WO2021073524A1 - Analysis method for tracing source of congestion traffic flow - Google Patents
Analysis method for tracing source of congestion traffic flow Download PDFInfo
- Publication number
- WO2021073524A1 WO2021073524A1 PCT/CN2020/120829 CN2020120829W WO2021073524A1 WO 2021073524 A1 WO2021073524 A1 WO 2021073524A1 CN 2020120829 W CN2020120829 W CN 2020120829W WO 2021073524 A1 WO2021073524 A1 WO 2021073524A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- source
- vehicle
- neural network
- deep neural
- spatial
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Definitions
- the invention relates to the field of traffic control, in particular to a method for tracing the source of congested traffic flow.
- Tracing the source of congested traffic flow refers to tracing the source of traffic flow at the time and space level.
- spatial traceability refers to tracing the starting position of a vehicle outside a certain spatial range
- time traceability refers to estimating the travel time required for the vehicle to reach a specific spatial position from the starting position. Since the penetration rate of connected vehicles will continue to maintain a low market penetration rate for a long period of time in the future, it is impossible to accurately determine the source of vehicles in congested areas.
- the traceability analysis of congested traffic flow traces the traffic by analyzing the existing incomplete data. The source of the flow is expected to become the key input information for network traffic control strategies.
- VPR Vehicle Path Reconstruction
- Automatic vehicle identifier data is a kind of data that is more suitable for traffic flow traceability.
- the trajectory data contains more information, in addition to the aforementioned constraints caused by the low permeability, the permeability itself is random, and its estimation is also a difficult point.
- cross-section sensors such as bayonet detection equipment, can detect information about all passing vehicles and have become popular in many large cities.
- Traffic flow traceability methods can provide new ideas for existing traffic jam mitigation strategies.
- Signal-based control For example, typical signal control systems: Sydney Coordinated Adaptive Traffic System (SCATs) and Sydney Coordinated Adaptive Traffic System (SCOOTs) ); 2) Optimization based on road facilities: For example, the utilization of time and space resources can be improved through the setting of variable lanes and dedicated bus lanes; 3) Based on travel mode; 4) Based on the ratio of turns at intersections.
- SCATs Sydney Coordinated Adaptive Traffic System
- SCOOTs Sydney Coordinated Adaptive Traffic System
- Optimization based on road facilities For example, the utilization of time and space resources can be improved through the setting of variable lanes and dedicated bus lanes; 3) Based on travel mode; 4) Based on the ratio of turns at intersections.
- congestion charging policies the development of electric car time-sharing and other measures.
- several of the above-mentioned congestion mitigation measures do not consider the source information of the traffic flow in the congested area, and thus do not have the ability to mitigate congestion
- the existing traffic flow traceability method does not consider the source information of the traffic flow in the congested area, so it does not have the ability to reduce the congestion from the network level.
- the purpose of the present invention is to provide a method for tracing the source of congested traffic flow in order to overcome the defects of the prior art.
- a method for tracing the source of congested traffic flow includes the following steps:
- Step S1 Based on the automatic vehicle identifier data and vehicle road source data of vehicles in the congested area, construct a deep neural network multi-classification model to obtain the spatial source of the vehicle;
- Step S2 Based on the spatial source of the vehicle and the data of the automatic vehicle identifier, a deep neural network regression model is constructed to obtain the time traceability result of the vehicle.
- the step S1 includes:
- Step S11 Perform one-hot encoding on the automatic vehicle identifier data and the vehicle road source data to obtain the automatic vehicle identifier one-hot encoding data and the vehicle road source one-hot encoding data respectively;
- Step S12 Construct a loss function of a deep neural network multi-classification model related to the spatial source
- Step S13 Based on the one-hot encoding data of the automatic vehicle recognizer, the one-hot encoding data of the source of the vehicle and the loss function of the deep neural network multi-classification model, the deep neural network multi-classification model is obtained through the optimization algorithm and the first accuracy algorithm;
- Step S14 Obtain the spatial source of the vehicle based on the multi-classification model of the deep neural network.
- the calculation formula of the loss function of the deep neural network multi-class model is:
- N is the number of vehicles
- m is the tag number of the spatial source
- p ⁇ m is the probability that the vehicle ⁇ belongs to the spatial source m
- y ⁇ m is the spatial source
- the first accuracy calculation method is:
- EE ⁇ represents the correctness of the spatial source area of the vehicle ⁇ .
- the spatial source area includes a boundary road segment and its adjacent boundary road segments on both sides, N is the number of vehicles, and SEA is the accuracy.
- the step S2 includes:
- Step S21 Perform one-hot encoding on the space source of the vehicle and the data of the automatic vehicle identifier to obtain the one-hot encoding space source and the one-hot encoding data of the automatic vehicle identifier;
- Step S22 Construct a loss function of the deep neural network regression model related to the time traceability result
- Step S23 Based on the one-hot encoding data of the automatic vehicle identifier, the one-hot encoding space source and the loss function of the deep neural network regression model, the deep neural network regression model is obtained through the optimization algorithm and the second accuracy algorithm;
- Step S24 Obtain the time traceability result of the vehicle based on the deep neural network regression model.
- the calculation formula of the second accuracy algorithm is the same as the calculation formula of the loss function of the deep neural network regression model.
- the present invention has the following advantages:
- a spatio-temporal analysis framework for traceability namely, deep neural network multi-classification model and deep neural network regression model, can avoid the problem of gradual increase in errors in the traceability method based on the ratio of intersection turns when the traceability distance increases.
- Figure 1 is a flow chart of the present invention
- FIG. 2 is a schematic road network diagram of traceability according to an embodiment of the present invention.
- FIG. 3 is a schematic diagram of the spatial error of an embodiment of the present invention.
- FIG. 4 is a schematic diagram of inputting a deep neural network multi-classification model according to an embodiment of the present invention.
- Fig. 5 is a comparison diagram of traceability results between an embodiment of the present invention and traditional machine learning.
- This embodiment provides a method for tracing the source of congested traffic flow, as shown in FIG. 1, which includes two steps:
- Step S1 Based on the automatic vehicle identifier data and vehicle road source data of vehicles in the congested area, construct a deep neural network multi-classification model to obtain the spatial source of the vehicle;
- Step S2 Based on the spatial source of the vehicle and the data of the automatic vehicle identifier, a deep neural network regression model is constructed to obtain the time traceability result of the vehicle.
