A three-color early warning method for high-speed sections

Abstract

The invention discloses a three-color early warning method for a highway section. Based on uploaded event data, traffic operation data and vehicle monitoring information, a graded early warning model for the danger of the section is established. Based on a weight vector determined by a hierarchical analysis method, a fuzzy comment set and a score vector of the comment, a multi-period safety evaluation is realized in combination with factor weights. The safety grade score of each period of the section is calculated, the main early warning factors of the section are determined, and the overall early warning state of the road is determined by the proportion of the early warning states of the sections constituting the road, so that the traffic police department can control the overall state of the road and the specific accident conditions of the section.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic movement control, and in particular to a three-color early warning method for a high-speed road section. Background Art

[0002] The collection of highway road condition information is mainly through field monitoring cameras installed on both sides of the highway and road inspections by highway maintenance departments and operation departments. Emergency incidents are mostly passive telephone alarms. At the same time, the operation and management modes of highways in various provinces are inconsistent, the highway road condition information release platform is not unified, there are many release channels, and the timeliness and closure of information release are not strong, which leads to poor highway road condition information collection. For example, if a traffic accident occurs on a certain section of the road, the traffic accident information cannot be transmitted to the Internet platform in time, which results in the vehicles traveling on the current section being unable to obtain the traffic accident information in time. Due to the high speed of the highway section, if the driver lacks experience in handling traffic accidents, the accident warning system is not perfect, and rescue cannot be carried out in a timely and effective manner, it is easy to cause secondary accidents, and the traffic safety situation is severe.

[0003] According to data, although there are some highway warning systems on the market, most of these systems serve drivers and have relatively simple functions. The intelligent alarm warning system allows drivers to quickly understand their specific location when an alarm sounds on the highway, and after the alarm sounds an audible and visual alarm to remind the vehicles behind to give way, but the traffic police department cannot understand the overall operation status of the highway and the specific accident situation based on the system. Another warning system specifically targets the impact of the meteorological environment on the road, but ignores the operation status of the road itself. It cannot help the traffic police department control the overall status of the road and the specific accident situation on the road section.

[0004] Chinese patent document CN112950935A discloses a "high-speed traffic control system based on a big data platform". It includes a control center connected to the following devices, wherein the control center can display the data collected by the corresponding devices in response to user operations: video monitoring device, including: a monitoring unit, used to collect and store real-time monitoring videos and patrol videos of the expressway section; an inspection unit, used to detect the monitoring unit; a monitoring and early warning device, including: a road condition monitoring unit, used to monitor and display the road condition information of the expressway section; a traffic information unit, used to manage the road condition information of the expressway section; a normal management device, including: a trajectory analysis unit, used to obtain the driving trajectory of the vehicle on the expressway section, and can judge the safety status of the vehicle based on the driving trajectory. The above technical solution has a single evaluation and cannot help the traffic police department to control the global status of the road and the specific accident situation of the road section. Summary of the invention

[0005] The present invention mainly solves the technical problem that the original technical solution has a single evaluation and cannot help the traffic police department to control the overall state of the road and the specific accident situation of the road section. It provides a three-color early warning method for a highway section. Based on uploaded event data, traffic operation data and vehicle monitoring information, a graded early warning model for the danger of the road section is established. Based on the weight vector, fuzzy comment set and comment score vector determined by the hierarchical analysis method, a multi-period safety evaluation is realized in combination with the factor weights, the safety level score of each period of the road section is calculated, the main early warning factors of the road section are determined, and the overall early warning state of the road is determined by the proportion of the early warning state of the road section constituting the road, so as to facilitate the traffic police department to control the overall state of the road and the specific accident situation of the road section.

[0006] The above technical problem of the present invention is mainly solved by the following technical solution: The present invention comprises the following steps:

[0007] S1 collects and uploads event data, traffic operation data and vehicle monitoring information;

[0008] S2 constructs a judgment matrix and calculates the corresponding eigenvectors of each judgment matrix, which is the factor weight vector, determines the relevant factors of the road section warning score and the relative weights between the factors, and treats all uploaded events as the same factor. Combined with traffic operation factors and vehicle monitoring factors, the road section safety warning score is calculated to determine the safety warning grade of the road section, and the specific deduction of each indicator is calculated to determine the main warning indicator of the road section;

[0009] S3 determines the fuzzy comment set and determines the score vector of the comment;

[0010] S4 establishes a graded warning model for road section danger and conducts multi-cycle safety evaluation.

