A method for identifying highway congestion based on ETC transaction data

Abstract

The application discloses a highway congestion identification method based on ETC transaction data, constructs a section data set based on ETC track data, constructs three data dimensions after analyzing the traffic characteristics of the highway, carries out data weighting by using an analytic hierarchy process, and finally combines a section label and a fuzzy evaluation to evaluate a score, so that the integrity and accuracy of the congestion section are guaranteed. The application avoids the congestion misjudgment problem of some sections to a great extent, and provides a certain degree of data support and auxiliary decision for relieving the highway congestion problem.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of highway management, and particularly relates to a highway congestion identification method based on ETC transaction data. BACKGROUND

[0002] The current traffic congestion judgment algorithm can be mainly divided into three categories: (1) direct detection method, such as directly judging the congestion of the highway section through video, which requires too many cameras and has high cost. Ameni Chetouan[5]formulated four research schemes based on monitoring video to detect congestion, and the implementation effect of congestion can be proportional to the requirement of hardware, and the dimension of recognition is poor, which is difficult to achieve the desired congestion recognition effect; (2) indirect detection method, mainly detecting the existence of events according to the influence of traffic indicators on traffic flow, such as analyzing the congestion of the current section through common traffic indicators such as speed and flow. This method has low cost and simple operation, but the detection rate is low and the false alarm rate is high. W. Pattara-atikom[7]uses speed to divide congestion levels, and Ahanin[8]also clusters the speed of each time period in the section, and finally uses MDL principle to estimate the missing traffic state. However, this method has obvious test effect fluctuation, and the traffic state does not fluctuate like traffic volume every day, resulting in low detection rate of the recognition method through indicators. As Wei-Hua Lin[9]once said, the effect of detecting traffic volume to detect traffic state is not as significant as directly detecting traffic state. Therefore, in order to identify the traffic congestion of the highway, the traffic state must be directly identified. Therefore, most people began to distinguish the traffic state through traffic indicators. Jianzhen Liu

[10] established a traffic state recognition model based on speed and traffic density to directly distinguish the traffic state, and the overall accuracy rate was above 95%. Yue Tu

[11] generated congestion index by mining free flow speed and flow, Hong Gao

[12] explored the time pattern of multiple congestion points and the spatial pattern of frequently congested sections, and analyzed multiple indicators, which achieved good results. Wan-Xiang Wang

[13] proposed three calculation methods of traffic congestion index, and compared them to adopt a more reasonable scheme. However, the above methods ignore the correlation between traffic indicators, and often produce congestion misjudgment when identifying congestion, resulting in that congestion events cannot be avoided and the recognition effect is not ideal. (3) Based on theoretical model: design the algorithm of judging traffic indicators, which includes some mature theoretical models such as gray system theory, cluster analysis and fuzzy comprehensive evaluation.

[0003] ETC gate generated transaction data is the primary guarantee for the realization of highway congestion identification algorithm, which can cover almost all sections of the highway traffic situation. At present, the average use rate of ETC network has exceeded 66% [1], ETC gate can automatically identify cars and upload ETC intelligent transaction data to provide data support and strong guarantee for highway management department [2]. The current congestion identification data sources mainly use video data and floating car data, such as Xiangjie Kong [4] uses particle swarm optimization algorithm based on floating car data to detect and predict traffic congestion, but the trajectory data dimension is simple, the data is complicated, and the detection and prediction of congestion have certain limitations; Ameni Chetouan [5] based on monitoring video developed four research programs to detect congestion, the implementation effect of congestion is proportional to the requirement of hardware, and the recognition dimension is small, which is difficult to achieve the desired congestion identification. Compared with these floating car data and video data, ETC system generates a large amount of ETC data, with 103 types of data, wide coverage, real-time, high reliability, easy analysis and processing, and the generated results are rich and have obvious advantages, which is considered as an important measure to solve this problem [6].

[0004] In the past few decades, people have done a lot of research on traffic congestion, and now most of the research tends to achieve the effect of identifying highway congestion by predicting traffic volume. W. Pattara-atikom [7] uses speed to divide congestion levels, and Ahanin [8] also clusters the speed of each time period in the road section, and finally uses MDL principle to estimate the missing traffic state. But this test effect fluctuates obviously, and the traffic state does not fluctuate like traffic volume every day, just like Wei-Hua Lin [9] said, detecting traffic volume to achieve the effect of detecting traffic state is not as significant as directly detecting traffic state. Therefore, to identify highway traffic congestion, it is best to directly identify the state. Therefore, most people began to distinguish traffic state by traffic indicators. Jianzhen Liu

[10] established a traffic state identification model based on speed and traffic density to directly distinguish traffic state, with an overall accuracy rate of more than 95%, Yue Tu

[11] generated congestion index by mining free flow speed and flow, Hong Gao

[12] explored the time pattern of multiple congestion points and the spatial pattern of frequently congested road sections, and used multiple indicators for analysis, which achieved good results, and Wan-Xiang Wang

[13] proposed three calculation methods of traffic congestion index, compared with each other to take a more reasonable scheme. But the above methods ignore the correlation between traffic indicators, which often leads to congestion misjudgment when identifying congestion, resulting in congestion events that cannot be avoided, and the identification effect is not ideal.

