A Low-Speed Vehicle Hierarchical Warning Method for Highway Scenarios

Abstract

The present invention discloses a method for classifying and warning low-speed vehicles in a highway scenario. The method is as follows: collect data from each ETC gantry and fuse them to obtain a dataset to be processed; identify the passing speed characteristics of different vehicle types in different sections according to the driving speeds of vehicles under different conditions to obtain an interval passing speed characteristic dataset; calculate the time difference when the vehicle passes through the service area and update the passing duration data in the trajectory data; calculate the passing speed of the vehicle and compare it with the road passing speed under the same conditions to determine whether it is a low-speed vehicle; construct a smaller membership function to calculate the membership of the current vehicle; calculate the weight vector of the factors affecting driving behavior safety based on the proportional scale table of the factors affecting driving behavior safety; calculate the driving behavior score of the vehicle section based on the membership of the vehicle and the weight vector of the factors affecting driving behavior safety, so as to divide different warning levels. The present invention accurately identifies low-speed vehicles and classifies and warns them.

Description

Technical Field

[0001] The present invention relates to the technical field of highway management, and particularly to a method for classifying and warning low-speed vehicles in highway scenarios. Background Art

[0002] The existence of low-speed vehicles on highways seriously affects the overall traffic efficiency. Especially during peak traffic hours, this impact is more significant, often leading to large-scale traffic congestion. More seriously, the composition of low-speed vehicles is extremely likely to cause vicious traffic accidents such as consecutive rear-end collisions. To effectively solve this problem, it is particularly crucial to establish a fast and accurate low-speed vehicle identification system. The establishment of this system can provide data support for the refined management of the road network, significantly improve the response speed of emergency management departments, provide a basis for formulating scientific emergency strategies, and thus better protect the lives and property safety of drivers and passengers.

[0003] Existing low-speed vehicle identification methods mostly rely on technologies such as video surveillance, manual inspections, and vehicle-mounted positioning. However, the application of these methods faces a series of problems:

[0004] 1. Manual inspections are inefficient and time-consuming: Manual inspections require a large amount of manpower and material resources, and are limited by the experience and attention of inspectors, making it difficult to achieve real-time monitoring throughout the day and across the entire road section. On long-distance highways, the round-trip time of inspection vehicles is relatively long, making it difficult to detect and handle low-speed vehicles in a timely manner. In addition, under adverse weather conditions, the efficiency and safety of manual inspections will be significantly reduced.

[0005] 2. Video surveillance has blind spots: Existing video surveillance systems are limited by the installation position and angle of cameras, making it difficult to cover all lanes and road sections. In special sections such as curves and slopes, the monitoring effect is often not ideal. At the same time, the recognition accuracy of video surveillance significantly decreases at night or in rainy and foggy weather. The monitoring blind spots between multiple cameras also pose difficulties for the continuous tracking of low-speed vehicles.

[0006] 3. The coverage of vehicle-mounted positioning is low: Relying on vehicle-mounted positioning systems for low-speed vehicle identification has obvious limitations. First of all, not all vehicles are equipped with positioning devices, especially some old vehicles and freight vehicles. Secondly, positioning signals are easily interfered with in special sections such as tunnels and mountainous areas, resulting in data loss or deviation. In addition, the data standards of positioning devices from different manufacturers are not unified, which also brings difficulties to data integration.

[0007] 4. False alarms are caused by vehicles entering and leaving service areas: When a vehicle normally enters a service area, its speed change is easily misjudged by the system as low-speed driving. Existing recognition algorithms often have difficulty distinguishing between vehicles that normally enter service areas and truly low-speed vehicles, resulting in a relatively high false alarm rate. Summary of the Invention

[0008] The object of the present invention is to provide a method for classifying and warning low-speed vehicles in the expressway scenario. Based on the transaction data collected by expressway ETC gantries and integrating multi-source data such as capture records in service areas, a statistical algorithm evaluation index, etc. is used to construct a low-speed vehicle classification and warning algorithm to realize the classification and warning of low-speed vehicles, and provide efficient and accurate warning information of low-speed vehicles on the expressway for traffic management departments and emergency rescue departments.