- Step S1 includes:
- Step S11 Perform one-hot encoding on the automatic vehicle identifier data and the vehicle road source data to obtain the automatic vehicle identifier one-hot encoding data and the vehicle road source one-hot encoding data;
- Step S12 Construct a loss function of a deep neural network multi-classification model related to the spatial source
- Step S13 Based on the one-hot encoding data of the automatic vehicle recognizer, the one-hot encoding data of the source of the vehicle and the loss function of the deep neural network multi-classification model, the deep neural network multi-classification model is obtained through the optimization algorithm and the first accuracy algorithm;
- Step S14 Obtain the spatial source of the vehicle based on the multi-classification model of the deep neural network.
- the deep neural network multi-classification model is based on the deep learning classifier (DNN Classifier);
- the road sections are adjacent to each other and form a sub-network in the urban road network.
- the vehicle road source data is the data of the sub-network, and the research is conducted on the vehicles within the sub-network.
- ⁇ as the number of the vehicle
- the spatial source of the vehicle is the set of boundary road segments Any boundary section in the. definition Is the spatial error
- the value of represents the number of boundary road sections between the true spatial source of the vehicle and the spatial source of the vehicle inferred by the deep neural network multi-classification model, so It can only be a non-negative integer.
- the output label of the deep neural network multi-class model As It represents the spatial source of the vehicle, and is specifically expressed as a collection of boundary road sections A certain road section in.
- 1 represents the space source of the vehicle, and there is one and only one 1 in a vector, and the rest are all 0s.
- E.g: Indicates that the source of the vehicle is the third boundary section (m 3).
- N represents the number of vehicles
- m represents the tag number of the spatial source
- y ⁇ m represents the spatial source of the vehicle
- p ⁇ m represents The probability that the vehicle ⁇ belongs to the spatial source m.
- Step S13 The essence of the deep neural network algorithm is to find the negative gradient and iterate until the optimal solution is found. This process is called gradient descent. This method uses the most commonly used optimization algorithms AdaGrad and Adam in the Google open source machine learning library TensorFlow.
- SEA is accuracy.
- Step S2 includes:
- Step S21 Perform one-hot encoding on the space source of the vehicle and the data of the automatic vehicle identifier to obtain the one-hot encoding space source and the one-hot encoding data of the automatic vehicle identifier;
- Step S22 Construct a loss function of the deep neural network regression model related to the time traceability result
- Step S23 Based on the one-hot encoding data of the automatic vehicle identifier, the one-hot encoding space source and the loss function of the deep neural network regression model, the deep neural network regression model is obtained through the optimization algorithm and the second accuracy algorithm;
- Step S24 Obtain the time traceability result of the vehicle based on the deep neural network regression model.
- the deep neural network regression model is based on the deep learning regressor (DNN Regressor).
- the defined travel time represents the elapsed time for the vehicle ⁇ from the starting boundary road segment to reach the road segment to be traced. Since the initial road section does not necessarily have an automatic vehicle recognition detector, in the time traceability model, a regression method is used to estimate the travel time, and the travel time obtained by the deep neural network regression model (that is, the time traceability result) is defined as
- the input information of the deep neural network regression model in step S21 is still in the form of one-hot encoding, which defines To input information, it mainly consists of two parts: the first part Is the output result of the deep neural network multi-classification model, namely the second part
- the information contained is the detector number where the vehicle ⁇ was first detected in the sub-network, and the time difference between reaching the road segment to be traced. For example, suppose Is the timestamp when the vehicle ⁇ is first detected by the detector ⁇ in the sub-network, then among them, The number of elements contained is equal to the number of detectors in the sub-road network, for The ⁇ th element of represents that it was captured by the ⁇ detector, and the remaining elements are all 0.
- step S23 AdaGrad and Adam in Google's open source machine learning library TensorFlow are also used as model optimization algorithms.
- the second accuracy is defined as TEE, and its algorithm is the same as the loss function of the deep neural network regression model, namely:
- the sub-network is composed of 25 intersections and several road sections, and several automatic vehicle identifiers are distributed on the road sections.
- the gray circle represents the ordinary intersection
- the black circle represents the boundary intersection
- the set of boundary sections is The section to be traced is r 14-15 .
- the true spatial source of the vehicle is r 3-8
- the spatial error values of different inferences of the deep neural network multi-classification model are shown in Figure 3. As shown in one column. It can be seen that the spatial errors are all non-negative integers.
- Figure 4 shows the input automatic vehicle identifier data form of the trajectory I and trajectory II in the deep neural network multi-classification model. If the vehicle passes through the road section with the automatic vehicle identifier, then The value of the corresponding element is 1, otherwise it is 0.
- the classification and regression based on deep neural networks used in this method is compared with the classification and regression based on traditional machine learning.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Traffic Control Systems (AREA)
Abstract
An analysis method for tracing a source of a congestion traffic flow, comprising the following steps: step S1: constructing, on the basis of automatic vehicle identifier data and vehicle path source data of vehicles in a congestion area, a deep neural network multi-classification model, so as to obtain a spatial source of the vehicles (S1); and step S2: constructing, on the basis of the spatial source and the automatic vehicle identifier data of the vehicles, a deep neural network regression model, so as to obtain a time source tracing result of the vehicles (S2). Compared with the prior art, the present invention takes source information of the traffic flow in the congestion area into account, thus has the capability of relieving congestion from a network level, and provides a new research perspective for relieving congestion; and compared with traditional machine learning algorithms, the present invention can significantly improve reasoning accuracy.
Description
本发明涉及交通控制领域,尤其是涉及一种拥堵交通流溯源分析方法。The invention relates to the field of traffic control, in particular to a method for tracing the source of congested traffic flow.
拥堵交通流溯源是指在时间与空间层面上对交通流的来源进行追溯。其中,空间溯源是指追溯车辆在一定空间范围之外的起点位置,时间溯源是指估计该车辆从起点位置到达某一具体空间位置所需要的行程时间。由于网联车渗透率在未来较长一段时间内将持续保持较低的市场渗透率,对拥堵区域车辆来源是无法准确判断的,拥堵交通流溯源分析通过解析现有的不完整的数据追溯交通流的来源,有望成为网络交通控制策略的关键输入信息。Tracing the source of congested traffic flow refers to tracing the source of traffic flow at the time and space level. Among them, spatial traceability refers to tracing the starting position of a vehicle outside a certain spatial range, and time traceability refers to estimating the travel time required for the vehicle to reach a specific spatial position from the starting position. Since the penetration rate of connected vehicles will continue to maintain a low market penetration rate for a long period of time in the future, it is impossible to accurately determine the source of vehicles in congested areas. The traceability analysis of congested traffic flow traces the traffic by analyzing the existing incomplete data. The source of the flow is expected to become the key input information for network traffic control strategies.
