多传感器信息融合方法在智能交通控制中的研究与应用

发布时间：2018-01-05 09:22

本文关键词：多传感器信息融合方法在智能交通控制中的研究与应用　出处：《东北大学》2015年硕士论文　论文类型：学位论文

【摘要】：多传感器信息融合技术,是对源于多种不同传感器检测的不确定信息进行多层次、多方面的综合分析,能够获得比单个传感器更加全面、准确、完整的信息处理过程。信息融合是一类智能化的综合信息分析技术,将各个传感器检测值信息相互补充,并运用计算机的高效计算,分别从时间序列或空间序列对信息进行综合分析,得到检测目标更加准确的数据信息,实现更加精确的识别和决策功能,加强结果信息的可信度与实用性。智能交通控制运用智能的信息处理方法,研究传感器所检测到的车流量数值,计算出信号灯的信号时长策略,由此对路口各方向的交通流实施控制。如何利用信息融合技术获得实时准确的路口车流量信息,是本文的重点研究目标。本文根据实际需求分析了智能交通控制领域信息融合技术的使用情况,针对实时的,分布的,复杂的交通融合环境,建立了在数据层使用最小二乘加权算法处理传感器数据,在决策层使用进化规则改进过的BP网络对车流量做出预测的信息融合模型。为了克服BP神经网络因初始权值随机选取而易陷入局部最小解的问题,本文研究了以进化规则算法优化BP神经网络初始权值来解决该问题的方法。进化规则算法对神经网络的随机初始权值进行编码,并通过变异算子与适当的选择策略找到最优的适应度个体,减小了神经网络学习过程中产生的经验误差,从而改进其训练效率,加快收敛速率。最后,基于预测出的路口车流量设计了一个信号灯控制系统。该系统对传感器数据进行处理分析,预测路口下一时间段车流量并以此分配路口信号灯时长,管理员能够实时了解信号灯、传感器和车流量信息并对信号灯模式和传感器进行控制,使用VISSIM软件进行路口三维建模以直观的方式显示路口车流状态。
[Abstract]:Multi sensor information fusion technology is the source of various sensors to detect the uncertain information of multi-level, comprehensive analysis, can earn more than a single sensor is more comprehensive, accurate and complete information processing. Information fusion is a kind of intelligent comprehensive information analysis technology, each sensor detection value of information complement each other, and the use of efficient calculation of the computer, from the time series or sequence of space information comprehensive analysis, detect the object more accurate data, more accurate recognition and decision-making functions, strengthen the information credibility and practicality. The intelligent traffic control using intelligent information processing method, traffic flow numerical research on sensor detected, calculate the length of the signal lamp strategy, the intersection traffic flow control. How to use financial information With technology to obtain real-time and accurate traffic flow information is the key goal of this paper. This paper analyses the intelligent traffic control information fusion technology is used according to the actual demand for real-time traffic, distribution, integration of environmental complex, established in the data processing of the sensor data using the least squares algorithm, information model in the BP network layer using decision fusion rules to improve the evolution of traffic flow forecast. In order to overcome the shortcoming of BP neural network for initial weights were randomly selected and easy to fall into local minimum solution of the problem, this paper studies the method to solve the problem by using evolutionary rules algorithm optimization the initial weights of BP neural network. The evolution rules of neural network algorithm the random initial weights for encoding, and find the optimal individual fitness by mutation operator and the appropriate selection strategy to reduce neural network Experience of errors in the learning process, to improve the training efficiency, speed up the convergence rate. Finally, the prediction of the traffic flow to design a signal lamp control system. The system based on analyzing the sensor data, traffic flow time and the distribution of the next prediction intersection intersection signal length, the administrator can understand the real-time signal lamp, sensor and traffic information and control the signal lamp mode and sensor, using VISSIM software for 3D modeling shows that intersection intersection traffic state in an intuitive way.

【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495;TP202;TP18

【相似文献】