[0009] The technical solution adopted by the present invention is as follows:

[0010] A method for classifying and warning low-speed vehicles in the expressway scenario, which includes the following steps:

[0011] Step 1, data collection and preprocessing: Real-time collect data of each ETC gantry and perform fusion processing to obtain a dataset to be processed. The collected data includes ETC system transaction data, ETC service area data, and expressway section QD.

[0012] Step 2, mining the passing speed characteristics of the expressway section: Identify the passing speed characteristics of the section according to the driving speeds of vehicles of different types on the expressway under different section conditions, and mine the passing speed characteristics of the vehicles to obtain an interval passing speed characteristic dataset.

[0013] Step 3, processing abnormal data in the service area: Obtain service area data and vehicle trajectory data , calculate the time difference when the vehicle passes through the service area and update the passing duration data in the vehicle trajectory data ; Eliminate the interference of the passing time in the service area on the time of the vehicle trajectory data .

[0014] Step 4, discrimination of low-speed vehicles: Calculate the passing speed of the corresponding vehicle based on the vehicle trajectory data , and compare the passing speed of the corresponding vehicle with the road passing speed under the same conditions in the interval passing speed characteristic dataset to determine whether the current vehicle belongs to a low-speed vehicle.

[0015] Step 5, classification scoring and warning of low-speed vehicles: Construct a partial small membership function as the membership function of low-speed vehicles to calculate the membership of the current vehicle; at the same time, calculate the weight vector of the factors affecting driving behavior safety based on the proportional scale table of the mutual ratio of factors affecting driving behavior safety; calculate the driving behavior score of the current vehicle section based on the membership of the current vehicle and the weight vector of the factors affecting driving behavior safety, so as to divide different warning levels based on the driving behavior score of the section.

[0016] Further, step 2 specifically includes the following steps:

[0017] Step 2-1: The data set is classified by hour to obtain the overall road section speed in different time periods.

[0018] Specifically, the data set can be divided into 24 categories accordingly.

[0019] Step 2-2: On the basis of the hour classification, the data is further refined according to the vehicle type vehclass, which can effectively solve the problem of different passing speeds of different vehicle types.

[0020] Step 2-3: According to the spatial regularity, the data set is classified according to the vehicle trajectory data to obtain different section passing data sets, so as to deeply explore the spatial regularity of vehicles.

[0021] Specifically, due to external conditions such as the number of lanes and road conditions, the passing speeds in different sections of highway vehicle driving behaviors are also different. Therefore, according to the spatial regularity, the data set is classified according to the vehicle trajectory data into different section passing data sets, so as to deeply explore the spatial regularity of vehicles.

[0022] Step 2-4: Extract the passing time of vehicles in the target section through the divided section passing data sets; at the same time, extract the passing distance of the target section through the gantry distance matrix D, and use the non-intrusive speed measurement model to calculate the driving speeds of all vehicles in the data set.

[0023] Specifically, the element in the gantry distance matrix D is the distance between gantry i and gantry j. If they are connected, dis is the distance between the two gantries. If they are the same node, it is 0, and the distance between unconnected nodes is inf; the element in the gantry distance matrix D has the following expression:

[0024] (1);

[0025] Step 2-5: Delete the outlier abnormal data outside the upper and lower edges based on the data cleaning model of the box plot.

[0026] Step 2-6: Calculate the average value of the driving speeds of all vehicles as the passing speed of the road , so as to more accurately and effectively approach the true value; the calculation formula of the passing speed of the road is as follows:

[0027] ;

[0028] where represents the vehicles passing through this section , represents the total number of vehicles on this section of the road.

[0029] Step 2-7: Add the traffic speed of the road to the highway section traffic speed feature dataset To further eliminate the impact of vehicles entering the service area, through the following capture technology and ETC data fusion technology, the impact of the service area is eliminated.