[0005] Trinh

[14] pointed out that fuzzy logic is a qualitative method that is close to human observation, reasoning and decision-making. Fuzzy comprehensive evaluation uses the weight ratio of each traffic indicator to evaluate the fuzzy information in the interval based on fuzzy logic. Maja Kalinic

[15] used fuzzy inference model to solve the correlation problem between variables. He used flow and density as input to the model and the detection effect was relatively ideal. The author concluded that fuzzy inference model is flexible in dealing with subjectivity, fuzziness, imprecision and uncertainty, but ignores the differences between traffic indicators. Khaliun

[16] used AHP to weigh the weight of variables and combined it with FCE to detect congestion, but ignored the road characteristics, resulting in an insignificant effect. Zhu Dandan

[17] used fuzzy comprehensive evaluation to identify the state of congestion based on the floating car driving characteristics and considered the influence of surrounding roads, service areas, merging and diverging, etc., but did not combine the road characteristics with weights to detect congestion, and could not achieve congestion identification well. Summary of the Invention

[0006] The purpose of this invention is to provide a method for identifying highway congestion based on ETC transaction data.

[0007] The technical solution adopted in this invention is:

[0008] A method for identifying highway congestion based on ETC transaction data includes the following steps:

[0009] Step 1: Obtain highway ETC transaction data, topology data, and highway vehicle trajectory data.

[0010] Step 2: Based on the spatiotemporal information of ETC transaction data, match highway ETC transaction data with topological data to construct highway vehicle trajectory data and segment datasets.

[0011] Step 3: Perform data cleaning on the highway vehicle trajectory data to remove unnecessary data;

[0012] Step 4: Construct a vehicle trajectory set TrajS from the cleaned ETC transaction data in chronological order. The vehicle trajectory set refers to the set of multiple ETC gantries that all vehicles pass through during their journey on the highway.

[0013] Step 5: Congestion feature matching was performed between the topology data and the segment dataset to construct segment labels;

[0014] Step 6: Construct and calculate three segment-dimensional parameters of traffic congestion based on the objective laws of highways. The three segment-dimensional parameters are segment average speed, segment flow rate, and segment average delay time. The segment average speed is used as the main evaluation index in the congestion identification method, and the segment flow rate and segment average delay time are used as correction features.

[0015] Step 7: Construct a judgment matrix X for the three segment dimension parameters, calculate the eigenvalues ​​and eigenvectors of the judgment matrix X respectively, and obtain the final weights after passing the consistency judgment test of the discriminant matrix.

[0016] Step 8: Analyze the segment profiles in the highway to obtain the relationship between different types of segment dimension parameters, and assign values ​​to the corresponding traffic congestion weight indicators.

[0017] Step 9: Construct a fuzzy matrix R by analyzing the relationship between traffic indicators and congestion assessment levels.

[0018]

[0019] Where R1, R2, and R3 correspond to the fuzzy matrices of the three segment dimension parameters, namely, the segment average speed, the segment flow rate, and the segment average delay, respectively. 11 ~r 16 This represents the corresponding position value of R1; r 21 ~r 26 This represents the corresponding position value of R2; r 31 ~r 36 This indicates the corresponding position value of R3;

[0020] Step 10: Combine the weights A = [a1, a2, a3] obtained by the analytic hierarchy process with the fuzzy matrix R to obtain matrix B. Based on the principle of maximum membership, select the maximum value score in matrix B as the congestion score C of the segment at this time.

[0021] B = A * R (20)

[0022] Among them, a1, a2, and a3 correspond to the weight values ​​of the three segment dimension parameters: segment average speed, segment flow, and segment average delay, respectively.

[0023] Step 11: Determine whether the congestion score C is greater than the set threshold; if yes, determine that there is no congestion in the current segment; otherwise, determine that there is congestion in the current segment.

[0024] Furthermore, in step 3, Chebyshev's theorem is used to determine the superposition degree SO of the segments. QD Perform data segmentation, removing segments with low overlap from TrajS and retaining SO. QD Congestion information is assessed for sections with high frequency of traffic congestion.

[0025] Further, the segment superposition degree SO QD represents the superposition times of each segment QD on the highway in a certain time period. The segment QD represents the combination of adjacent boundary points DND on the highway into a segment; the boundary points DND include gantries FND, toll stations SND.

[0026] Further, in step 7, the three dimensions are analyzed and weighted using the analytic hierarchy process, and for the same level of an element, the importance of a certain criterion in the previous level is compared two by two, a two-by-two comparison matrix is constructed, and a judgment matrix X is constructed by constructing a matrix of all dimensions.

[0027]

[0028] wherein x 11 ~ x mn represents the importance comparison degree between two elements, which is used to construct the judgment matrix.

[0029] Further, in step 8, the judgment index of the segment image is as follows:

[0030] L QD = {p1, p2, p3, Level} (13)

[0031]

[0032]

[0033]

[0034] wherein L QD is the judgment index of the segment image, p1 captures the uncertainty of the congestion information evaluation caused by the highway toll station, p2 captures the uncertainty of the congestion information evaluation caused by the highway diverging and merging area, and p3 captures the uncertainty of the congestion information evaluation caused by the highway service area; Level is the joint effect coefficient of the segment image, SND, VND, and YND are three uncertainty factors, and Level directly represents the number of uncertainty factors; adjacent DND on the highway are combined into a segment QD, and according to the uncertainty factor UF, the segment type can be further determined, wherein when UF in QD is SND, the segment is called SD, and when UF does not exist, it is a normal segment FD:

[0035] QD = {DND1, UF, DND2} (1)

[0036] SD = {DND1, SND, DND2} (2)

[0037] FD = {DND1, DND2} (3)

[0038] Wherein, DND represents a junction point, DND1 is a first junction point, and DND2 is a second junction point; UF represents an uncertainty factor,

[0039] When UF in QD is VND, the section is referred to as VD;

[0040] VD = {DND1, VND, DND2} (4)

[0041] When UF in QD is YND, the section is referred to as YD

[0042] YD = {DND1, YND, DND2} (5)

[0043] Wherein, FND represents a gantry, SND represents a toll station, VND represents a service area, and YND represents a split-merge point,

[0044] The above formula describes the data index of the influence factors of the expressway.

