基于交通网络数据集的动态路径诱导系统规划与实现探讨

智能交通是当前公认的有效解决交通拥堵、能源消耗及交通污染等突出问题的根本途径,而动态的交通诱导系统则是智能交通的核心研究领域之一,并一直被学术界和企业界所关注。当前已有诸多成果,但能广泛应用于实际交通环境的不多,通过着重探讨基于交通网络数据集下的动态路径诱导系统的规划设计与实现方式,有利于交通动态路径诱导系统的开发应用。

基于深度学习的短时交通流预测方法综述与仿真研究

近年来,随着城市路网交通检测设备和城市数据存储基础设施的升级换代以及深度学习技术的快速发展,应用深度学习技术解决城市路网短时交通流预测问题已成为智能交通领域的一个研究热点。不同于传统短时交通流预测方法,基于深度学习的短时交通流预测方法能充分利用海量交通流数据,深入挖掘路网中不同交通节点间流量的隐藏特征与复杂时空关联,能有效提升预测短时交通流的精度。首先,简要回顾短时交通流预测方法的发展历史,重点分析、讨论基于深度学习模型的短时交通流预测方法最新技术进展和理论研究结果。其次,梳理、总结国内外广泛用于验证算法有效性和进行比较分析的公开交通流数据集。再次,阐述基于深度学习模型的短时交通流预测算法解决实际交通流预测问题的具体过程和详细步骤,基于公开测试数据集PEMS04分别对基于深度学习模型长短时记忆网络(LSTM)和门控循环单元(GRU)的短时交通流预测算法进行仿真研究,以验证算法的有效性及其相较于传统方法的优势。最后,总结、展望基于深度学习模型的短时交通流预测方法在实际应用中存在的挑战和未来研究方向。

基于交通背景知识攻击模型的卡口车牌识别数据链接攻击研究

本文基于卡口过车数据集与停车场流水数据集,探究两者之间的数据关联性,并建立基于交通背景知识的攻击模型,对数据集中的个体进行链接攻击进行评估分析,以试图发现交通数据集中的链接攻击的现象和规律。但是本文的案例分析中,简化了较多因素,考虑不够全面,分析的准确性还有待提高,建议相关部门在实际中更深入分析数据之间的关联性,有效减少个人隐私数据的泄露。

基于城市居住区POI数据的昆明市公共交通站点可达性分析

本文利用2016年昆明市主城五区的POI数据,采用GIS分析工具,构建昆明市可达性分析模型,基于昆明路网建立步行交通网络数据集,通过最小阻抗和出行范围的两种可达性评价展开研究。通过针对服务于居民的居住区片区公共服务设施的空间布置方法,尝试通过考虑交通的转弯和通达性、时间成本以及距离等因素进行可达性分析,以昆明市公交站点和住区点为基础对昆明市公共交通可达性进行分析评价。

交通标志的检测与识别方法研究综述

交通标志的检测与识别是当今目标检测领域一项较为关键的技术,在智能交通系统和辅助驾驶技术中发挥着重要作用。对交通标志检测与识别领域早期的传统方法和现阶段较流行的深度学习方法分别做了阐述和分析,相较于传统的交通检测和交通识别方法,基于深度学习的方法更能兼顾准确性和实时性,对于遮挡、光线变化、目标太小等情况有很好的鲁棒性。最后总结了国内外常用的交通标志数据集,并对深度学习方法在该领域的发展进行了展望。

基于双向嵌套级联残差的交通标志检测方法

交通标志检测是自动驾驶领域的一个重要课题,其对于检测系统的实时性和精度都有非常高的要求。目标检测领域中的YOLOv3算法是业界公认在精度和速度上都处于前列的一种算法。文中以YOLOv3检测算法作为基础网络,提出一种双向嵌套级联残差单元(bid⁃NCR),替换掉原网络中顺序堆叠的标准残差块。双向嵌套级联残差单元的两条残差边采用相同的结构,都是一次卷积操作加上一次级联残差处理,两条边上级联的标准残差块的数量可以调节,从而形成不同的深度差。然后将两条边的结果逐像素相加,最后再做一次卷积操作。相较于标准残差块,双向嵌套级联残差单元拥有更强的特征提取能力和特征融合能力。文中还提出跨区域压缩模块(CRC),它是对2倍率下采样卷积操作的替代,旨在融合跨区域的通道数据,进一步加强主干网络输入特征图所包含的信息。实验结果表明:提出的模型在CCTSDB数据集上mAP(0.5)、mAP(0.5∶0.95)分别达到96.86%、68.66%,FPS达到66.09帧。相比于YOLOv3算法,3个指标分别提升1.23%、10.35%、127.90%。

复杂环境下的交通标志检测与识别方法综述

交通标志检测与识别是无人驾驶三大模块中环境感知的研究热点之一,检测和识别交通标志可以向无人车传递道路交通信息,优化行车决策。在暴雨、大雾以及光线昏暗等复杂环境下,拍摄到的图像往往会被遮挡,变得模糊。这不仅影响图像的质量,还会对后期标志的检测与识别带来巨大的困难。简述了交通标志检测与识别方法,对近年来国内外学者解决各类复杂环境下交通标志检测与识别的方法、原理和步骤进行了总结归纳,有利于人们更好地解决此类问题。同时,对常用的交通标志数据集进行了总结,并对数据集里在复杂环境下拍摄的图像比例给予了说明。