[0030] Furthermore, Step 4 specifically includes the following steps:

[0031] Step 4-1: Extract vehicle trajectory data and transaction data, and calculate the section traffic speed of the vehicle ;

[0032] Step 4-2: Compare the section traffic speed of the vehicle with the traffic speed of the road under the corresponding conditions in the section traffic speed feature dataset to determine whether there is a low-speed driving behavior; if so, record the traffic information of this low-speed driving and execute Step 4-3; otherwise, execute Step 4-3; Specifically, in Step 4-2, the average value of the driving speeds of all vehicles is calculated as the traffic speed of the road

[0033] ; ;

[0034] Step 4-3: Determine whether all vehicle trajectories have been traversed; if so, output the recording result; otherwise, execute Step 4-1.

[0035] Furthermore, Step 5 specifically includes the following steps:

[0036] Step 5-1: Construct a sub-minimum membership function as the membership function of low-speed vehicles. The expression of the membership function of low-speed vehicles is as follows:

[0037]

[0038] where represents the vehicle traffic speed; is the traffic speed of the road in the highway section traffic speed feature dataset that is the same as the current situation;

[0039] Step 5-2: Calculate and obtain the low-speed membership of the current vehicle ;

[0040] Step 5-3: Based on the proportional scale table of the safety factors affecting driving behavior, construct a judgment matrix R, where each element value in the judgment matrix R represents the proportional scale between different safety factors affecting driving behavior.

[0041] Step 5-4: Normalize the column vectors of the judgment matrix R to obtain the normalized judgment matrix ; Calculate the mean value of the row vectors of the normalized judgment matrix to obtain the row vector , which serves as the weight vector of the driving behavior safety factors in the corresponding row.

[0042] Step 5-5: Calculate the section driving behavior score of the current vehicle according to the weights of different driving behavior safety factors; specifically, the expression of the section driving behavior score is as follows:

[0043] ;

[0044] where represents the section driving behavior score of the vehicle; A is the set of vehicles with dangerous driving behaviors occurring within the section, is the membership degree of the driving behavior of vehicle a, is the driving behavior weight of vehicle a.

[0045] Step 5-6: Divide the vehicle itinerary into different warning levels according to the section driving behavior score of the vehicle.

[0046] Furthermore, the vehicle speed classification standard corresponding to the membership function of low-speed vehicles in Step 5 is as follows: In the first section, when the driving speed of a motor vehicle is greater than 80% of the road passing level, it is considered to be in the normal level, and the membership degree is 0; in the second section, when the driving speed of a motor vehicle is less than or equal to 80% of the road passing speed and greater than 60% of the road passing speed, calculate the membership degree of this driving behavior; in the third section, when the driving speed of a motor vehicle is less than or equal to 60% of the road passing speed and greater than 40% of the road passing speed, calculate the membership degree of this driving behavior; in the fourth section, when the passing speed of the vehicle is less than or equal to 40% of the road passing speed, the membership degree is 1.

[0047] The present invention adopts the above technical solutions. Through intelligent data processing and algorithms, combined with ETC gantry data, road network topology, and service area capture data, etc., it can judge stranded vehicles in real time and accurately identify them. According to different factors such as vehicle types and road sections, a flexible low-speed time threshold is designed to ensure accurate identification of low-speed vehicles in different road sections. After the stranded vehicles are identified, the system automatically generates warning information and sends alerts to traffic management departments and emergency rescue personnel in a timely manner, providing support for precise rescue and safety guarantee.

[0048] Compared with the prior art, the present invention has the following advantages and beneficial effects: 1) Using service area information to avoid a large number of false alarms: By fusing the capture records in the service area with the ETC transaction records, the filtering of low-speed vehicles misidentified when entering the service area is realized, solving the problem of misidentification of low-speed vehicles caused by entering the service area. 2) Low computational complexity and real-time computability: Compared with deep learning algorithms such as neural networks, it has the advantages of a small number of parameters and low computational complexity. In addition, due to its extremely low computational complexity, it can be applied to the real-time early warning of low-speed vehicles. 3) Strong interpretability: Compared with deep learning methods, this statistical method has strong interpretability and can provide a clearer theoretical basis for management personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The present invention will be further described in detail below in conjunction with the drawings and specific embodiments;