[0045] Specifically, the section image is performed for the service area, the toll station, the split-merge section, and the joint influence section, so as to avoid the difference between the influence factors of different sections.

[0046] The above technical scheme is adopted, the dimension information of the expressway ETC transaction data is deeply mined, an expressway congestion recognition method combined with a section label is proposed, the fuzzy comprehensive evaluation is improved through the dimension weighting of the analytic hierarchy process and the section feature extraction, and the congestion in the section is accurately recognized. Compared with the traditional congestion recognition method, the method can greatly avoid the congestion misjudgment problem of part of the section, and provides a certain degree of data support and auxiliary decision for relieving the congestion problem of the expressway. BRIEF DESCRIPTION OF DRAWINGS

[0047] The application will be further described in detail below in combination with the drawings and specific embodiments.

[0048] Figure 1 It is a section superposition degree diagram of the application;

[0049] Figure 2 It is a congestion recognition framework diagram of the application;

[0050] Figure 3 It is a hierarchical structure diagram of the application;

[0051] Figure 4 It is a highway network diagram of Fujian Province;

[0052] Figure 5 It is a Chebyshev diagram;

[0053] Figure 6For the service area section speed analysis chart;

[0054] Figure 7 For the traffic index law chart;

[0055] Figure 8 For the section image quantity chart;

[0056] Figure 9 For the congestion identification experiment chart;

[0057] Figure 10 For the video data schematic diagram;

[0058] Figure 11 The application discloses a highway congestion identification method based on ETC transaction data. DETAILED DESCRIPTION

[0059] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0060] Definition 1 transaction node: the highway network is connected by countless nodes ND, and the nodes mainly comprise gantries FND, toll stations SND, service areas VND and diverging and merging points YND, wherein the FND and the SND are collectively referred to as boundary points DND.

[0061] Definition 2 uncertainty factor: the highway contains various types of sections, and the influence factors of the sections are different, wherein the uncertainty factor determines the section type on the highway, and is denoted by UF, and the uncertainty factor comprises the SND, the VND and the YND.

[0062] Definition 3 highway section: adjacent DNDs on the highway are combined into a section QD, and according to the uncertainty factor UF, the section type can be further determined, wherein when the UF in the QD is the SND, the section is referred to as a SD, and when the UF does not exist, the section is referred to as a normal section FD.

[0063] QD={DND1, UF, DND2} (1)

[0064] SD={DND1, SND, DND2} (2)

[0065] FD={DND1, DND2} (3)

[0066] If the UF in the QD is the VND, the section is referred to as a VD;

[0067] VD={DND1, VND, DND2} (4)

[0068] If the UF in the QD is the YND, the section is referred to as a YD

[0069] YD = {DND1, YND, DND2} (5)

[0070] Definition 4 Vehicle Trajectory: Each vehicle on the highway will have a separate trajectory, which is composed of a series of DNDs that the vehicle has passed through. It is called Traj (n≥2)

[0071] Traj = {DND1, DND2,..., DND n} (6)

[0072] Definition 5 Vehicle Trajectory Set: As shown in the above definition, the trajectory of a single vehicle is Traj, and the highway is often driven by multiple vehicles at the same time. The trajectories of multiple vehicles are defined as a set, which is called the vehicle trajectory set TrajS.

[0073] TrajS = {Traj1, Traj2,..., Traj n} (7)

[0074] Definition 6 Delay Time: Each section of the highway will have many vehicles passing through within a certain period of time, but due to congestion or personal habits, the vehicle travel time will be more or less, so there will be a certain delay time, which is represented by

[0075] Delay = d i -d QD (8)

[0076] where d i is the travel time of each vehicle, d QD is the travel time of the free flow in this section, and the section travel time is represented by Delta.

[0077] Definition 7 Section Traffic Index: The traffic index on the highway describes the congestion information in the section of the highway, and the most representative indexes are speed and flow and travel time. This invention uses the average speed of the section and the flow of the section to further evaluate the congestion information of the section, which are represented by Speed and Flow, respectively.

[0078] Definition 8 Section Superposition Degree: In the highway, each QD has different degrees of importance to the road network. The driving route of a single vehicle in the sample time is not the same, but it will pass through the same QD. Therefore, in a certain period of time, each QD has a certain degree of superposition times, and this superposition times often represents the importance of the QD in the overall road network, which is called the section superposition degree SO QD , the section superposition degree is represented as Figure 1 .

[0079] For example,Figures 1 to 11 The application discloses a highway congestion identification method based on ETC transaction data.

[0080] The application proposes a new highway congestion identification framework. The framework is a section congestion information evaluation algorithm based on fuzzy comprehensive evaluation, including 4 modules as shown in the Figure 2 The modules include a highway section spatiotemporal data set construction module, an input module, a section congestion data comprehensive evaluation module, and an output module.

[0081] (1) Highway section spatiotemporal data set construction module: The module is mainly composed of highway ETC transaction data, topological data, and highway vehicle trajectory data, etc. Based on the spatiotemporal information of the ETC transaction data, the application matches the highway ETC transaction data and the topological data to construct the highway vehicle trajectory data set, wherein the abnormal data in the ETC transaction data is filtered, and the application is based on SO QD The vehicle trajectory set is split into a highway section spatiotemporal data set, and module input is performed.

[0082] (2) Input module: The input module uses the section spatiotemporal data set constructed in the road network to input the section congestion data comprehensive evaluation module.

[0083] (3) Section congestion data comprehensive evaluation module: The module uses the weighted data of the AHP model to perform comprehensive scoring evaluation through FCE, performs multi-weight calculation through a unique section image, finally obtains a section congestion comprehensive score, and obtains a congestion level after classification;

[0084] (4) Output module: The output module mainly outputs the congestion spatiotemporal data of each section in the road network, and obtains a congestion identification result after considering objective evaluation.