基于三尺度嵌套残差结构的交通标志快速检测算法

智能驾驶对交通标志自动检测的实时性及鲁棒性有着极高要求.目标检测中YOLOv3-tiny检测算法是轻量网络,实时性好、但准确率不高.将YOLOv3-tiny检测算法作为基础网络,提出了一种三尺度嵌套残差结构的交通标志快速检测算法.首先,在基础网络上采用逐像素相加的跨层连接,并未增加特征图的通道数,同时网络中形成1个小残差结构.其次,通过同样的跨层连接方式,增加了1层空间分辨率更高的预测输出,使得该尺度输出包含更丰富的空间信息,进而构成大残差结构.最终,将2个残差结构进行嵌套,形成了1个三尺度预测的嵌套残差网络模型,使得Tiny检测算法的部分主网络位于这2个残差结构中,起到3次调参的作用.实验结果表明:提出的算法能够快速鲁棒地检测真实场景中的交通标志.在德国交通标志检测数据集(German traffic sign detection benchmark, GTSDB)上交通标志总F1值为91.77%、检测时间为5 ms;在长沙理工大学中国交通标志检测数据集(CSUST Chinese traffic sign detection benchmark, CCTSDB)上指示、禁令、警告三大类交通标志F1值分别为92.41%,93.91%,92.03%,检测时间为5 ms.

基于Trafficnet的交通标志识别分析

交通标志识别是无人驾驶技术中的重要组成部分。随着深度学习近些年在全世界范围的快速发展以及无人驾驶技术中对识别效率越来越高的要求。文章提出一种基于改进googlenet网络,以德国交通标志数据集(GTSRB)为测试数据、caffe框架为工具,通过对输入图像进行卷积和下采样操作,提取出交通标志特征,该方法最大的优势就是提升了计算资源的利用率。

Data Reconstruction in Internet Traffic Matrix

Traffic matrix is an abstract representation of the traffic volume flowing between sets of source and destination pairs.It is a key input parameter of network operations management,planning,provisioning and traffic engineering.Traffic matrix is also important in the context of OpenFlow-based networks.Because even good measurement systems can suffer from errors and data collection systems can fail,missing values are common.Existing matrix completion methods do not consider traffic exhibit characteristics and only provide a finite precision.To address this problem,this paper proposes a novel approach based on compressive sensing and traffic self-similarity to reconstruct the missing traffic flow data.Firstly,we analyze the realworld traffic matrix,which all exhibit lowrank structure,temporal smoothness feature and spatial self-similarity.Then,we propose Self-Similarity and Temporal Compressive Sensing(SSTCS) algorithm to reconstruct the missing traffic data.The extensive experiments with the real-world traffic matrix show that our proposed SSTCS can significantly reduce data reconstruction errors and achieve satisfactory accuracy comparing with the existing solutions.Typically SSTCS can successfully reconstruct the traffic matrix with less than 32%errors when as much as98%of the data is missing.

Large-Scale Traffic Characterization of Chinese Multimedia Messaging Service Using Hadoop

Recently, there have been many mo- bile value-added services in the Chinese mo- bile telecommunication market nowadays. Am- ong them, the characteristics of Multimedia Mes- saging Service （MMS） have not yet been fully understood. In this paper, with the help of a cloud computing platform, we investigated the flow-level charactefistcs of Chinese MMS. All of the experimental data were collected by the TMS equipment deployed in a major node in Sou- them China. The collection time spanned six mo- nths. We performed high-level analysis to show the basic distributions of MMS characteristics. Then, by analysing the detailed MMS features, we determined the distribution of personal MMS, and made a comprehensive comparison between 2G and 3G MMS. Finally, we tried to build a model on the personal MMS inter-arrival time, and we found that the Weibull distribution was optimum.

Management Procedures for Data Collected via Intelligent Transportation Systems

Addressing transportation planning, operation and investment challenges requires increasingly sophisticated data and information management strategies. ITS （intelligent transportation systems） and CV （connected vehicle） technologies represent a new approach to capturing and using needed transportation data in real time or near real time. In the case of Michigan, several ITS programs have been launched successfully, but independently of each other. The objective of this research is to evaluate and assess all important factors that will influence the collection, management and use of ITS data, and recommend strategies to develop integrated, dynamic and adaptive data management systems for state transportation agencies.

Dynamic evaluation of traffic flows on city roads

This paper presents an analysis of the random fluctuations, deferred conduction effect and periodicity of road traffic based on the basic features of road networks. It also discusses the limitations of road network evaluation theories based on road "V/C". In addition, it proposes a set of theoretical and technical methods for the real-time evaluation of traffic flows for entire road networks, and for solving key technical issues, such as real-time data collection and processing in areas with no blind zones, the spatial-temporal dynamic analysis of road network traffic, and the calibration of key performance index thresholds. It also provides new technical tools for the strategic transportation planning and real-time diagnosis of road traffic. The new tools and methodology presented in this paper are validated using a case study in Beijing.