[0050] Figure 1 It is a schematic flow diagram for identifying the passing speed characteristics of the sections of the present invention;

[0051] Figure 2 It is a schematic flow diagram for identifying low-speed driving behaviors of the present invention;

[0052] Figure 3 It is a schematic diagram of a relatively small S function;

[0053] Figure 4 It is a schematic diagram of the top 10 low-speed risk behavior vehicles. SPECIFIC EMBODIMENTS

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0055] As Figures 1 to 4 shown in one of them, the present invention discloses a method for grading and early warning of low-speed vehicles in a highway scenario, specifically including the following steps:

[0056] Step 1) Data collection and preprocessing: Using big data technology to collect and fuse the data of each ETC gantry in real time, including the following data:

[0057] ETC system transaction data: The vehicle passing records collected by the ETC toll collection system on the highway, including the vehicle entry and exit toll station information. It can be defined as vehicle trajectory data : The gantry trajectory formed by the vehicle passing through section QD is called and the definition expression is as follows:

[0058] ;

[0059] Where Representing vehicle trajectory data is the starting point, representing vehicle trajectory data is the ending point.

[0060] ETC service area data: The capture discipline of vehicles entering and leaving the service area, including vehicle license plate information, the time when the vehicle enters and leaves the service area, the longitude and latitude of the service area, and the name of the service area.

[0061] ;

[0062] ;

[0063] Among them, represents the entire service area dataset, represents the capture data of service area s, is the minimum capture unit data of service area s, represents the number of unit data captured; represents the license plate; is the trip id of the vehicle; represents the time of entering the service area; represents the time of leaving the service area;

[0064] Highway section QD: All the gantries and toll station entrances and exits (including cross-provincial entrances and exits) on the highway are collectively called nodes Node. Two adjacent nodes Node form a highway section QD, simply referred to as a section, and the representation form is as follows

[0065] ;

[0066] Among them, QD is the highway section, Node1 is the starting point of the section, and Node2 is the ending point of the section.

[0067] Step 2) Mining the passing speed characteristics of highway sections: According to the driving speeds of vehicles of different types on different sections of the highway under different conditions, an algorithm for identifying passing speed characteristics of sections is established to mine the passing speed characteristics of the vehicles therein, and a dataset of passing speed characteristics for intervals is obtained.

[0068] Due to the spatial regularity and time-variability of driving behaviors, the danger levels of vehicles driving with the same driving behavior on different sections of the highway are different, and for different types of vehicles, the speed limits are also different, and the corresponding speed limit information is not constant. Therefore, according to the driving speeds of vehicles of different types on different sections of the highway under different conditions, the present invention establishes an algorithm for identifying passing speed characteristics of sections to mine the passing speed characteristics of the vehicles therein, and a dataset of passing speed characteristics for intervals is obtained, so as to more flexibly identify the driving behavior characteristics of vehicles. The specific process is as follows:

[0069] 1) Since highway travel is subject to time constraints, mainly including factors such as morning and evening rush hours and weather conditions, the dataset is classified by hour to more effectively extract time features. Thus, the data can be divided into 24 categories. The overall section speeds for different time periods.

[0070] 2) According to national regulations, the vehicles currently traveling on highways are divided into 18 categories, and the specific classification is shown in Table 1.

[0071] Table 1 Vehicle Classification Table

[0072]

[0073] Different vehicle types have different passing speeds on the same road. Therefore, on the basis of dividing the dataset by hour, the data is further divided according to the vehicle type vehclass. Through the refined division of vehicle types, the problem of different passing speeds for different vehicle types can be effectively solved.

[0074] 3) Due to external conditions such as the number of lanes and road conditions, the passing speeds in different sections of highway vehicle driving behavior are also different. Therefore, according to spatial regularity, the dataset is classified into different section passing datasets according to vehicle trajectory data , so as to deeply explore the spatial regularity of vehicles.