[0011] The specific implementation algorithm is divided into 5 steps: feature extraction and analysis, determining the weight of each factor, calculating the safety level score of each cycle of the road section, calculating the warning score of the road section, and determining the overall warning status of the road. Feature extraction and analysis mainly determines the factors that affect the status of the road section, including all data items in vehicle monitoring, traffic operation and uploaded event data. It should be noted that the number of speeding vehicles needs to be combined with the number of speeding vehicles in vehicle monitoring and the number of abnormal speeding events in the uploaded event data. Abnormal slow driving is the sum of the number of suspected parking in vehicle monitoring and the number of abnormal slow driving events in the uploaded event number. After determining the factors that affect the status of the road section, the weight of each factor is obtained through a single-layer hierarchical analysis method. The three factors with the largest weights are traffic accidents, reversing / reverse driving, and abnormal parking.

[0012] The third step starts with the business module of the algorithm. First, calculate the safety level score of each period of the road section. Through the weights of each factor obtained previously and the specific values ​​of the corresponding data items, we can get the safety level score of each factor. For example, if a traffic accident occurs on a road section during a certain period, then the score of the traffic accident factor of the road section during this period is 0. Assuming that the weight of the traffic accident is 0.4, and no other accidents occur on the road section during this period, the operation is smooth, and there are no speeding and parked vehicles, then the safety level score of the road section is 60 points.

[0013] The fourth step is to use the exponential weighted average of the safety level scores of the road section's historical cycle to predict the safety level score of the next cycle, which is called the road section warning score here. Compared with other prediction methods, the advantage of exponential weighted average is that it makes full use of events that have occurred and the current state to reasonably estimate the future state. For example, when using exponential weighted average to predict the temperature of the next day, because the temperature will not change suddenly and drastically, the exponential weighted average can well capture the temperature change trend in the past few days and assign a weight to each historical data. The closer the time, the greater the weight, and the closer to the front, the smaller the weight. Similar to temperature, the trend of changes in road section status will not change drastically in a short period of time, so exponential weighted average can also be used for prediction. After obtaining the warning scores of all sections on the road, the current overall warning status of the road can be given by the proportion of each warning status.

[0014] Preferably, the step S2 constructs a judgment matrix and introduces a matrix judgment scale so that the importance of each element in the matrix can be quantitatively displayed: The importance is from 1 to 12, with larger numbers indicating The more important.

[0015] a. Traffic accident > b. Reversing / reverse driving > c. Abnormal parking > d. Speeding vehicle > e. Abnormal slow driving > f. Pedestrian / non-motor vehicle intrusion > g. Outdoor flame / tunnel flame = h. Outdoor smoke / tunnel smoke = i. Fog > j. Traffic congestion > k. Construction violation = l. Road construction = m. Road surface obstacle > n. Traffic level > o. Proportion of large vehicles > p. Vehicle speed standard deviation.

[0016] Calculate the corresponding eigenvectors of each judgment matrix, which is the factor weight vector: The factor weight vector of the road section warning classification:

[0017] W1=[0.5139, 0.2801, 0.1003, 0.069, 0.0366]

[0018] The factor weight vector of the road section warning misclassification:

[0019] W2=[0.2304,0.1784,0.1361,0.1024,0.0758,0.0546,0.0376,0.0376,0.0376,0.0259,0.0178,0.0178,0.0178,0.0128,0.0097,0.0078]

[0020] Preferably, the step S3 of determining the fuzzy comment set specifically includes dividing the warning level and the level of each underlying factor into three levels, that is, the fuzzy comment set of the warning level is:

[0021]

[0022] Determine the classification factor level division rules for section warning and the classification factor level division rules for section warning failure, and determine the fuzzy comment set for each factor according to the warning level.

[0023] No warning needed: [1, 0, 0]

[0024] Level 1 warning: [0.1, 0.6, 0.3]

[0025] Level 2 warning: [0, 0, 1]

[0026] Preferably, the step S3 determines the score vector of the comment to perform quantitative processing on the comment, determines the score interval of the comment, and then determines the score vector of the comment according to the score interval of the comment.

[0027] S = (100, 50, 0)

[0028] Preferably, the step S4 includes calculating the safety level score of each period of the road section, combining the weight vector determined by the hierarchical analysis method with the fuzzy comment set to perform fuzzy comprehensive evaluation modeling, and the fuzzy evaluation is as follows:

[0029]

[0030] Among them, W1 is the factor weight vector of the road section warning classification, A is the matrix composed of the fuzzy comment set corresponding to the road section warning classification factors, is the factor in the fuzzy evaluation R.

[0031] As a preferred method, the comments are quantified based on the score vector of the fuzzy evaluation to determine the score of the road section safety level, thereby determining the warning comment based on the score range of the comment:

[0032]

[0033] Among them, R is the fuzzy evaluation, is the factor in fuzzy evaluation R, is the transpose of the review score vector.