Impacts of Coordinated Traffic Signal Control Strategies and Bus Priority

Delay in signalized intersections may constitute a significant part of bus journey times in urban environment. Providing priority for buses at traffic signals can be an effective measure to reduce this delay. Bus priority in Swedish urban traffic signal systems are normally coordinated with fixed time plan selection. Within this framework local traffic actuated signal timing adjustments are applied based on detector inputs aimed to reduce the number of vehicles in the dilemma zone. Active bus priority is also achieved with the aim to display green signal at the arrival of the bus to the stop line. Due to lack of knowledge of traffic performance impacts of these techniques a major research study was undertaken funded by the Swedish Road Administration. The aim was to evaluate the following control strategies using Stockholm as case study： （1） Fixed time coordination （FTC）; （2） Fixed time coordination with local signal timing adjustment （FTC-LTA）; （3） FTC-LTA with active bus priority （PRIBUSS）; （4） Self-optimizing control （SPOT） with active bus priority. The methodologies for the study included field data collection using mobile and stationary techniques, offiine signal timing calculations with TRANSYT, microscopic simulation modeling using the HUTSIM model. The study obtained the following results： （1） Local traffic adjustment with the manual FTC reduced total delay by 1%. （2） Signal timings determined using TRANSYT reduced the average intersection delay by 9% compared to manual signal settings. （3） Local traffic adjustment reduced total delay by a further 5%. （4） Bus travel time was reduced by 11% using PRIBUSS, and 28% using SPOT. （5） Travel time for all vehicles did not increase using PRIBUSS, and was reduced by 6.5% with SPOT. Results of comparing PRIBUSS and SPOT to FTC-LTA were shown to be statistically significant.

基于机器学习的交通流预测方法综述

通过文献梳理、专家访谈和试验场景构建等方法,分析了道路指定断面和区域路网宏观交通流预测的国内外研究现状和发展趋势,归纳了局部断面交通流预测方法,包括传统机器学习、递归神经网络和混合模型,分析了卷积神经网络、图神经网络和融合多因素网络的特点,阐述了方法的原理、优势、局限性和应用场景,总结了现有场景交通数据集类别,从采样周期与采集方式角度归纳了国内外主流交通数据集。分析结果表明:递归神经网络可以有效获取交通数据的历史规律,但存在梯度爆炸、计算复杂度高、长时预测准确度不佳等问题;图神经网络针对路网拓扑连接关系引入了图结构,在考虑路网和交通流数据的时空相关性上具有明显优势;融合多因素网络充分考虑天气、道路、事故等内外部因素的影响,有效提升了交通流预测的实时性和鲁棒性;由于交通数据采集困难、外部因素影响难以量化、机器学习方法可解释性差等原因,交通流预测方法的改进受到了限制;未来应从交通信息有效挖掘和图卷积方法完善两方面入手,拓宽图结构在交通领域的应用和考虑非常态交通场景,进一步揭示交通数据的内在规律,开发更准确、高效的交通流预测方法,推动交通流预测在工业界的落地应用。

Exploring adoption of augmented reality smart glasses： Applications in the medical industry

This study explores the use of augmented reality smart glasses(ARSGs) by physicians and their adoption of these products in the Turkish medical industry.Google Glass was used as a demonstrative example for the introduction of ARSGs. We proposed an exploratory model based on the technology acceptance model by Davis. Exogenous factors in the model were defined by performing semi-structured in-depth interviews, along with the use of an expert panel in addition to the technology adoption literature. The framework was tested by means of a field study, data was collected via an Internet survey, and path analysis was used. The results indicate that there were a number of factors to be considered in order to understand ARSG adoption by physicians.Usefulness was influenced by ease of use, compatibility,ease of reminding, and speech recognition, while ease of use was affected by ease of learning, ease of medical education, external influence, and privacy. Privacy was the only negative factor that reduced the perceived ease of use,and was found to indirectly create a negative attitude.Compatibility emerged as the most significant external factor for usefulness. Developers of ARSGs should pay attention to healthcare-specific requirements for improved utilization and more extensive adoption of ARSGs in healthcare settings. In particular, they should focus on how to increase the compatibility of ARSGs. Further research needs to be conducted to explain the adoption intention of physicians.

Survival analysis of car travel time near a bus stop in developing countries

This paper presents an important investigation into car travel time affected by mixed traffic flow near a bus stop on the basis of survival analysis theory.Travel time data associated with mixed traffic characteristics near a bus stop were collected by video cameras.A hazard-based duration model was introduced to analyze the effects of mixed traffic flow on car travel time.The results indicate that mixed traffic flow impacts car travel time significantly.And the presence of bus berthing violation would delay car travel time.The proposed model can be used to forecast temporal shifts in car travel time due to changes in mixed traffic flow.The influential factors related to mixed traffic flow should be given full consideration in the planning and designing of bus stops in developing countries.