[0075] 4) According to the already divided dataset and the gantry distance matrix D, the passing time of vehicles in the target section is extracted from the dataset, and the passing distance of the target section is extracted from the distance matrix D. Using the non-intrusive speed measurement model, the driving speeds of all vehicles in the dataset are calculated.

[0076] The process of calculating the QD distance using the Amap API: First, extract the connected gantries in the adjacency matrix G ( , ), and use the gantry positioning method above to obtain the accurate gantry positions , . Input the gantry geographical locations into the Amap API to obtain the actual driving distance of the highway QD.

[0077] To effectively store and judge the actual driving distance of QD, the present invention constructs a gantry distance matrix D, where dis is the distance between the i-th gantry and the j-th gantry. If they are connected, dis is the distance between the two gantries; if they are the same node, it is 0; the distance between unconnected nodes is inf. The calculation formula is shown in Formula (1) below and is constructed into a matrix as shown in Table 2.

[0078] (1);

[0079] Table 2 Distance Matrix

[0080]

[0081] 5) To calculate the average speed of the road, the arithmetic mean method is used to effectively remove the extreme values in the dataset, so as to calculate the speed more accurately. The formula for calculating the average speed of different vehicle types is as follows.

[0082] ;

[0083] where is the average speed of vehicle p in vehicle type , is the total number of vehicles in vehicle type , is the vehicle type 's average speed;

[0084] Table 3 Average Driving Speeds of Different Vehicle Types

[0085]

[0086] For the estimation of road traffic speed, generally, the average speed of vehicles passing through this section within a certain time range is used as the estimated value. Therefore, reducing outliers can effectively improve the estimation accuracy. However, the method mentioned above can only effectively remove larger outliers and cannot effectively eliminate the situation of reducing vehicle driving speed such as staying in the service area.

[0087] Based on the box plot data cleaning model, the outlier abnormal data outside the upper and lower edges needs to be deleted, which can effectively remove the excessive or too small speed caused by unexpected situations or entering the service area. Then, the average value of the driving speeds of all vehicles is calculated using the following formula as the traffic speed of the road , so as to more accurately and effectively approach the true value.

[0088] ;

[0089] where represents the vehicles passing through this section , represents the total number of vehicles in this section.

[0090] Add the traffic speed of the road to the freeway section traffic speed feature dataset . To further eliminate the influence of vehicles entering the service area, the following capture technology and ETC data fusion technology are used to eliminate the influence of the service area.

[0091] Step 3) Processing of abnormal service area data: Obtain service area data and vehicle trajectory data , calculate the time difference when the vehicle passes through the service area and update the vehicle trajectory data in the passing duration data; eliminate the interference of the service area passing time on the vehicle trajectory data time;

[0092] Specifically, extract the capture records of the service area from the service area data , match the trajectory data of the vehicle passing through the service area , calculate the time difference when the vehicle enters and exits the service area; update the passing duration (excluding the residence time in the service area) of the section where the vehicle passes through the service area , and update the time difference to the trajectory passing duration data of the vehicle, avoiding interference with the vehicle speed calculation.

[0093] Step 4) Discrimination of low-speed vehicles: Low-speed driving on expressways has a great impact on road safety. Therefore, the present invention proposes an algorithm for identifying the characteristics of low-speed driving behavior on expressways

[0094] 1) Based on the vehicle trajectory data and the gantry distance matrix D established above, obtain the section driving distance and travel passing time time of the vehicle, and use the non-intrusive speed measurement model to calculate the section driving speed of the vehicle, and the calculation formula is as shown below.

[0095] ;

[0096] 2) Compare the passing speed of the vehicle with under the same traffic conditions in to determine whether the vehicle has speeding behavior. When , it is determined that there is low-speed driving behavior and the passing information of this time is recorded; on the contrary, when 0.8 , it is determined that the vehicle does not have low-speed driving behavior.

[0097] 3) Determine that the vehicle trajectory data has been traversed, and output the low-speed information of the vehicle.