从定义上来看,拥堵交通流溯源与车辆轨迹重构(Vehicle Path Reconstruction,VPR)既存在相似性也存在不同点:相同点在于,两者的目的均在于获取车辆来源更详细的信息;其差异在于,车辆轨迹重构的目的在于获取单量车的具体轨迹,而交通溯源则只需要获取车辆来源信息,而无需获取完整的路径信息。From a definition point of view, congested traffic flow tracing and Vehicle Path Reconstruction (VPR) both have similarities and differences: the same point is that the purpose of the two is to obtain more detailed information about the source of the vehicle; their differences It is that the purpose of vehicle trajectory reconstruction is to obtain the specific trajectory of a single vehicle, while traffic tracing only needs to obtain vehicle source information without obtaining complete path information.
得益于车联网技术的逐渐发展,蕴含丰富交通运行信息的浮动车轨迹数据的获取变得更加容易,给交通参数估计、交通管控策略的研究提供了丰富的想象空间,现有应用包括排队长度估计,信号配时优化等等。然而,大部分研究均依赖于较高的市场渗透率。Thanks to the gradual development of Internet of Vehicles technology, it has become easier to obtain floating car trajectory data containing rich traffic operation information, which provides a wealth of imagination for traffic parameter estimation and traffic control strategy research. Existing applications include queue length. Estimate, signal timing optimization and so on. However, most studies rely on high market penetration.
自动车辆识别器数据是一种更适合交通流溯源的数据。虽然轨迹数据包含更多的信息,但是除了前述的低渗透率造成的约束之外,渗透率本身存在随机性,其估计也是一个难点。相比而言,断面传感器,例如卡口检测设备,能够检测到所有经过的车辆信息,并已经在许多大城市内普及。Automatic vehicle identifier data is a kind of data that is more suitable for traffic flow traceability. Although the trajectory data contains more information, in addition to the aforementioned constraints caused by the low permeability, the permeability itself is random, and its estimation is also a difficult point. In contrast, cross-section sensors, such as bayonet detection equipment, can detect information about all passing vehicles and have become popular in many large cities.
交通流溯源方法可以给现有交通缓堵策略提供新的思路。目前,在交通拥堵缓解策略领域已有大量的研究与成果,主要可归纳为1)基于信号控制:例如,典型的信号控制系统:Sydney Coordinated Adaptive Traffic System(SCATs)和Sydney Coordinated Adaptive Traffic System(SCOOTs);2)基于道路设施优化:例如,通过可 变车道、公交专用道的设置来提高时空资源的利用率;3)基于出行模式;4)基于交叉口转弯比例。例如,实施拥堵收费政策、发展电动汽车分时租赁等措施。然而,上述的若干缓堵措施,均没有考虑拥堵区域交通流的来源信息,从而不具备从网络层面进行缓堵的能力。Traffic flow traceability methods can provide new ideas for existing traffic jam mitigation strategies. At present, there have been a lot of research and results in the field of traffic congestion mitigation strategies, which can be summarized as 1) Signal-based control: For example, typical signal control systems: Sydney Coordinated Adaptive Traffic System (SCATs) and Sydney Coordinated Adaptive Traffic System (SCOOTs) ); 2) Optimization based on road facilities: For example, the utilization of time and space resources can be improved through the setting of variable lanes and dedicated bus lanes; 3) Based on travel mode; 4) Based on the ratio of turns at intersections. For example, the implementation of congestion charging policies, the development of electric car time-sharing and other measures. However, several of the above-mentioned congestion mitigation measures do not consider the source information of the traffic flow in the congested area, and thus do not have the ability to mitigate congestion from the network level.
目前存在的问题:已有的交通流溯源方法没有考虑拥堵区域交通流的来源信息,从而不具备从网络层面进行缓堵的能力。The current problem: The existing traffic flow traceability method does not consider the source information of the traffic flow in the congested area, so it does not have the ability to reduce the congestion from the network level.
发明内容Summary of the invention
本发明的目的就是为了克服上述现有技术存在的缺陷而提供一种拥堵交通流溯源分析方法。The purpose of the present invention is to provide a method for tracing the source of congested traffic flow in order to overcome the defects of the prior art.
本发明的目的可以通过以下技术方案来实现:The purpose of the present invention can be achieved through the following technical solutions:
一种拥堵交通流溯源分析方法,该方法包括以下步骤:A method for tracing the source of congested traffic flow, the method includes the following steps:
步骤S1:基于拥堵区域的车辆的自动车辆识别器数据和车辆道路来源数据,构建深度神经网络多分类模型,得到车辆的空间来源;Step S1: Based on the automatic vehicle identifier data and vehicle road source data of vehicles in the congested area, construct a deep neural network multi-classification model to obtain the spatial source of the vehicle;
步骤S2:基于车辆的空间来源和自动车辆识别器数据,构建深度神经网络回归模型,得到车辆的时间溯源结果。Step S2: Based on the spatial source of the vehicle and the data of the automatic vehicle identifier, a deep neural network regression model is constructed to obtain the time traceability result of the vehicle.
所述的步骤S1包括:The step S1 includes:
步骤S11:将自动车辆识别器数据和车辆道路来源数据进行独热编码,分别得到自动车辆识别器独热编码数据和车辆道路来源独热编码数据;Step S11: Perform one-hot encoding on the automatic vehicle identifier data and the vehicle road source data to obtain the automatic vehicle identifier one-hot encoding data and the vehicle road source one-hot encoding data respectively;
步骤S12:构建与空间来源有关的深度神经网络多分类模型损失函数;Step S12: Construct a loss function of a deep neural network multi-classification model related to the spatial source;
步骤S13:基于自动车辆识别器独热编码数据、车辆道路来源独热编码数据和深度神经网络多分类模型损失函数,通过优化算法和第一准确度算法得到深度神经网络多分类模型;Step S13: Based on the one-hot encoding data of the automatic vehicle recognizer, the one-hot encoding data of the source of the vehicle and the loss function of the deep neural network multi-classification model, the deep neural network multi-classification model is obtained through the optimization algorithm and the first accuracy algorithm;
步骤S14:基于深度神经网络多分类模型,得到车辆的空间来源。Step S14: Obtain the spatial source of the vehicle based on the multi-classification model of the deep neural network.
所述的深度神经网络多分类模型损失函数的计算式为:The calculation formula of the loss function of the deep neural network multi-class model is:
其中,N为车辆的数量,m为空间来源的标签编号,p
ωm为车辆ω属于空间来源m的概率;y
ωm为空间来源,y
ωm=1表示空间来源m为车辆ω的正确空间来源,y
ωm=0表示空间来源m不是车辆ω的正确空间来源。
Among them, N is the number of vehicles, m is the tag number of the spatial source, p ωm is the probability that the vehicle ω belongs to the spatial source m; y ωm is the spatial source, and y ωm =1 indicates that the spatial source m is the correct spatial source of the vehicle ω, y ωm = 0 means that the space source m is not the correct space source of the vehicle ω.
