基于路网结构的城市交通事故短期风险预测方法

城市交通事故一般都发生在公共道路上,然而现有的交通事故风险预测算法都通过对预测区域进行规则网格化来确定预测空间单位,导致预测精度不高且实用价值较低。本文将道路路段作为预测单位,采用图卷积和长短期记忆网络,构建了一种基于路网结构的城市交通事故短期风险预测方法(traffic accidents risk prediction based on road network,TARPBRN)。该方法能对指定路段短期内的交通事故风险进行预测,从而可以有针对性地进行治理,减少交通事故的发生。本文使用杭州市西湖区的交通事故数据对模型进行了训练,并与4种常用的计量经济学模型和3种已有的深度学习预测算法进行了对比。实验结果证明本文算法在准确度、正确率和漏报率等方面都优于已有算法。     

Model-based object tracking in monocular image sequences of road traffic scenes

Moving vehicles are detected and tracked automatically in monocular image sequences from road traffic scenes recorded by a stationary camera. In order to exploit the a priori knowledge about shape and motion of vehicles in traffic scenes, a parameterized vehicle model is used for an intraframe matching process and a recursive estimator based on a motion model is used for motion estimation. An interpretation cycle supports the intraframe matching process with a state MAP-update step. Initial model hypotheses are generated using an image segmentation component which clusters coherently moving image features into candidate representations of images of a moving vehicle. The inclusion of an illumination model allows taking shadow edges of the vehicle into account during the matching process. Only such an elaborate combination of various techniques has enabled us to track vehicles under complex illumination conditions and over long (over 400 frames) monocular image sequences. Results on various real-world road traffic scenes are presented and open problems as well as future work are outlined.     

主成分分析法在城市道路交通安全评价中的应用

城市道路交通安全评价是一个综合性的课题,是改善道路交通安全、实现城市交通安全快捷不可或缺的部分。利用主成分分析的方法,选定几个评价其安全的主要指标,利用数学软件计算得出不同的几条路面的综合安全系数,得分越大表明危险系数越高;然后将所得结果与已有的参考文献中利用模糊综合评价法得出的结果比较,得出主成分法是可行的,而且能利用到实际的测评中来衡量某个城市区域交通规划的合理性,并在一些特定场合有其特有的优越性。     

Aerial image sequence geolocalization with road traffic as invariant feature

The geolocalization of aerial images is important for extracting geospatial information (e.g. the position of buildings, streets, and cars) and for creating maps. The standard is to use an expensive aerial imaging system equipped with an accurate GPS and IMU and/or do laborious ground control point measurements. In this paper we present a novel method to recognize the geolocation of aerial images automatically without any GPS or IMU. We extract road segments in the image sequence by detecting and tracking cars. We search in a database created from a road network map for the best matches between the road database and the extracted road segments. Geometric hashing is used to retrieve a shortlist of matches. The matches in the shortlist are ranked by a verification process. The highest scoring match gives the location and orientation of the images. We show in the experiments that our method can correctly geolocalize the aerial images in various scenes: e.g. urban, suburban, and rural with motorway. Besides the current images only the road map is needed over the search area. We can search an area of 22,500 km² containing 32,000 km of streets within minutes on a single cpu.     

Adaptive routing for road traffic

The article reports on an integrated system which uses road congestion information to guide routing, both in advance and while in transit. It offers two novel features: historic information about congestion is collected and retained for use when planning routes; and GPS tracks vehicles while they undertake journeys, and the Global System for Mobile Communications (GSM) Short Message Service (SMS) maintains communications between a moving vehicle and a central planning service to suggest revised routes avoiding congestion     

基于延迟函数次梯度启发式道路交通补偿策略

拥挤收费被认为是解决交通拥挤的有效方法,解决道路交通拥堵的主要想法是,对于有些容易造成拥堵的道路进行收费,而对于其他未充分利用的道路进行适当补偿,对此提出一种基于延迟函数的次梯度启发式道路交通补偿策略。首先,给出道路集的收费/补贴的非线性规划模型,主要是基于Beckmann最小化目标函数实现,然后利用库恩-希尔斯条件和拉格朗日乘子建立模型的条件约束;其次,基于启发式算法建立道路交通的定价补偿策略,利用边际成本建立延迟函数分析模型,然后基于次梯度法进行模型的优化;最后,通过在真实道路网络上的仿真实验,显示所提算法在旅行时间、交通流量、收敛性等指标上均具有较好的性能,验证了算法的有效性。     