[0098] Step 5) Classification, scoring and early warning of low-speed vehicles: The present invention uses the larger and smaller S-shaped membership functions as the quantization model of the driving behavior risk degree, and the function schematic diagram is as Figure 2 , and the smaller S-shaped membership function is used for low-speed driving.​

[0099] As shown by Figure 3 the membership function is divided into four parts. The first part is relatively smooth, the second part starts to drop steeply, the third part drops slowly, and the fourth part is parallel to the x-axis. The reasons for choosing this type of membership function are summarized as follows: when the speed is within a reasonable range, it basically does not belong to dangerous driving behavior, and the penalty factor is small at this time. As the speed continues to decrease, the penalty factor increases, the curve becomes steeper, and the function value continues to increase until it finally reaches 1. At the same time, this membership function quantifies the degree of danger of dangerous driving within the range of [0,1], thus effectively realizing the quantification of the risk level of driving behavior.

[0100] After the above analysis of the driving characteristics of highway vehicles, combined with the membership function selected above, the membership function is specifically determined, and the membership of the vehicle at low speed is calculated:

[0101] ;

[0102] where is the road traffic speed in the highway section traffic speed characteristic dataset that is the same as the current situation.

[0103] As shown in the above formula, for the membership function of low-speed driving behavior, the present invention divides it into four parts. Since the newly introduced traffic law does not classify low-speed driving and only uniformly processes vehicles below the specified speed limit, the present invention establishes a classification standard in accordance with the speeding rules. In the first part, when the driving speed of a motor vehicle is greater than 80% of the road traffic level, it is considered a normal level; in the second part, when the driving speed of a motor vehicle is less than or equal to 80% of the road traffic speed and greater than 60% of the road traffic speed, the membership of this driving behavior is calculated; in the third part, when the driving speed of a motor vehicle is less than or equal to 60% of the road traffic speed and greater than 40% of the road traffic speed, the membership of this driving behavior is calculated; in the fourth part, when the traffic speed of the vehicle is less than or equal to 40% of the road traffic speed, the membership is 1.

[0104] When constructing the judgment matrix, the importance degree of the compared indicators has a great influence on the values of the elements in the matrix for the influence level of driving safety.

[0105] Table 4 Ratio Scale Table

[0106]

[0107] According to the ratio scale table after stratifying the driving safety, comparing the importance between the bottom-layer elements, the present invention establishes a judgment matrix R = rij(n×n), as shown in Table 5:

[0108] Table 5 Judgment Matrix

[0109]

[0110] Among them, f1, f2, f3, …, fn represent various factors affecting driving behavior safety after classification, and the elements in the matrix The value of is determined by experts through the proportional scale table for factors and factor The corresponding importance scale is obtained from the importance between (for example, Factor is significantly more important than Factor, then The scale value of is 5, The scale value of is 1 / 5). This table classifies the importance scale of elements into 1-9. The larger the importance scale, the more important the element is relative to another element. In order to further determine the weight of the vehicle and facilitate calculation, the column vectors of matrix R are normalized to obtain the normalized matrix , and the results are shown in Table 6. Finally, calculate the mean value of the middle row vector of matrix to obtain the row vector , as shown in the following formula, as the weight vector of the elements.

[0111] Table 6 Normalized judgment matrix

[0112] Among them, Represents the value of the element in the first column of the second row of the judgment matrix, that is, Value in Table 5; Represents the judgment matrix Row Column element value, that is, Value in Table 5;

[0113] (2)

[0114] Among them, Represents the value of the element in the first column of the first row of the normalized judgment matrix; The value of the element in the first column of the Row of the normalized judgment matrix;

[0115] After determining the weights of different driving behaviors, it is necessary to calculate the driving safety of a vehicle's certain trip according to the calculated weights. The present invention proposes a section driving behavior scoring model, and the calculation formula is as shown in the following formula.

[0116] ;

[0117] Among them, ​Represents the section driving behavior score of the vehicle; A is the set of vehicles with dangerous driving behaviors occurring within the section, is the membership degree of the driving behavior of vehicle a, is the driving behavior weight of vehicle a in formula (2).