第一准确度计算方法为:The first accuracy calculation method is:
其中,EE
ω表示车辆ω的空间来源区域的正确性,所述的空间来源区域包括一条边界路段及其两侧相邻的边界路段,N为车辆的数量,SEA为准确度。
Among them, EE ω represents the correctness of the spatial source area of the vehicle ω. The spatial source area includes a boundary road segment and its adjacent boundary road segments on both sides, N is the number of vehicles, and SEA is the accuracy.
所述的步骤S2包括:The step S2 includes:
步骤S21:将车辆的空间来源和自动车辆识别器数据进行独热编码,得到独热编码空间来源和自动车辆识别器独热编码数据;Step S21: Perform one-hot encoding on the space source of the vehicle and the data of the automatic vehicle identifier to obtain the one-hot encoding space source and the one-hot encoding data of the automatic vehicle identifier;
步骤S22:构建与时间溯源结果有关的深度神经网络回归模型损失函数;Step S22: Construct a loss function of the deep neural network regression model related to the time traceability result;
步骤S23:基于自动车辆识别器独热编码数据、独热编码空间来源和深度神经网络回归模型损失函数,通过优化算法和第二准确度算法得到深度神经网络回归模型;Step S23: Based on the one-hot encoding data of the automatic vehicle identifier, the one-hot encoding space source and the loss function of the deep neural network regression model, the deep neural network regression model is obtained through the optimization algorithm and the second accuracy algorithm;
步骤S24:基于深度神经网络回归模型,得到车辆的时间溯源结果。Step S24: Obtain the time traceability result of the vehicle based on the deep neural network regression model.
所述的深度神经网络回归模型损失函数的计算式为:The calculation formula of the loss function of the deep neural network regression model is:
所述的第二准确度算法的计算式与深度神经网络回归模型损失函数的计算式相同。The calculation formula of the second accuracy algorithm is the same as the calculation formula of the loss function of the deep neural network regression model.
所述的优化算法为AdaGrad和Adam。The optimization algorithms described are AdaGrad and Adam.
与现有技术相比,本发明具有以下优点:Compared with the prior art, the present invention has the following advantages:
(1)提出了溯源的时空分析框架,即深度神经网络多分类模型和深度神经网络回归模型,能够避免随着溯源距离上升时,基于交叉口转弯比例的溯源方法中误差逐级提升问题。(1) A spatio-temporal analysis framework for traceability, namely, deep neural network multi-classification model and deep neural network regression model, can avoid the problem of gradual increase in errors in the traceability method based on the ratio of intersection turns when the traceability distance increases.
(2)基于深度神经网络,相比传统机器学习算法,在推理准确度上能够明显提高。(2) Based on deep neural networks, compared with traditional machine learning algorithms, the inference accuracy can be significantly improved.
(3)考虑拥堵区域交通流的来源信息,从而具备从网络层面进行缓堵的能力,提供了缓解拥堵的新研究视角。(3) Consider the source information of the traffic flow in the congested area, so as to have the ability to alleviate the congestion from the network level, and provide a new research perspective on alleviating congestion.
(4)自动车辆识别器定点设置,只依赖定点检测设备的数据,具有较好的适应性。(4) The fixed-point setting of the automatic vehicle identifier only relies on the data of the fixed-point detection equipment, which has good adaptability.
图1为本发明的流程图;Figure 1 is a flow chart of the present invention;
图2为本发明实施例的溯源示意路网图;2 is a schematic road network diagram of traceability according to an embodiment of the present invention;
图3为本发明实施例的空间误差示意图;FIG. 3 is a schematic diagram of the spatial error of an embodiment of the present invention;
图4为本发明实施例的深度神经网络多分类模型输入示意图;4 is a schematic diagram of inputting a deep neural network multi-classification model according to an embodiment of the present invention;
图5为本发明实施例与传统机器学习溯源结果对比图。Fig. 5 is a comparison diagram of traceability results between an embodiment of the present invention and traditional machine learning.
下面结合附图和具体实施例对本发明进行详细说明。本实施例以本发明技术方案为前提进行实施,给出了详细的实施方式和具体的操作过程,但本发明的保护范围不限于下述的实施例。The present invention will be described in detail below with reference to the drawings and specific embodiments. This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation mode and specific operation process are given, but the protection scope of the present invention is not limited to the following embodiments.
实施例Example
本实施例提供一种拥堵交通流溯源分析方法,如图1所示,包括两个步骤:This embodiment provides a method for tracing the source of congested traffic flow, as shown in FIG. 1, which includes two steps:
步骤S1:基于拥堵区域的车辆的自动车辆识别器数据和车辆道路来源数据,构建深度神经网络多分类模型,得到车辆的空间来源;Step S1: Based on the automatic vehicle identifier data and vehicle road source data of vehicles in the congested area, construct a deep neural network multi-classification model to obtain the spatial source of the vehicle;
步骤S2:基于车辆的空间来源和自动车辆识别器数据,构建深度神经网络回归模型,得到车辆的时间溯源结果。Step S2: Based on the spatial source of the vehicle and the data of the automatic vehicle identifier, a deep neural network regression model is constructed to obtain the time traceability result of the vehicle.
具体而言:in particular:
一、步骤S1包括:1. Step S1 includes:
步骤S11:将自动车辆识别器数据和车辆道路来源数据进行独热编码,得到自动车辆识别器独热编码数据和车辆道路来源独热编码数据;Step S11: Perform one-hot encoding on the automatic vehicle identifier data and the vehicle road source data to obtain the automatic vehicle identifier one-hot encoding data and the vehicle road source one-hot encoding data;
步骤S12:构建与空间来源有关的深度神经网络多分类模型损失函数;Step S12: Construct a loss function of a deep neural network multi-classification model related to the spatial source;
步骤S13:基于自动车辆识别器独热编码数据、车辆道路来源独热编码数据和深度神经网络多分类模型损失函数,通过优化算法和第一准确度算法得到深度神经网络多分类模型;Step S13: Based on the one-hot encoding data of the automatic vehicle recognizer, the one-hot encoding data of the source of the vehicle and the loss function of the deep neural network multi-classification model, the deep neural network multi-classification model is obtained through the optimization algorithm and the first accuracy algorithm;
步骤S14:基于深度神经网络多分类模型,得到车辆的空间来源。Step S14: Obtain the spatial source of the vehicle based on the multi-classification model of the deep neural network.