[0085] ETC data cleaning: The ETC gantry can interact with users, and a large amount of transaction data is generated in a short time. However, due to device failure, weather influence, and system error, ETC transaction data may have data missing and data abnormality. The collected ETC gantry transaction data mainly has the following problems. 1) Data redundancy: There are transaction data with completely same dimension data in the transaction data. 2) Data abnormality: There are data records that do not conform to normal highway cognition, including that the transaction time in the track is almost equal to the last transaction time, the entrance time of the entrance toll station is less than the exit time, and the entrance number cannot be matched with the transaction data. 3) Data missing: The transaction data collected by the vehicle passing through the gantry cannot be effectively collected. These abnormal data greatly reduce the potential value of mining ETC transaction data. In order to better realize the fine detection of highway congestion, the above data needs to be cleaned.

[0086] Segment dataset construction: ETC transaction data is composed of continuous discrete points obtained by sampling, which has discrete and continuous mathematical characteristics. Obviously, it is time-consuming and inefficient to directly mine congestion information from a large amount of transaction data. After constructing the segment data set, the congestion information mined from it is real, effective and intuitive.

[0087] After the transaction data collected by the ETC gantry system is preliminarily cleaned, the vehicle trajectory set TrajS is constructed in time sequence. The vehicle trajectory set refers to the ETC gantry set passed by multiple vehicles during driving on the highway. According to the dimension information of the transaction data of a single trip code, the vehicle trajectory set TrajS is obtained after the de-duplication operation on part of the data. TrajS contains congestion segments and non-congestion segments, wherein the congestion segments account for a small proportion. When the congestion segments are relatively few, the congestion information is necessarily insufficient. In order to more fully mine the congestion information of the segment, the mathematical expectation E(X) = mu and the variance D(X) = sigma 2 According to Chebyshev's theorem,

[0088]

[0089] That is,

[0090]

[0091] Transformed into

[0092]

[0093] Therefore, for deep mining of congestion segments, the present application uses Chebyshev's theorem to cut SO QD Data, select segments with low segment superposition degree, and then remove TrajS to quickly identify segment congestion information in a large amount of ETC trajectory.

[0094] Highway segment congestion information evaluation: In the present application, the congestion of the highway is not determined by a single factor

[17] , and the fuzzy comprehensive evaluation often ignores the correlation between traffic indicators in the application of highway congestion identification. In order to more accurately evaluate the congestion information of different segments of the highway, the present application combines the analytic hierarchy process and proposes an interval portrait to improve the fuzzy comprehensive evaluation, strengthens the connection between traffic indicators and distinguishes the characteristics between different segments to achieve accurate identification, as shown in Algorithm 1.

[0095]

[0096]

[0097] Algorithm 1 algorithm flow

[0098] Analytic hierarchy process: in order to better evaluate the congestion information of the expressway, according to the general objective law on the expressway, three parameters commonly used in traffic congestion, speed, flow and section delay time are constructed, the analytic hierarchy process divides a complex problem into a structure with several layers, for example, object layer, index layer and sub-index layer

[18] It is also a method for solving evaluation problems, and is a more subjective method, and the subjective factor accounts for a large proportion when the weight vector is assigned, and the most influential congestion factor can be selected

[19] In order to ensure the scientific rationality and universality of the expressway congestion identification framework, the analytic hierarchy process is used to analyze and weight the three dimensions. First, the relationship between each factor in the system is analyzed, and the hierarchical structure diagram is constructed as follows: for a single section of the expressway, each factor is interrelated and interrelated, and the average speed of a section has a positive or negative relationship with the traffic flow or the average delay time. Therefore, the importance of a certain element in the same level with respect to a certain criterion in the previous level needs to be compared, and a two-by-two comparison matrix is constructed, so that all dimensions are constructed into a matrix, and a judgment matrix X is constructed.

[0099]

[0100] The importance of the two elements is compared, and the importance is shown in Table 3-1

[0101] Table 3-1 importance comparison table

[0102] Scale Meaning 1 Indicates that two factors are of equal importance compared to each other 3 Indicates that one factor is slightly more important than the other compared to each other 5 Indicates that one factor is significantly more important than the other compared to each other 7 Indicates that one factor is strongly more important than the other compared to each other 9 Indicates that one factor is extremely more important than the other compared to each other 2、4、6、8 The midpoint of the two adjacent judgments above Reciprocal If the scale is 3 for A and B, then it is 1 / 3 for B and A

[0103] The traffic indicators on the expressway have different standards for congestion information judgment, and a large number of papers [20-25] and experimental verification, the judgment matrix X is constructed for the three dimensions, the eigenvalue and eigenvector of the judgment matrix X are calculated, since the scores of the judgment matrix may conflict, the consistency of the judgment matrix needs to be judged before the final weight can be obtained, so the eigenvalue and eigenvector of the matrix are first tested, and any element of the matrix is positively judged, and the result is less than 0.1 after the CR is calculated, the consistency is passed, at this time the weight set A is obtained after the eigenvector is normalized.

[0104] Table 3-2 random consistency index RI

[0105] n 1 2 3 4 5 6 7 8 9 10 11 RI 0 0 0.58 0.90 1.12 1.24 1.32 1.41 1.45 1.49 1.51

[0106] Segment image: the traditional highway congestion recognition method cannot well evaluate the highway segment congestion information, different segment types have different needs for recognition, and the characteristics of the segment need to be further extracted and different weight schemes are constructed. Based on the highway segment data, the characteristics of the segment image are modeled, and the uncertainty of the highway segment congestion information evaluation mainly comes from three different sources, namely the on-ramp, the service area and the toll station, and the traffic flow patterns formed have different local space-time characteristics. First, previous studies have shown [26-27] that service areas or toll stations may cause changes in highway traffic conditions, greatly affecting the accuracy of segment travel time and congestion information evaluation. When the uncertainty factors are not determined, the evaluation of the main congestion information may be difficult to measure and obtain, especially when the uncertainty produced by different influencing factors overlaps, the uncertainty will increase sharply.