[0118] The present invention divides the low-speed journey into 4 categories, and the classification meanings are shown in Table 7 below.

[0119] Table 7 Vehicle Driving Behavior Definition Table

[0120] Identification Meaning 1 The current trip of the vehicle is a normal trip 2 The current trip of the vehicle is a mild low-speed trip 3 The current trip of the vehicle is a moderate low-speed risk trip 4 The current trip of the vehicle is a severe low-speed risk trip

[0121] In order to effectively verify the effectiveness and reliability of the model, the present invention uses the following formula to convert the scoring results above into 4 types in the table:

[0122] ;

[0123] Among them, x represents the score obtained from the above evaluation model, and C(x) represents the classification result.

[0124] Through different scoring results, hierarchical classification early warning of low-speed vehicles is achieved.

[0125] Selected some sections from September 2024 to December 2024 for algorithm effect evaluation. 124 vehicles were monitored and a total of 2,862 low-speed behaviors occurred. Among them, the top 10 vehicles with serious low-speed risks are as Figure 4 shown.

[0126] The present invention is particularly applicable to automatically identifying low-speed risk vehicles and giving safety early warnings in scenarios of bad weather (such as typhoons, heavy rains, etc.).

[0127] The present invention adopts the above technical solutions. Through intelligent data processing and algorithms, combined with ETC gantry data, road network topology structure, service area capture data, etc., it can judge the stranded vehicles in real time and accurately identify them. According to different vehicle types, road sections and other factors, a flexible low-speed time threshold is designed to ensure accurate identification of low-speed vehicles in different road sections. After the stranded vehicles are identified, the system automatically generates early warning information and sends alarms to traffic management departments and emergency rescue personnel in a timely manner, providing support for precise rescue and safety guarantee.

[0128] Compared with the prior art, the present invention has the following advantages and beneficial effects: 1) Avoiding a large number of false alarms by using service area information: By fusing the captured records in the service area with the ETC transaction records, the filtering of the low-speed vehicles misidentified when entering the service area is realized, and the problem of misidentifying low-speed vehicles caused by entering the service area is solved. 2) Low computational complexity and real-time computability: Compared with deep learning algorithms such as neural networks, it has the advantages of a small number of parameters and low computational complexity. In addition, due to its extremely low computational complexity, it is applicable to the real-time early warning of low-speed vehicles. 3) Strong interpretability: Compared with the deep learning method, this statistical method has strong interpretability and can provide a clearer theoretical basis for managers.

[0129] Obviously, the described embodiments are part of the embodiments of the present application, rather than all embodiments. Without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. Generally, the components of the embodiments of the present application described and illustrated in the accompanying drawings here 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 present application claimed, but merely represents the selected embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the scope of protection of the present application.