其中,深度神经网络多分类模型基于深度学习分类器(DNN Classifier);Among them, the deep neural network multi-classification model is based on the deep learning classifier (DNN Classifier);
进一步地,设
为距离待溯源路段一定空间距离的边界路段集合,第m条路段的标签即为m,例如,若有11条边界路段,则m=1,2,...,11,且这11条边界路段两 两相邻,构成城市路网中的一个子网络,车辆道路来源数据即为子网络的数据,研究针对这个子网络范围内的车辆进行研究。定义ω为车辆的编号,车辆的空间来源为边界路段集
中任一边界路段。定义
为空间误差,
的值表示车辆的真实空间来源与深度神经网络多分类模型推测得到的车辆的空间来源之间所间隔的边界路段数,因此
只可能为非负整数。
Further, suppose It is a set of boundary road sections with a certain spatial distance from the road section to be traced, the label of the mth road section is m, for example, if there are 11 boundary road sections, then m=1, 2,...,11, and these 11 boundaries The road sections are adjacent to each other and form a sub-network in the urban road network. The vehicle road source data is the data of the sub-network, and the research is conducted on the vehicles within the sub-network. Define ω as the number of the vehicle, and the spatial source of the vehicle is the set of boundary road segments Any boundary section in the. definition Is the spatial error, The value of represents the number of boundary road sections between the true spatial source of the vehicle and the spatial source of the vehicle inferred by the deep neural network multi-classification model, so It can only be a non-negative integer.
步骤S11中使用独热编码(One-hot encoding)技术处理深度神经网络多分类模型的输入数据格式,输入量为自动车辆识别器数据和车辆道路来源数据,例如,车辆ω的输入自动车辆识别器数据为特征矢量
其中μ表示自动车辆识别器的编号。若车辆ω经过自动车辆识别器μ,则ω
μ=1,否则,ω
μ=0。
In step S11, one-hot encoding technology is used to process the input data format of the deep neural network multi-classification model, and the input is automatic vehicle identifier data and vehicle road source data, for example, input of vehicle ω to automatic vehicle identifier Data as feature vector Where μ represents the number of the automatic vehicle identifier. If the vehicle ω passes the automatic vehicle identifier μ, then ω μ =1, otherwise, ω μ =0.
定义深度神经网络多分类模型输出的标签为
其表示车辆的空间来源,具体表现为边界路段集合
中的某一条路段。在标签中,1表示车辆的空间来源,且一个矢量中有且仅有一个1,其余均为0。例如:
表示车辆的来源为第3个边界路段(m=3)。
Define the output label of the deep neural network multi-class model as It represents the spatial source of the vehicle, and is specifically expressed as a collection of boundary road sections A certain road section in. In the label, 1 represents the space source of the vehicle, and there is one and only one 1 in a vector, and the rest are all 0s. E.g: Indicates that the source of the vehicle is the third boundary section (m=3).
步骤S12中深度神经网络多分类模型损失函数的具体计算式为The specific calculation formula of the loss function of the deep neural network multi-classification model in step S12 is
其中,N表示车辆的数量;m表示空间来源的标签编号;y
ωm表示车辆的空间来源,y
ωm=1表示空间来源m为车辆ω的正确空间来源,反之则y
ωm=0;p
ωm表示车辆ω属于空间来源m的概率。
Among them, N represents the number of vehicles; m represents the tag number of the spatial source; y ωm represents the spatial source of the vehicle, y ωm =1 indicates that the spatial source m is the correct spatial source of the vehicle ω, otherwise y ωm =0; p ωm represents The probability that the vehicle ω belongs to the spatial source m.
步骤S13中:深度神经网络算法本质是通过找到负梯度,不断迭代直至找到最优解,这个过程称为梯度下降,本方法采用谷歌开源代码机器学习库TensorFlow中最常用的优化算法AdaGrad与Adam。Step S13: The essence of the deep neural network algorithm is to find the negative gradient and iterate until the optimal solution is found. This process is called gradient descent. This method uses the most commonly used optimization algorithms AdaGrad and Adam in the Google open source machine learning library TensorFlow.
定义一条边界路段及其两侧相邻的边界路段共同构成一个空间来源区域,用EE
ω表示车辆ω的空间来源区域的正确性。
Define a boundary road segment and the adjacent boundary road segments on both sides to form a space source area, and use EE ω to represent the correctness of the space source area of the vehicle ω.
当深度神经网络多分类模型推测的车辆的空间来源在车辆的真实空间来源所在的空间来源区域内时
即认为模型对车辆的空间来源获得了准确推测,即EE
ω=1;当模型推测得的车辆的空间来源在车辆的真实空间来源所在的空间来源区域之外时
即认为深度神经网络多分类模型对车辆的空间来源未获得准确推测,即EE
ω=0。
When the space source of the vehicle predicted by the deep neural network multi-classification model is in the space source area where the true space source of the vehicle is located That is to say, it is considered that the model has obtained an accurate estimation of the space source of the vehicle, that is, EE ω = 1; when the space source of the vehicle estimated by the model is outside the space source area where the real space source of the vehicle is located That is to say, it is believed that the deep neural network multi-classification model has not obtained an accurate estimation of the spatial source of the vehicle, that is, EE ω =0.
上述内容可表述为如下公式:The above content can be expressed as the following formula:
进而,第一准确度算法计算公式如下:Furthermore, the calculation formula of the first accuracy algorithm is as follows:
其中,SEA为准确度。Among them, SEA is accuracy.
二、步骤S2包括:2. Step S2 includes:
步骤S21:将车辆的空间来源和自动车辆识别器数据进行独热编码,得到独热编码空间来源和自动车辆识别器独热编码数据;Step S21: Perform one-hot encoding on the space source of the vehicle and the data of the automatic vehicle identifier to obtain the one-hot encoding space source and the one-hot encoding data of the automatic vehicle identifier;
步骤S22:构建与时间溯源结果有关的深度神经网络回归模型损失函数;Step S22: Construct a loss function of the deep neural network regression model related to the time traceability result;
步骤S23:基于自动车辆识别器独热编码数据、独热编码空间来源和深度神经网络回归模型损失函数,通过优化算法和第二准确度算法得到深度神经网络回归模型;Step S23: Based on the one-hot encoding data of the automatic vehicle identifier, the one-hot encoding space source and the loss function of the deep neural network regression model, the deep neural network regression model is obtained through the optimization algorithm and the second accuracy algorithm;
步骤S24:基于深度神经网络回归模型,得到车辆的时间溯源结果。Step S24: Obtain the time traceability result of the vehicle based on the deep neural network regression model.
其中,深度神经网络回归模型基于深度学习回归器(DNN Regressor)。Among them, the deep neural network regression model is based on the deep learning regressor (DNN Regressor).