[0107] The segment image is to describe the dimension information of different types of segments and extract the corresponding characteristics. The segment image reflects the weight proportion of the corresponding characteristics of these different types of segments. In this mode, each type of segment has its own segment image and weight scheme. The invention needs to extract the characteristics of the highway network information, identify various segment types, and extract representative segment images according to the characteristics of each segment to evaluate the authenticity of the congestion information, so as to determine the segment congestion level. The following is the judgment index of the segment image, where L QD is the judgment index of the segment image, and the following variable index ρ is:

[0108] L QD ={ρ1,ρ2,ρ3,Level} (13)

[0109]

[0110]

[0111]

[0112] The above formula describes the data index of the highway influencing factors, ρ1 captures the uncertainty of the congestion information evaluation caused by the highway toll station, ρ2 captures the uncertainty of the congestion information evaluation caused by the highway on-ramp, and ρ3 captures the uncertainty of the congestion information evaluation caused by the highway service area. Level is the joint effect coefficient of the segment image, where SND, VND and YND are three uncertainty factors, and Level can be directly expressed as the number of uncertainty factors.

[0113] Table 3-3 Level explanation table

[0114] Level Meaning 0 No uncertain factor 1 One uncertain factor, such as containing SND or VND or YND 2 Two uncertain factors, containing two cases of uncertain factors 3 Three uncertain factors, containing all cases of uncertain factors

[0115] Fuzzy comprehensive evaluation: the congestion state in the section is subjective, everyone's cognition of the traffic state is fuzzy, and the fuzzy comprehensive evaluation is a comprehensive evaluation method based on fuzzy mathematics, Hua Liu

[28] and S Y Hao[ 29 ] all apply the fuzzy comprehensive evaluation combined with multiple factors to the related application scenarios for research. According to the membership degree theory of fuzzy mathematics, the comprehensive evaluation method converts qualitative evaluation into quantitative evaluation, that is, the fuzzy mathematics is used to make a general evaluation on a thing or object restricted by multiple factors, and the traffic condition in the section can be evaluated clearly and systematically.

[0116] Firstly, the traffic state in the section is defined as a fuzzy set μ A : x→[0, 1], and the vector representation method is used to represent the membership degree. Further, the membership function is determined, and the assignment method is used to determine the membership function. According to the objective law and general needs of the section average speed, section traffic flow and section delay time on the expressway, a large type of membership function is selected to assign the section average speed index, and a small type of membership function is used to assign the section delay time.

[0117] Small type:

[0118]

[0119] Large type:

[0120]

[0121] Through the relationship between the traffic index and the congestion evaluation grade, the fuzzy matrix R is constructed.

[0122]

[0123] The weight A obtained by the analytic hierarchy process is [a1, a2, a3], and the fuzzy matrix is comprehensively evaluated.

[0124] B=A*R (20) According to the maximum membership degree principle, the maximum score in the B matrix is selected as the congestion score of the section at this time, which is called C in the application.

[0125] In the present application, the congestion score C of the highway section identified by fuzzy comprehensive evaluation, the output of C is defined as a continuous value [0, 1] by the present application, according to the Technical Specification for Highway Travel Information Service, the section traffic speed is taken as the main index to evaluate the highway traffic congestion, and according to the threshold value of the speed, C is classified, and the section traffic flow Flow and section delay time Delay are taken as auxiliary indexes, so as to judge the congestion state. That is, if C is greater than or equal to 0.35, the present application considers that no congestion occurs in the section, otherwise, it is considered that congestion has occurred in the section.

[0126]

[0127]

[0128] The environment of the present application adopts python 2.75 as the programming language, and Linux 7 as the environment system. The 351 toll stations of the expressway in Fujian Province have realized full coverage of ETC lanes, and 1021 sets of main line ETC gantry systems have been built to cover all expressway toll sections, and the user side ETC issuance has broken through 5.5 million vehicles, accounting for 85% of the total vehicle population in the province. The large-scale deployment of the ETC system records the passing state of most vehicles on the expressway, in order to verify the congestion identification accuracy of the method of the present application, the present application takes the ETC system distribution of the expressway in 8 prefecture-level cities and 13 county-level cities in Fujian Province as the research area, and the spatial positions of the gantries and toll stations are as shown in the figure.

[0129] Experimental data introduction: the ETC leader gantry system is an important part of the expressway ETC system, which generates intelligent transaction data that can describe the specific situation on the expressway in detail, realize congestion identification, gantry detection, path identification and other functions. There are three types of experimental data. One is the intelligent transaction data generated by each ETC gantry of the expressway in Fujian Province from May 1, 2021 to May 3, 2021 to May 5, 2021, as shown in Table 4-1, a total of 42809819 transaction data, of which 2599805 trajectory data and 17049087 section data are fitted, the other data is the topological relationship diagram between the expressway gantries and the distance between them. The third kind of data is the latitude and longitude information of the gantry and the toll station. These data come from Fujian Expressway Information Technology Co., Ltd.