A model identification scheme for driver-following dynamics in road traffic

The driver-following, or car-following, model is one of the most fundamental driver behavior models that are applied in intelligent transport applications. Its fidelity determines the applicability of microscopic traffic simulators, where the model is often implemented to mimic real traffic. Meanwhile, the behavioral model is fundamental to the development of advanced driving assistance systems (ADAS). This paper develops a dynamic model identification approach based on iterative usage of the extended Kalman Filtering (EKF) algorithm. Among other things, this allows to carry out model identification using a rather general optimization objective on the whole physical states of the following vehicle. In particular, the method is established on the basis of the equivalence between the Kalman filter and the recursive least squares (RLS) method in a specific context of parameter identification. To illustrate the method, two car-following models are studied in numerical experiments using real car-following data. The method has shown advantages in replication and prediction of vehicle dynamics in car-following over the conventional approaches. It has also the potential to be further extended for building tactical driving controllers in intelligent transportation applications.     

基于非参数核密度模型的交通图像目标提取

针对现有目标提取和去噪方法不能很好地满足城市交通图像车辆目标提取的要求,提出基于概率比较结合形态学闭操作的目标提取去噪方法。通过非参核密度估计算法建立背景模型,获得每个像素点上各灰度值的出现概率,提取出前景目标;分别计算前景目标是属于车辆移动还是树叶抖动的概率,通过概率比较去除噪声,用形态学闭操作进一步去噪。实验结果表明,提出的算法较好地实现了树叶噪声与车辆目标的分离,能有效去除树叶抖动噪声,正确提取车辆目标,具有良好的抗噪性。     

Representation of occurrences for road vehicle traffic

Our 3D-model-based Computer Vision subsystem extracts vehicle trajectories from monocular digitized videos recording road vehicles in inner-city traffic. Steps are documented which import these quantitative geometrical results into a conceptual representation based on a Fuzzy Metric-Temporal Horn Logic (FMTHL, see [K.H. Schäfer, Unscharfe zeitlogische Modellierung von Situationen und Handlungen in Bildfolgenauswertung und Robotik, Dissertation, 1996]). The facts created by this import step can be understood as verb phrases which describe elementary actions of vehicles in image sequences of road traffic scenes. The current contribution suggests a complete conceptual representation of elementary vehicle actions and reports results obtained by an implementation of this approach from real-world traffic videos.     

Detailed traffic animation for urban road networks

We present a new agent-based system for detailed traffic animation on urban arterial networks with diverse junctions like signalized crossing, merging and weaving areas. To control the motion of traffic for visualization and animation purposes, we utilize the popular follow-the-leader method to simulate various vehicle types and intelligent driving styles. We also introduce a continuous lane-changing model to imitate the vehicle’s decision-making process and dynamic interactions with neighboring vehicles. By applying our approach in several typical urban traffic scenarios, we demonstrate that our system can well visualize vehicles’ behaviors in a realistic manner on complex road networks and generate immersive traffic flow animations with smooth accelerating strategies and flexible lane changes.     

一种基于图像特征的移动流量分类方法

随着移动设备的快速发展和应用激增,其产生的移动流量也迅猛增加且众多操作系统皆存在着巨大的安全风险,能够从巨大的网络流量中有效地区分出来自移动端的流量并识别其操作系统对后续的移动安全的分析有着重大的安全意义。基于传统特征的流量分析技术存在着因过分依赖特征选择而导致无法稳定有效地分类移动流量的问题,提出了一种基于图像特征的移动流量分类方法。该方法将流量样本进行可视化转换成灰度图像,从而提取其图像的GLCM特征进行分类。实验结果表明,该方法较传统方法精确率最高提升22.4%,有效地解决了传统方法的特征选择以及没有良好的扩展性等问题。此外,研究了流量研究粒度（flow到stream）、分类粒度（二分类到多分类）和数据集的均衡性（均衡与不均衡）对移动流量检测方法的影响,结果表明分类粒度对分类准确率的影响影响极小,准确率最大降低2.6%。该实验结果进一步说明了提出方法的扩展性,能够有效地用于后续的移动流量的安全研究。     

Pattern recognition for road traffic accident severity in Korea

An increasing number of road traffic accidents (RTA) in Korea has emerged as being harmful both for the economy and for safety. An accurately estimated classification model for several severity types of RTA as a function of related factors provides crucial information for the prevention of potential accidents. Here, three data-mining techniques (neural network, logistic regression, decision tree) are used to select a set of influential factors and to build up classification models for accident severity. The three approaches are then compared in terms of classification accuracy. The finding is that accuracy does not differ significantly for each model and that the protective device is the most important factor in the accident severity variation.     