进一步地,设车辆ω到达待溯源路段的时刻为
定义行程时间表示车辆ω从起始边界路段开始,到达待溯源路段所经过的时间。由于起始路段不一定会有自动车辆识别检测器,因此,在时间溯源模型中,采用回归的方式来推测行程时间,定义深度神经网络回归模型得到的行程时间(即时间溯源结果)为
Further, suppose that the time when the vehicle ω reaches the road section to be traced is The defined travel time represents the elapsed time for the vehicle ω from the starting boundary road segment to reach the road segment to be traced. Since the initial road section does not necessarily have an automatic vehicle recognition detector, in the time traceability model, a regression method is used to estimate the travel time, and the travel time obtained by the deep neural network regression model (that is, the time traceability result) is defined as
步骤S21中深度神经网络回归模型的输入信息依旧采用独热编码的形式,定义
为输入信息,其主要包含两部分:第一部分
是深度神经网络多分类模型的输出结果,即
第二部分
包含的信息是车辆ω第一次在子网络被检测到的检测器编号,以及到达待溯源路段之间的时间差。例如,设
为车辆ω第一次被子网络中的检测器μ检测到的时间戳,则有
其中,
所包含的元素个数等于子路网中拥有的检测器个数,
为
的第μ个元素,代表其被μ号检测器捕捉到,其余元素均为0。
The input information of the deep neural network regression model in step S21 is still in the form of one-hot encoding, which defines To input information, it mainly consists of two parts: the first part Is the output result of the deep neural network multi-classification model, namely the second part The information contained is the detector number where the vehicle ω was first detected in the sub-network, and the time difference between reaching the road segment to be traced. For example, suppose Is the timestamp when the vehicle ω is first detected by the detector μ in the sub-network, then among them, The number of elements contained is equal to the number of detectors in the sub-road network, for The μth element of represents that it was captured by the μ detector, and the remaining elements are all 0.
步骤S22中深度神经网络回归模型损失函数具体计算公式如下:The specific calculation formula of the loss function of the deep neural network regression model in step S22 is as follows:
步骤S23中同样采用谷歌开源代码机器学习库TensorFlow中的AdaGrad与Adam 作为模型优化算法。In step S23, AdaGrad and Adam in Google's open source machine learning library TensorFlow are also used as model optimization algorithms.
第二准确度定义为TEE,其算法与深度神经网络回归模型损失函数相同,即:The second accuracy is defined as TEE, and its algorithm is the same as the loss function of the deep neural network regression model, namely:
下面结合一个具体的例子说明本方法:The following is a specific example to illustrate this method:
如图2所示,为示例应用场景,该子网络由25个交叉口及若干条路段组成,在路段上分布着若干自动车辆识别器。其中,D
μ(μ=1,2,...,10)表示第μ个自动车辆识别器,灰色圆圈代表普通交叉口,黑色圆圈代表边界交叉口,与边界交叉口相邻的黑色虚线路段为边界路段,由边界路段构成的集合为
待溯源路段为r
14-15。
As shown in Figure 2, it is an example application scenario. The sub-network is composed of 25 intersections and several road sections, and several automatic vehicle identifiers are distributed on the road sections. Among them, D μ (μ=1, 2,...,10) represents the μth automatic vehicle identifier, the gray circle represents the ordinary intersection, the black circle represents the boundary intersection, and the black dashed line segment adjacent to the boundary intersection Is a boundary section, and the set of boundary sections is The section to be traced is r 14-15 .
若在图2子网络中,车辆的真实空间来源是r
3-8,则深度神经网络多分类模型的不同推测的空间误差数值如图3中
一栏所示。可以看到,空间误差均为非负整数。
If in the sub-network of Figure 2, the true spatial source of the vehicle is r 3-8 , then the spatial error values of different inferences of the deep neural network multi-classification model are shown in Figure 3. As shown in one column. It can be seen that the spatial errors are all non-negative integers.
图2中给出了两条示例轨迹(I和II),它们在这个子网络内的起点边界路段为l
2和l
3,均经过待溯源路段r
14-15。
Two example trajectories (I and II) are shown in Figure 2. Their starting and boundary road sections in this sub-network are l 2 and l 3 , and they all pass through the road section to be traced r 14-15 .
以图2中的两条示例轨迹为例,图4为轨迹I与轨迹II在深度神经网络多分类模型中的输入自动车辆识别器数据形式,若车辆经过带有自动车辆识别器的路段,则对应元素的值为1,否则为0。Taking the two example trajectories in Figure 2 as an example, Figure 4 shows the input automatic vehicle identifier data form of the trajectory I and trajectory II in the deep neural network multi-classification model. If the vehicle passes through the road section with the automatic vehicle identifier, then The value of the corresponding element is 1, otherwise it is 0.
如图5所示,将本方法所采用的基于深度神经网络的分类和回归,与基于传统机器学习的分类和回归进行了对比。结果发现,基于深度神经网络的分类和回归在效果上全面优于基于传统机器学习的分类和回归的效果。As shown in Figure 5, the classification and regression based on deep neural networks used in this method is compared with the classification and regression based on traditional machine learning. The results found that the classification and regression based on deep neural networks are overall better than those based on traditional machine learning.
Claims (8)
- 一种拥堵交通流溯源分析方法,其特征在于,该方法包括以下步骤:A method for tracing the source of congested traffic flow, which is characterized in that the method includes the following steps:步骤S1:基于拥堵区域的车辆的自动车辆识别器数据和车辆道路来源数据,构建深度神经网络多分类模型,得到车辆的空间来源;Step S1: Based on the automatic vehicle identifier data and vehicle road source data of vehicles in the congested area, construct a deep neural network multi-classification model to obtain the spatial source of the vehicle;步骤S2:基于车辆的空间来源和自动车辆识别器数据,构建深度神经网络回归模型,得到车辆的时间溯源结果。Step S2: Based on the spatial source of the vehicle and the data of the automatic vehicle identifier, a deep neural network regression model is constructed to obtain the time traceability result of the vehicle.