[0130] Table 4-1 ETC transaction data

[0131] Name Types Examples PassID String 01350119382305xxxxxxxx0501163019 Obuplate String Blue xxx9742 Entime DateTime 2021-05-01xx:xx:xx Enstation FixedString(4) 35xx flagid String 34xx01 tradetime DateTime 2021-05-01xx:xx:xx

[0132] Segment dataset construction: matching the transaction data generated by the ETC gantry and the topological data, the traffic flow pattern of the vehicle trajectory data formed has different local space-time characteristics, and the congestion information therein cannot be well evaluated. In order to more effectively evaluate the congestion information of the highway segment, the present application selects 2599805 pieces of trajectory data for abnormal data screening, eliminates 139197 pieces of trajectory data that cannot be matched due to the absence of topological data, and uses Chebyshev theorem combined with SO QD Cutting the trajectory data, as shown in Figure 5 SO QD Mostly below 6773, in order to better evaluate the congestion information of the highway segment, the present application focuses on evaluating the congestion information of the few segments with high SO QD times, and constructs 12292350 pieces of segment dataset.

[0133] In order to maximize the extraction of congestion features in the segment dataset, the present application performs congestion data screening on the segment dataset and designs a weight scheme for segments with external factors, and performs congestion evaluation combined with fuzzy comprehensive evaluation, wherein 2769 kinds of topological data and segment dataset are matched to construct segment labels. Through experimental verification, there are 8 types of segment labels, of which the most common segment type is the split and merge segment, followed by the split and merge and service area coexistence segment, with 8225424 and 2884151 respectively. The segment type on the highway has an indispensable influence on the segment dimension information. In order to design the weight scheme, the dimension of each type of segment needs to be analyzed. The present application selects the service area segment with the greatest impact on the dimension information for analysis, Figure 6 Taking the first 200 data of the service area.

[0134] Segment feature selection: effectively extracting highway segment features is conducive to accurate evaluation of congestion information, and the segment average speed can well reflect the traffic running state of the segment. The present application takes the segment average speed as the main evaluation index in the congestion recognition method, and takes the segment flow and the segment average delay time as the correction features, and combines the relevant standards of the service level classification of the highway to recognize the congestion of the segment.

[0135] The general trend of analyzing the characteristics of the highway segment is shown in Figure 7 The statistical distribution of the characteristic dimension of 12292350 pieces of segment data from May 3, 2021 to May 5, 2021 is shown. Figure 7 All the statistical curves are generated from the segment dataset, which reveals the objective law of vehicle travel on the highway and explains the reason for selecting appropriate parameters for the model.

[0136] Effectiveness of section image: there are differences in the evaluation of congestion information of different section types. In order to solve various uncertainties between different section types, the application proposes a section image mainly for three types of sections, such as service area, toll station, split and merge section and jointly affected section. Table 4-2 is the definition of the section image.

[0137] Table 4-2 Section image

[0138] Service area Toll station Split and merge FD × × × SD × √ × VD √ × × YD × × √ SVD √ √ × SYD × √ √ VYD √ × √ SVYD √ √ √

[0139] The number of each section type is shown in Table 4-2. Figure 8 Through experiments, it is verified that the speed of the service area section will decrease greatly at a certain moment, the average speed of the section is greatly affected, and the traffic flow of the section is easy to cause congestion. The existence of ramp in the toll station section causes the speed of the vehicle to be limited, which causes errors in the evaluation of congestion information. Therefore, the section image is proposed to avoid the differences caused by different influencing factors between different sections.

[0140] The application identifies congestion of 12292350 sections of the expressway. Through consistency test of the analytic hierarchy process, the application constructs a congestion identification data set based on different weight schemes of different section types with a time interval of 15 minutes. Among 1804 sections, 182 possible congestion sections are selected, and 6 sections with obvious congestion characteristics are selected, as shown in Table 4-3. Figure 9

[0141] In some sections, a small part of vehicles travels at a slow speed, which is detected as congestion by the previous congestion identification algorithm, and the "low flow and low speed" and "high flow and high speed" phenomena occur. Through comparison of data on May 1, 2021, it can be seen that the congestion identification method proposed by the application can avoid such problems. Figure 9 The left side of the figure in Table 4-4 is the result of fuzzy comprehensive evaluation identification, and the right side is the result of the method of the application. It can be seen that the congestion state curve of Bailong Hub to Minhou Cane section (340D19-340D1B) starts to fluctuate around 1 am, which is often caused by a small number of vehicles driving at low speed at night to ensure safety. In the Minhou Sand Dike to Jingxi Hub section (340D19-350133), the identification result of the fuzzy comprehensive evaluation in the congestion period 10:00-10:40 fluctuates between slight congestion and basic stability, while the result of the method proposed by the application is stable in moderate congestion. This is because the Minhou Sand Dike to Jingxi Hub section is a split and merge section, and the proportion of section traffic flow needs to be larger than that of general section traffic flow, so that the evaluation of congestion information is wrong, which proves the effectiveness of the section image.

[0142] ​Method test result verification: the present application regards congestion levels above moderate congestion as congestion, and uses ETC transaction data provided by Fujian Provincial Highway Information Technology Co., Ltd. on May 1, 2021 and May 3, 2021 to May 5, 2021 as a verification set to verify, and the video data Figure 10 Verification, the segment congestion recognition is basically accurate.

[0143] The present application proposes a highway congestion recognition method based on ETC intelligent transaction data. First, a segment data set is constructed based on ETC trajectory data. After analyzing the traffic characteristics of the highway, three data dimensions are constructed, and the data is weighted using the analytic hierarchy process. Finally, the segment label and fuzzy evaluation are combined to evaluate the score, ensuring the integrity and accuracy of the congestion segment.

[0144] The experimental results show that: the combination of the analytic hierarchy process and the segment label improves the fuzzy comprehensive evaluation, which reasonably evaluates the fuzzy situation in the segment, effectively improves the recognition of the situation in the congestion segment and reduces the occurrence of congestion misjudgment, and can evaluate the congestion information combined with the characteristics of different segments. However, there are still some problems that have not been solved:

[0145] (1) This work will be affected by some special situations of the highway, such as weather factors, road maintenance factors and traffic accidents, etc. More data features can be considered for extraction and modeling to improve the recognition effect.

