基于机器视觉的车辆检测与参数识别研究进展

近年来公路交通运输快速增长,交通车辆的快速准确检测与识别对智能交通系统和交通基础设施运维具有重要意义.随着机器视觉和深度学习技术的迅速发展及其在目标检测领域的广泛应用,车辆目标检测和参数识别也取得新的突破.该文从车辆参数的识别方法和应用研究两方面梳理了机器视觉和深度学习在车辆检测与参数识别领域的研究现状、最新研究成果和...     

FMS故障诊断的神经网络方法研究

文中对神经网络理论应用于FMS(Flexible Manufacture System,柔性制造系统)故障诊断进行研究,在介绍了神经网络基本理论的基础上,提出一种FMS故障诊断的神经网络计算方法。     

基于BP神经网络的预测模型

预测是数据分析的基本任务之一,传统方法对非线性数据的预测不易找到简单而有效的模型,神经网络的产生为处理非线性问题提供了一条新途径。文中运用智能计算技术建立了BP网络模型。通过珠江三角洲天河水文站的水位预测结果表明,BP网络模型有较好的泛化能力,预测更为可靠。     

基于MapBasic的战时运输最优路径问题研究

战时运输最优路径问题是一个多目标多约束随机动态路网寻优问题。在分析战时运输最优路径问题特性前提下,着重研究战时运输路阻函数模型,求出时间阻抗、风险阻抗和费用阻抗,标定阻抗参数μ1,μ2和μ3,及确定函数模型的MapBasic表达,在给出最优路径模型基础上,利用改进的Dijkstra算法求解。实例验证表明研究成果满足实用要求。     

The use of demand correlation in the modeling of air carrier departure delays as first‐order autoregressive random processes

A method for modeling air carrier departure delays at commercial‐service airports as autoregressive random processes is presented. This method employs the correlation of a priori demand data to significantly reduce prediction error in the optimal least‐squares estimator for additive white noise. The reduction factor of the prediction error is demonstrated to be on the order of 10² over that of the unbiased estimator. Copyright © 2011 John Wiley & Sons, Ltd.     

基于GRNN的GSM-R场强覆盖预测算法

在比较Hata模型修正方法和GRNN算法场强覆盖预测效果的基础上,仿真分析了训练集构成和平滑因子选择对GRNN算法预测效果的影响,给出了训练集构成和平滑因子选择的方法;提出了电波传播环境相似系数来表征GRNN模型在不同环境下的适用性。仿真结果表明,两种传播环境的相似系数越大,由一种环境下测试数据确定的GRNN在另一环境下的预测精度越高。     

基于社交网络的民航旅客偏好建模

从社会网络的视角，提出了一种旅客个体偏好与关系偏好相结合的建模方法．首先，从旅客的历史出行记录中，构建基于共同出行关系的旅客社会网络；然后，构建旅客个体偏好模型和旅客关系偏好模型；最后，基于旅客偏好模型给旅客推荐座位．在民航领域的一个真实的数据集上进行了实验，证明本文提出的偏好模型能够有机地将旅客个体偏好与关系偏好结合起来，较好地描述旅客对航班座位的偏好．     

Resolving the property right of transportation emissions through public–private partnerships

The application of public–private partnerships (P3’s) in the transportation sector has grown in popularity worldwide. Despite this important shift in the provision of transportation service, there are clear gaps in knowledge about the impacts of P3 projects, especially on emissions from transportation systems as a whole. Not only should policy makers evaluate the emissions impacts from P3 projects, but they should also think about innovative models that address or charge for emissions into P3 contracts. This addition to P3 contracts could provide a new solution to the long-existing property right paradox: who owns (is responsible for) emissions from transportation systems? This study attempts to fill the research gap by analyzing these innovative models. Using the road network of Fresno, California, as our case study, we offer a number of interesting insights for policy makers. First, average peak emissions costs range from 1.37 cents per mile (the do-nothing case) to 1.20 cents per mile (profit-maximizing cases) per vehicle. Although emissions costs from the P3 projects are lowest for the profit-maximizing cases, the system-wide emissions costs of these cases are highest because of spillover effects. Second, charging project owners for the emissions costs of P3 projects is not an effective way to reduce emissions or the total costs of travel, especially on a VMT basis. Instead, the public sector should implement emissions-included social cost-based price ceilings. When employing these limits, project owners could still be charged for the emissions costs. Finally, using total travel time as the only objective function for evaluating P3 projects can be misleading. Several P3 projects have shown better outcomes using total travel cost with the inclusion of emissions and fuel consumption costs, instead of using total travel time as the only objective function.     

短时段交通流预测的RBF神经网络模型

介绍了基于RBF神经网络的短时段交通量预测模型,并利用该模型对高速公路所采集的数据进行仿真预测分析。预测结果表明RBF神经网络预测方法通过定义合理的网络结构参数可以获得较高的预测精度,能够满足路网调度对短时段交通流预测的需求。     

基于神经网络的高速公路出入口OD矩阵估计方法研究

通过高速公路出入口OD矩阵计算,得到高速公路的断面交通流量,这些数据对于高速公路各项管理措施的制定是非常重要的.针对高速公路出入口OD矩阵推算方法中假设每一留线车辆以等概率驶离高速公路的不足,提出了基于改进BP神经网络的高速公路出入口OD矩阵推算模型,并设计了OD推算神经网络.实例分析表明,该模型推算结果理想,且推算精度得到一定提高.