Pattern recognition for road traffic accident severity in Korea

An increasing number of road traffic accidents (RTA) in Korea has emerged as being harmful both for the economy and for safety. An accurately estimated classification model for several severity types of RTA as a function of related factors provides crucial information for the prevention of potential accidents. Here, three data-mining techniques (neural network, logistic regression, decision tree) are used to select a set of influential factors and to build up classification models for accident severity. The three approaches are then compared in terms of classification accuracy. The finding is that accuracy does not differ significantly for each model and that the protective device is the most important factor in the accident severity variation.     

雾天交通场景中单幅图像去雾

目的 针对雾天交通场景中通常含有大片天空区域,传统去雾方法在复原这些图像时容易产生光晕、色彩失真等现象。提出一种基于天空区域分割和暗通道先验理论的图像去雾方法。方法 首先提出一种基于最大类间方差法（OTSU）的图像分割算法来自动分离天空区域,然后将天空区域的平均强度值作为大气光值,从而改进场景传输率的估计。结果 本文方法不仅有效地解决了图像去雾后出现光晕和色彩失真等问题,与He Kaiming以及Fattal的方法相比,本文方法去雾后的图像更加真实、自然、平滑。结论 实验结果表明,本文方法能够很好地完成对雾天交通图像的复原,可为道路交通监管提供一种有效的理论基础和技术支撑。     

交通道路行驶车辆车标识别算法

为解决交通道路行驶车辆车标识别中存在的目标小、噪声大、种类多的问题,提出了一种基于深度学习的目标检测算法与基于形态学模板匹配算法相结合的方法,并设计了一种高准确度且能应对新类型车标的识别系统.首先,采用通过K-Means++重新聚类锚框值,并引入残差网络的YOLOv4进行车标的一步定位;其次,通过对标准车标图像进行预处...     

基于视频信息的城市路段交通安全状态评估方法研究

城市道路交通安全状态实时评估是智能交系统的重要研究内容。针对现有交通安全状态实时评估方法的评估效果不理想的问题,建立了基于视频信息的城市路段交通安全状态评估方法。首先,分析了基于视频信息的交通流参数快速检测方法;然后,从车速离散度的角度提出了影响城市路段交通安全状态的路段车速离散度的概念;最后,建立了基于路段车速离散度的城市路段交通安全状态评估方法。实验表明,所提方法能够实时合理地对城市道路的安全状态水平进行评估,评估结果可以为交通管理部门制定有效的城市路段交通安全改进方案提供相应的依据。     

A new pyramid-based color image representation for visual localization

This paper presents a new algorithm to extract a color graph from a real image so that it can be efficiently represented in a compact way. The proposal presents two main novelties. First, a new PCA-based color feature vector has been defined. This vector is resistant against illumination changes, noise and microtextures. Second, the image is segmented by means of a fast, hierarchical split and merge algorithm based on a pyramid structure that implicitly provides a color graph. In order to test the suitability of the proposed representation technique, simple localization experiments have been conducted in different real environments.     

A robust method for image stitching

Pattern Analysis and Applications - We propose a novel method for large-scale image stitching that is robust against repetitive patterns and featureless regions in the imagery. In such cases,...     

一种改进的深度学习的道路交通标识识别算法

针对复杂的环境,结合图像预处理与深度学习神经网络,提出了一种道路交通标识识别算法。该方法不仅利用图像分割技术,而且利用卷积神经网络模型对道路交通标识进行了更准确的识别。首先,通过调节光照影响、去除复杂背景、数据增强和归一化等批量预处理操作,形成一个完整的数据集;然后,结合squeeze-and-excitation思想和残差网络结构,充分训练出自己的卷积神经网络模型;最后,将优化的网络模型用于道路交通标识的识别。实验结果表明,该方法使训练时间缩短了12%左右,识别精度可达99.26%。