[0146] (2) Because the speed limit of each segment of the highway is different, there is a certain difference between the average speed of each segment, which has a certain influence on the experimental results. For the subsequent experiment, the speed limit condition of each segment can be compared with the speed, and the comparison difference can be used as a reference to measure the congestion degree of the segment.

[0147] Obviously, the embodiments described are part of the embodiments of the present application, rather than all the embodiments. The embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of the present application is not intended to limit the scope of the claimed application, but only represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0148] Reference:

[0149] [1] The Office of the Ministry of Transport, on further improving the truck ETC issuance service related work notice. https: / / xxgk.mot.gov.cn / 2020 / jigou / glj / 202111 / t20211115_3626254.html

[0150] [2] The Office of the Ministry of Transport, on strengthening the construction of national highway inter-provincial bottleneck section notice https: / / xxgk.mot.gov.cn / 2020 / jigou / zhghs / 202201 / t20220124_3638100.html

[0151] [3] Yue Tu Shukuan, Lin Jianzhong, Qiao Bin Liu. Deep traffic congestion prediction model based on road segment grouping. Applied Intelligence (2021).

[0152] [4] Xiangjie Kone, Zhenzhen Xu, Guojiang Shen, Jinzhong Wang, Qiuyuan Yang, Benshi Zhang. Urban traffic congestion estimation and prediction based on floating car trajectory data. Future Generation Computer Systems. 2015.

[0153] [5] Ameni Chetouane, Sabra Mabrouk, Imen Jemili, Mohamed Mosbah. Vision-based vehicle detection for road traffic congestion classification. First published: 09 September 2020. Concurrency and Computation Practice and Experience.

[0154] [6] Fan Zhang, Shuyun Niu, Jian Gao, and Marianthi Mermygka. A Study on Highway Corridor Traffic State Identification Based on Data Fusion of DSRC Data and ETC Data. CICTP 2017.

[0155] [7] W. Pattara-atikom, P. Pongpaibool, and S. Thajchayapong. Estimating Road Traffic Congestion using Vehicle Velocity. 2006 6th International Conference on ITS Telecommunications Proceedings.

[0156] [8] Fatemeh Ahanin, Norwati Mustapha, Nasir Bin Sulaiman, Maslina Zolkepli. An efficient traffic state estimation model based of fuzzy c-mean clustering and mdl using FCE. Journal of Theoretical and Applied Information Technology. 31st July 2020. Vol. 98. No 14.

[0157] [9] Wei-Hua Lin, Qingying Lu, and Joy Dahlgren. Dynamic Procedure for Short-Term Prediction of Traffic Conditions. Transportation Research Record 1783.

[0158]

[10] Trinh Dinh Toana, Y. D. Wong. Fuzzy logic-based methodology for quantification of traffic congestion. Physica A. 2021.

[0159]

[11] Yue Tu, Shukuan Lin, Jianzhong Qiao, Bin Liu. Deep traffic congestion prediction model based on road segment grouping. Applied Intelligence (2021).

[0160]

[12] Hong Gao, Zhenjun Yan, Xu Hu, Zhaoyuan Yu, Wen Luo, Linwang Yuan, and Jiyi Zhang. A Method for Exploring and Analyzing Spatiotemporal Patterns of Traffic Congestion in Expressway Networks Based on Origin-Destination Data. Internation Journal of GeoInformation. 2021.

[0161]

[13] Wan-Xiang Wang, Rui-Jun Guo and Jing Yu. Research on road traffic congestion index based on comprehensive parameters: Taking Dalian city as an example. Advances in Mechanical Engineering 2018.

[0162]

[14] Maja Kalinic, Jukka M. Krisp. Fuzzy inference approach in traffic congestion detection. Annals of GIS. Volume 25, 2019-Issue 4.

[0163]

[15] Khaliun Nasanjargal, Jing Lu. Application of Analytic Hierarchy Process-Fuzzy Comprehensive Evaluation in Public Transport of Ulaanbaatar City, Mongolia. ICMLT 2020.

[0164]

[16] Zhudandan. Research on the Recognition Method of Inter-city Road Congestion State Based on Fuzzy Comprehensive Evaluation.

[0165]

[17] Xu Sun, Kun Lin, Pengpeng Jiao and Huapu Lu. The Dunamical Decision Model of Intersection Congestion Based on Risk Identification. Road Traffic Engineering and Sustainable Transportation. 2020.

[0166]

[18] Hai-Min Lyu, Wan-Huan Zhou, Shui-Long Shen, An-Nan Zhou. Foundation risk assessment of metro system using AHP and TFN-AHP in Shenzhen. Sustainable Cities and Society. Volume 56, May 2020, 102103.

[0167]

[19] Jianfeng Xi, Zhonghao Zhao, Wei Li, Quan Wang. A Traffic Accident Causation Analysis Method Based on AHP-Apriori. Procedia Engineering. 2016.

[0168]

[20] Gao Ningbo, Wang Wei, Huang Ying. Road Traffic Congestion Information Recognition Model Based on RFID. Transportation Technology and Economy. 2013.

[0169]

[21] Pang Haibo. Traffic Congestion Detection Based on Fuzzy Analytic Hierarchy Process. 2021. Dalian University of Technology.

[0170]

[22] Yuan Xin, Bu Henan, Niu Jianmin, Yu Wenjuan, Zhou Honggen, Ji Xingyu, Ye Pengfei. Coating matching recommendation based on improved fuzzy comprehensive evaluation and collaborative filtering algorithm. Scientific reports. 2021.

[0171]

[23] Shen Zhang, Hua Wang, Wei Quan and Xin Liu. Measuring Variability in Freeway Traffic States using Real-time Loop Data in Jilin. 13th COTA international Conference of Transportation Professionals. CICTP 2013.