- 根据权利要求1所述的一种拥堵交通流溯源分析方法,其特征在于,所述的步骤S1包括:The method for tracing the source of congested traffic flow according to claim 1, wherein said step S1 comprises:步骤S11:将自动车辆识别器数据和车辆道路来源数据进行独热编码,分别得到自动车辆识别器独热编码数据和车辆道路来源独热编码数据;Step S11: Perform one-hot encoding on the automatic vehicle identifier data and the vehicle road source data to obtain the automatic vehicle identifier one-hot encoding data and the vehicle road source one-hot encoding data respectively;步骤S12:构建与空间来源有关的深度神经网络多分类模型损失函数;Step S12: Construct a loss function of a deep neural network multi-classification model related to the spatial source;步骤S13:基于自动车辆识别器独热编码数据、车辆道路来源独热编码数据和深度神经网络多分类模型损失函数,通过优化算法和第一准确度算法得到深度神经网络多分类模型;Step S13: Based on the one-hot encoding data of the automatic vehicle recognizer, the one-hot encoding data of the source of the vehicle and the loss function of the deep neural network multi-classification model, the deep neural network multi-classification model is obtained through the optimization algorithm and the first accuracy algorithm;步骤S14:基于深度神经网络多分类模型,得到车辆的空间来源。Step S14: Obtain the spatial source of the vehicle based on the multi-classification model of the deep neural network.
- 根据权利要求2所述的一种拥堵交通流溯源分析方法,其特征在于,所述的深度神经网络多分类模型损失函数的计算式为:The method for tracing the source of congested traffic flow according to claim 2, wherein the calculation formula of the loss function of the deep neural network multi-classification model is:其中,N为车辆的数量,m为空间来源的标签编号,p ωm为车辆ω属于空间来源m的概率;y ωm为空间来源,y ωm=1表示空间来源m为车辆ω的正确空间来源,y ωm=0表示空间来源m不是车辆ω的正确空间来源。 Among them, N is the number of vehicles, m is the tag number of the spatial source, p ωm is the probability that the vehicle ω belongs to the spatial source m; y ωm is the spatial source, and y ωm =1 indicates that the spatial source m is the correct spatial source of the vehicle ω, y ωm = 0 means that the space source m is not the correct space source of the vehicle ω.
- 根据权利要求2所述的一种拥堵交通流溯源分析方法,其特征在于,第一准确度计算方法为:The method for tracing the source of congested traffic flow according to claim 2, wherein the first accuracy calculation method is:其中,EE ω表示车辆ω的空间来源区域的正确性,所述的空间来源区域包括一条边界路段及其两侧相邻的边界路段,N为车辆的数量,SEA为准确度。 Among them, EE ω represents the correctness of the spatial source area of the vehicle ω. The spatial source area includes a boundary road segment and its adjacent boundary road segments on both sides, N is the number of vehicles, and SEA is the accuracy.
- 根据权利要求1所述的一种拥堵交通流溯源分析方法,其特征在于,所述的 步骤S2包括:The method for tracing the source of congested traffic flow according to claim 1, wherein said step S2 comprises:步骤S21:将车辆的空间来源和自动车辆识别器数据进行独热编码,得到独热编码空间来源和自动车辆识别器独热编码数据;Step S21: Perform one-hot encoding on the space source of the vehicle and the data of the automatic vehicle identifier to obtain the one-hot encoding space source and the one-hot encoding data of the automatic vehicle identifier;步骤S22:构建与时间溯源结果有关的深度神经网络回归模型损失函数;Step S22: Construct a loss function of the deep neural network regression model related to the time traceability result;步骤S23:基于自动车辆识别器独热编码数据、独热编码空间来源和深度神经网络回归模型损失函数,通过优化算法和第二准确度算法得到深度神经网络回归模型;Step S23: Based on the one-hot encoding data of the automatic vehicle identifier, the one-hot encoding space source and the loss function of the deep neural network regression model, the deep neural network regression model is obtained through the optimization algorithm and the second accuracy algorithm;步骤S24:基于深度神经网络回归模型,得到车辆的时间溯源结果。Step S24: Obtain the time traceability result of the vehicle based on the deep neural network regression model.
- 根据权利要求5所述的一种拥堵交通流溯源分析方法,其特征在于,所述的深度神经网络回归模型损失函数的计算式为:The method for tracing the source of congested traffic flow according to claim 5, wherein the calculation formula of the loss function of the deep neural network regression model is:
- 根据权利要求5所述的一种拥堵交通流溯源分析方法,其特征在于,所述的第二准确度算法的计算式与深度神经网络回归模型损失函数的计算式相同。The method for tracing the source of congested traffic flow according to claim 5, wherein the calculation formula of the second accuracy algorithm is the same as the calculation formula of the loss function of the deep neural network regression model.
- 根据权利要求5所述的一种拥堵交通流溯源分析方法,其特征在于,所述的优化算法为AdaGrad和Adam。The method for tracing the source of congested traffic flow according to claim 5, wherein the optimization algorithms are AdaGrad and Adam.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910978947.2A CN110889427B (en) | 2019-10-15 | 2019-10-15 | Congestion traffic flow traceability analysis method |
CN201910978947.2 | 2019-10-15 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2021073524A1 true WO2021073524A1 (en) | 2021-04-22 |
Family
ID=69746149
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2020/120829 WO2021073524A1 (en) | 2019-10-15 | 2020-10-14 | Analysis method for tracing source of congestion traffic flow |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN110889427B (en) |
WO (1) | WO2021073524A1 (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115311854A (en) * | 2022-07-22 | 2022-11-08 | 东南大学 | Vehicle space-time trajectory reconstruction method based on data fusion |
CN116580586A (en) * | 2023-07-12 | 2023-08-11 | 中南大学 | Vehicle path induction method and system for balancing personal benefits and social benefits |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110889427B (en) * | 2019-10-15 | 2023-07-07 | 同济大学 | Congestion traffic flow traceability analysis method |
CN112328791A (en) * | 2020-11-09 | 2021-02-05 | 济南大学 | Text classification method of Chinese government affair information based on DiTextCNN |
CN113920719B (en) * | 2021-09-09 | 2022-09-30 | 青岛海信网络科技股份有限公司 | Traffic tracing method and electronic equipment |
CN116580563B (en) * | 2023-07-10 | 2023-09-22 | 中南大学 | Markov chain-based regional congestion traffic source prediction method, device and equipment |
CN117010667B (en) * | 2023-09-27 | 2024-02-27 | 深圳市城市交通规划设计研究中心股份有限公司 | Road traffic emission space tracing method, electronic equipment and storage medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105882695A (en) * | 2016-03-17 | 2016-08-24 | 北京交通大学 | Foresight associated control method for passenger flow congestion of urban railway traffic network |
CN106652441A (en) * | 2015-11-02 | 2017-05-10 | 杭州师范大学 | Urban road traffic condition prediction method based on spatial-temporal data |
US20190179328A1 (en) * | 2017-12-11 | 2019-06-13 | Volvo Car Corporation | Path prediction for a vehicle |
CN110111574A (en) * | 2019-05-16 | 2019-08-09 | 北京航空航天大学 | A kind of urban transportation imbalance evaluation method based on the analysis of flow tree |
CN110287995A (en) * | 2019-05-27 | 2019-09-27 | 同济大学 | The multiple features learning network model method of whole day overhead transport jam situation classification |