Claims

1. A method for classifying and warning low-speed vehicles in highway scenarios, characterized in that: It includes the following steps: Step 1, data collection and preprocessing: Real-time collect the data of each ETC gantry and perform fusion processing to obtain the dataset to be processed. The collected data includes ETC system transaction data, ETC service area data, and highway section QD; Step 2, mining the passing speed characteristics of highway sections: Identify the passing speed characteristics of sections according to the driving speeds of vehicles of different types on different sections of the highway, and mine the passing speed characteristics of vehicles to obtain the interval passing speed characteristic dataset; Step 3, processing abnormal data in the service area: Obtain service area data and vehicle trajectory data , calculate the time difference when the vehicle passes through the service area and update the vehicle trajectory data in the passing duration data; Eliminate the interference of the service area passing time on the vehicle trajectory data to the time; Step 4, low-speed vehicle discrimination: Based on vehicle trajectory data Calculate the corresponding vehicle passing speed , and compare the corresponding vehicle passing speed with the road passing speed under the same conditions in the interval passing speed feature dataset to determine whether the current vehicle is a low-speed vehicle; Step 5, grading and scoring warning for low-speed vehicles: Construct a smaller membership function as the membership function of low-speed vehicles to calculate the membership of the current vehicle; at the same time, calculate the weight vector of the factors affecting driving behavior safety based on the proportional scale table of the mutual influence of factors affecting driving behavior safety; Calculate the driving behavior score of the current vehicle section based on the membership of the current vehicle and the weight vector of the factors affecting driving behavior safety, so as to divide different warning levels based on the driving behavior score of the section. Step 5 specifically includes the following steps: Step 5-1, construct a smaller membership function as the membership function of low-speed vehicles. The expression of the membership function of low-speed vehicles is as follows: ； Among them, represents the vehicle passing speed; it is the road passing speed in the highway section passing speed characteristic data set that is the same as the current situation; Step 5-2, calculate and obtain the low-speed membership degree of the current vehicle ; Step 5-3, construct a judgment matrix R based on the proportional scale table of the mutual influence of factors affecting driving behavior safety. Each element value in the judgment matrix R represents the proportional scale between different factors affecting driving behavior safety; Step 5-4, normalize the column vectors of the judgment matrix R to obtain the normalized judgment matrix ; calculate the mean of the row vectors of the normalized judgment matrix to obtain the row vector , which serves as the weight vector of the driving behavior safety factor for the corresponding row; Step 5-5, calculate the driving behavior score of the current vehicle section according to the weights of different factors affecting driving behavior safety. Specifically, the expression of the driving behavior score of the section is as follows: ; Among them, represents the section driving behavior score of the vehicle; A is the set of vehicles with dangerous driving behaviors occurring within the section, is the membership degree of the driving behavior of vehicle a, is the driving behavior weight vector of vehicle a; Step 5-6, divide the vehicle itinerary into different warning levels according to the driving behavior score of the vehicle section.

2. The low-speed vehicle classification warning method for highway scenarios according to claim 1, wherein: Step 2 specifically includes the following steps: Step 2-1, classify the dataset by hour to obtain the overall road speed in different time periods; Step 2-2, on the basis of hourly classification, further refine the data according to vehicle types; Step 2-3: Classify the dataset according to the vehicle trajectory data according to spatial regularity to obtain different section passing datasets; Step 2-4, extract the passing time of vehicles in the target section through the divided passing dataset of sections; at the same time, extract the passing distance of the target section through the gantry distance matrix D, and use the non-sensing speed measurement model to calculate the driving speeds of all vehicles in the dataset; Step 2-5, delete the abnormal data of the driving speed of vehicles based on the data cleaning model of the box plot; Step 2-6, calculate the average value of all vehicle driving speeds as the traffic speed of the road , and the calculation formula for the traffic speed of the road is as follows: ； Among them, the vehicles representing those passing through the road section , represent the total number of vehicles on the road section; Step 2-7, add the passing speed of the road to the highway section passing speed feature dataset into it.

3. The low-speed vehicle classification warning method for highway scenarios according to claim 2, characterized in that: In Step 2-1, the dataset is divided into 24 categories.

4. A low-speed vehicle classification warning method for highway scenarios according to claim 2, characterized in that: The expression of the element in the gantry distance matrix D in Step 2-4 is as follows: (1)； Among them, dis is the distance between two connected gantries, and inf means that there is no connection between two gantry nodes.

5. A low-speed vehicle classification warning method for highway scenarios according to claim 1, characterized in that: Step 4 specifically includes the following steps: Step 4-1: Extract vehicle trajectory data and transaction data, and calculate the section passing speed of the vehicle ; Step 4-2: Compare the section passing speed of the vehicle with the passing speed of the road under corresponding conditions in the section passing speed feature data set to determine whether there is a low-speed driving behavior; if so, record the passing information of this low-speed driving and execute Step 4-3; otherwise, execute Step 4-3; ​ Step 4-3, judge whether all vehicle trajectories have been traversed; if so, output the record result; otherwise, execute Step 4-1.

6. The low-speed vehicle classification warning method for highway scenarios according to claim 5, wherein: In step 4-2, the average value of the driving speeds of all vehicles is calculated as the traffic speed of the road .

Images