[0172]

[24] Feifei He, Xuedong Yan, Yang Liu, Lu Ma. A Traffic Congestion Assessment Method for Urban Road Networks Based on Speed Performance Index. Procedia Engineering. 2016.

[0173]

[25] Han Long, Dong Kaifan, Feng Chunfang. Expressway traffic safety evaluation based on Fuzzy-AHP. China public safety. 2019.

[0174]

[26] Jiandong Zhao, Zhixin Yu, Xin Yang, Ziyou Gao, Wenhui Liu. Short term traffic flow prediction of expressway service area based on STL-OMS. Physica A 595 (2022) 126937.

[0175]

[27] Zhu Chen, Bangyu Wu, Bin Li, Housong Ruan. Expressway Exit Traffic Flow Prediction for ETC and MTC charging System Based on Entry Traffic Flows and LSTM Model.

[0176]

[28] Hua Liu, Bin Li, Yiqi Sun, Xiaozheng Dou, Yongfu Zhang and Xianghui Fan. Safety Evaluation of Large-size Transportation Bridges Based on Combination Weighting Fuzzy Comprehensive Evaluation. CEAEE 2021.

[0177]

[29] S Y Hao, N L Zhao, Y Z Wu, J H Li, K M Wu and C W Jiao. Research on Traffic Condition Discriminant Technology of Urban Intersections Under Intelligent Lighting System Based on Fuzzy Comprehensive Evaluation. CEAEE 2021.

Claims

1. A method for identifying highway congestion based on ETC transaction data, characterized in that: It includes the following steps: Step 1: Obtain highway ETC transaction data, topology data, and highway vehicle trajectory data. Step 2: Based on the spatiotemporal information of ETC transaction data, match highway ETC transaction data with topological data to construct highway vehicle trajectory data and segment datasets. Step 3: Clean the highway vehicle trajectory data to remove unwanted data; Step 4: The cleaned ETC transaction data is used to construct a vehicle trajectory set TrajS in chronological order. The vehicle trajectory set refers to the set of multiple ETC gantries that all vehicles pass through during their journey on the highway. Step 5: Congestion feature matching was performed between the topology data and the segment dataset to construct segment labels; Step 6: Construct and calculate three segment-dimensional parameters of traffic congestion based on the objective laws of highways. The three segment-dimensional parameters are the average speed of the segment, the flow rate of the segment, and the average delay time of the segment. Step 7: Construct a judgment matrix X for the three segment dimension parameters, calculate the eigenvalues ​​and eigenvectors of the judgment matrix X respectively, and obtain the final weights after passing the consistency judgment test of the discriminant matrix. In step 7, the analytic hierarchy process (AHP) is used to analyze and weight these three dimensions. The importance of a certain element at the same level with respect to a certain criterion at the previous level is compared pairwise, and a pairwise comparison matrix is ​​constructed. The judgment matrix X is obtained by constructing a matrix for all dimensions. (12) in, ~ This indicates the degree of importance between pairs of elements, used to construct a judgment matrix; Step 8: Analyze the segment profiles in the highway to obtain the relationship between different types of segment dimension parameters, and assign the corresponding final weights. Step 9: Construct a fuzzy matrix R based on the relationship between traffic indicators and congestion assessment levels; (19) in, , The fuzzy matrices correspond to the three segment dimensional parameters: segment average speed, segment flow rate, and segment average delay time, respectively. ~ express The corresponding position value; ~ express The corresponding position value; ~ express The corresponding position value; Step 10: Apply the final weights obtained by the analytic hierarchy process. The fuzzy matrix R is used to make a comprehensive evaluation to obtain matrix B. According to the principle of maximizing membership, the maximum value score in matrix B is selected as the congestion score C of the segment at this time. (20) in, The weight values ​​for the three segment dimension parameters, namely, segment average speed, segment flow, and segment average delay time, are respectively. Step 11: Determine whether the congestion score C is greater than the set threshold; if yes, determine that there is no congestion in the current segment; otherwise, determine that there is congestion in the current segment.

2. The highway congestion identification method based on ETC transaction data according to claim 1, characterized in that: In step 3, Chebyshev's theorem is used to assess the superposition degree of segments. Perform data segmentation, removing segments with low overlap from TrajS, and retaining... Congestion information is assessed for sections with high frequency of traffic congestion.

3. The highway congestion identification method based on ETC transaction data according to claim 2, characterized in that: Segment overlap It indicates the number of times each section QD is superimposed on the highway within a certain time period; section QD represents the combination of adjacent boundary points DND on the highway to form a section; boundary point DND includes gantry FND and toll station SND.

4. The highway congestion identification method based on ETC transaction data according to claim 1, characterized in that: The criteria for determining the segment profile in step 8 are as follows: (13) (14) (15) (16) in, Indicators for profiling a segment. The uncertainty in assessing congestion information caused by highway toll booths. The uncertainty in assessing congestion information caused by merging and diverging areas on highways. The uncertainty in assessing congestion information caused by highway service areas; Level indicates the number of uncertain factors. This indicates the section QD containing the toll plaza SND. This indicates the segment QD containing the service area VND. This indicates the section QD containing the merging and diverging point YND.

5. The highway congestion identification method based on ETC transaction data according to claim 4, characterized in that: Adjacent DNDs on a highway are combined to form a segment QD. The segment type is determined by the uncertainty factor UF. When the UF in QD is SND, the segment is called SD; if the UF is absent, it is called a normal segment FD. (1) (2) (3) Where DND represents the boundary point, The first boundary point, This is the second boundary point; UF represents the uncertainty factor. When UF in QD is VND, this segment is called VD; (4) When UF in QD is YND, the segment is called YD; (5) SND represents toll station, VND represents service area, and YND represents merging / diverting point.

Images