CN110889427A (en) * | 2019-10-15 | 2020-03-17 | 同济大学 | Congestion traffic flow traceability analysis method |
Family Cites Families (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2001283373A (en) * | 2000-03-30 | 2001-10-12 | Toshiba Corp | Traffic flow measuring system |
JP6558239B2 (en) * | 2015-12-22 | 2019-08-14 | アイシン・エィ・ダブリュ株式会社 | Automatic driving support system, automatic driving support method, and computer program |
CN106856049B (en) * | 2017-01-20 | 2020-04-24 | 东南大学 | Key intersection demand aggregation analysis method based on bayonet number plate identification data |
CN107564291A (en) * | 2017-10-20 | 2018-01-09 | 重庆市市政设计研究院 | A kind of volume of traffic Source Tracing method and system based on RFID |
CN108509978B (en) * | 2018-02-28 | 2022-06-07 | 中南大学 | Multi-class target detection method and model based on CNN (CNN) multi-level feature fusion |
CN109101997B (en) * | 2018-07-11 | 2020-07-28 | 浙江理工大学 | Traceability method for sampling limited active learning |
CN109361617B (en) * | 2018-09-26 | 2022-09-27 | 中国科学院计算机网络信息中心 | Convolutional neural network traffic classification method and system based on network packet load |
CN109087510B (en) * | 2018-09-29 | 2021-09-07 | 讯飞智元信息科技有限公司 | Traffic monitoring method and device |
CN109191849B (en) * | 2018-10-22 | 2020-10-09 | 北京航空航天大学 | Traffic jam duration prediction method based on multi-source data feature extraction |
CN109448367B (en) * | 2018-10-22 | 2020-04-10 | 南京理工大学 | Intelligent road traffic tracking management system based on big data image acquisition |
CN109492588A (en) * | 2018-11-12 | 2019-03-19 | 广西交通科学研究院有限公司 | A kind of rapid vehicle detection and classification method based on artificial intelligence |
CN109639739B (en) * | 2019-01-30 | 2020-05-19 | 大连理工大学 | Abnormal flow detection method based on automatic encoder network |
CN110136435A (en) * | 2019-04-17 | 2019-08-16 | 青岛大学 | A kind of congestion networking propagation model for infecting threshold value, more propagating and deposit more |
-
2019
- 2019-10-15 CN CN201910978947.2A patent/CN110889427B/en active Active
-
2020
- 2020-10-14 WO PCT/CN2020/120829 patent/WO2021073524A1/en active Application Filing
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106652441A (en) * | 2015-11-02 | 2017-05-10 | 杭州师范大学 | Urban road traffic condition prediction method based on spatial-temporal data |
CN105882695A (en) * | 2016-03-17 | 2016-08-24 | 北京交通大学 | Foresight associated control method for passenger flow congestion of urban railway traffic network |
US20190179328A1 (en) * | 2017-12-11 | 2019-06-13 | Volvo Car Corporation | Path prediction for a vehicle |
CN110111574A (en) * | 2019-05-16 | 2019-08-09 | 北京航空航天大学 | A kind of urban transportation imbalance evaluation method based on the analysis of flow tree |
CN110287995A (en) * | 2019-05-27 | 2019-09-27 | 同济大学 | The multiple features learning network model method of whole day overhead transport jam situation classification |
CN110889427A (en) * | 2019-10-15 | 2020-03-17 | 同济大学 | Congestion traffic flow traceability analysis method |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115311854A (en) * | 2022-07-22 | 2022-11-08 | 东南大学 | Vehicle space-time trajectory reconstruction method based on data fusion |
CN115311854B (en) * | 2022-07-22 | 2023-08-25 | 东南大学 | Vehicle space-time track reconstruction method based on data fusion |
CN116580586A (en) * | 2023-07-12 | 2023-08-11 | 中南大学 | Vehicle path induction method and system for balancing personal benefits and social benefits |
CN116580586B (en) * | 2023-07-12 | 2023-10-13 | 中南大学 | Vehicle path induction method and system for balancing personal benefits and social benefits |
Also Published As
Publication number | Publication date |
---|---|
CN110889427A (en) | 2020-03-17 |
CN110889427B (en) | 2023-07-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2021073524A1 (en) | Analysis method for tracing source of congestion traffic flow | |
US11783594B2 (en) | Method of segmenting pedestrians in roadside image by using convolutional network fusing features at different scales | |
CN109584553B (en) | Road section relevance missing completion method based on space-time information | |
US20240149882A1 (en) | Multiple exposure event determination | |
WO2021249071A1 (en) | Lane line detection method, and related apparatus | |
WO2023147786A1 (en) | Internet-of-vehicles intrusion detection method and device based on improved convolutional neural network | |
CN113077626B (en) | Lane selection guiding method, device, system, equipment and storage medium | |
CN112613225A (en) | Intersection traffic state prediction method based on neural network cell transmission model | |
Wang et al. | Vehicle reidentification with self-adaptive time windows for real-time travel time estimation | |
Kim et al. | Visual extensions and anomaly detection in the pNEUMA experiment with a swarm of drones | |
CN114694078A (en) | Traffic behavior judgment method based on multi-target tracking | |
Kumar et al. | Moving Vehicles Detection and Tracking on Highways and Transportation System for Smart Cities | |
CN110634289B (en) | Urban road traffic optimal path online planning method based on electric police data | |
CN110610118A (en) | Traffic parameter acquisition method and device | |
CN116386020A (en) | Method and system for predicting exit flow of highway toll station by multi-source data fusion | |
Akhand et al. | Traffic density estimation using transfer learning with pre-trained inceptionresnetv2 network | |
Priya et al. | Road traffic condition monitoring using deep learning | |
CN112347938B (en) | People stream detection method based on improved YOLOv3 | |
Zhang et al. | Robust vehicle detection and identification with single magnetic sensor | |
CN114219970A (en) | Image processing method and system for traffic management | |
Bhuptani et al. | Automating traffic signals based on traffic density estimation in bangalore using YOLO | |
Jain et al. | Machine Learning-Based Real-Time Traffic Control System | |
CN116631186B (en) | Expressway traffic accident risk assessment method and system based on dangerous driving event data | |
CN114662792B (en) | Traffic flow prediction method of recurrent neural network based on dynamic diffusion graph convolution | |
Li et al. | Multi-hierarchical spatial-temporal graph convolutional networks for traffic flow forecasting |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 20878008 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 20878008 Country of ref document: EP Kind code of ref document: A1 |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 20878008 Country of ref document: EP Kind code of ref document: A1 |