance and parking prediction subsystem.The communication format and interactive protocol are defined among the subsystems, and the data collection, transmission, processing and application are realized.Secondly, the prediction function of parking space information is introduced in the parking guidance subsystem, in order to predict the future parking space changes in a short time.Through the comparison of the traditional methods of time series prediction, the nonlinear prediction methods of time series and the neural network forecasting methods, the vehicle parking information prediction model based on BP neural network is established.The model is validated with actual data.Finally, according to the actual project requirements, the development and implementation of the system is completed.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491.7;TP18

【参考文献】