[0034] Preferably, the step S4 includes calculating a road section warning score, taking the road section safety level scores in the first n cycles, and calculating the road section safety warning score based on exponential weighted average:

[0035]

[0036]

[0037] …

[0038]

[0039] in, Represents the road safety warning score of the nth cycle, represents adjustable parameters, Represents the actual road safety level score value of the nth cycle.

[0040] As a preferred method, the road section warning score is compared with the road section safety level score, and the parameters are modified. and When the difference exceeds the error threshold a, the number of cycles n is changed , repeatedly try to modify the number of cycles n and record and If the difference gradually decreases, continue to try to modify the number of cycles; if the difference becomes larger, stop and record the number of cycles n when the error is the smallest.

[0041] Preferably, the step S4 comprises calculating the loss score of each factor in each cycle, the loss score of each factor of the road section is a score vector based on fuzzy evaluation, the comments are quantitatively processed, the safety score of each factor of the road section is determined, the safety score is subtracted from the full score, and then multiplied by the indicator weight.

[0042]

[0043] in, is the fuzzy comment set corresponding to factor i, is the transpose of the review score vector, It is the weight corresponding to the W2 factor i in the weight vector of the road section warning misclassification factor.

[0044] Preferably, the step S4 includes determining the overall warning state of the road, and determining the overall warning state of the road by the proportion of warning states of the road sections constituting the road:

[0045]

[0046]

[0047] Among them, warn_h represents the percentage threshold of the number of high-risk sections on the road, and warn_l represents the percentage threshold of the number of low-risk sections on the road. The final output is the road section warning record table.

[0048] The beneficial effects of the present invention are as follows: based on uploaded event data, traffic operation data and vehicle monitoring information, a graded warning model for road section danger is established; based on the weight vector, fuzzy comment set and comment score vector determined by the hierarchical analysis method, a multi-period safety evaluation is realized in combination with the factor weights; the safety level score of each period of the road section is calculated; the main warning factors of the road section are determined; and the overall warning status of the road is determined by the proportion of the warning status of the road sections that constitute the road, so that the traffic police department can control the global status of the road and the specific accident conditions of the road section. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a flow chart of the present invention.

[0050] Figure 2 This is a schematic diagram of a road section safety warning score of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the technical solution of the present invention is further described in detail below through embodiments and in combination with the accompanying drawings. It should be understood that the specific implementation method described here is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0052] The application scenario of this algorithm is to provide it to the traffic police, so that the traffic police department can control the overall status of the road and the specific accident situations on the road sections. Although there are already some highway warning systems on the market, those systems mostly serve drivers and have relatively simple functions. For example, the conventional intelligent alarm warning system allows drivers to quickly understand their specific location when an alarm sounds on the highway, and after the alarm, it will issue an audible and visual alarm to remind the vehicles behind to avoid it, but the traffic police department cannot understand the overall operation status of the highway and the specific accident situation based on the system. Another warning system specifically targets the impact of the meteorological environment on the road, but ignores the operation status of the road itself.

[0053] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the operations (or steps) as sequential processes, many of the operations (or steps) therein can be implemented in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but can also have additional steps not included in the drawings; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0054] The technical solution of the present invention is further specifically described below through embodiments and in conjunction with the accompanying drawings.

[0055] Example 1

[0056] A three-color early warning method for high-speed sections, such as Figure 1 As shown, the following steps are included:

[0057] S1 collects and uploads event data, traffic operation data and vehicle monitoring information.

[0058] S2 constructs a judgment matrix and calculates the corresponding eigenvectors of each judgment matrix, which is the factor weight vector, determines the relevant factors of the road section warning score and the relative values ​​of the weights between the factors, and uniformly regards all uploaded events as the same factor. It combines it with traffic operation factors and vehicle monitoring factors to calculate the road section safety warning score to determine the safety warning grade of the road section, and calculates the specific deduction of each indicator to determine the main warning indicator of the road section.

[0059] Construct a judgment matrix and introduce a matrix judgment scale so that the importance of each element in the matrix can be quantitatively displayed: The importance is from 1 to 12, with larger numbers indicating The more important.

[0060] a. Traffic accident > b. Reversing / reverse driving > c. Abnormal parking > d. Speeding vehicle > e. Abnormal slow driving > f. Pedestrian / non-motor vehicle intrusion > g. Outdoor flame / tunnel flame = h. Outdoor smoke / tunnel smoke = i. Fog > j. Traffic congestion > k. Construction violation = l. Road construction = m. Road surface obstacle > n. Traffic level > o. Proportion of large vehicles > p. Vehicle speed standard deviation.

[0061] Calculate the corresponding eigenvectors of each judgment matrix, which is the factor weight vector: The factor weight vector of the road section warning classification:

[0062] W1=[0.5139, 0.2801, 0.1003, 0.069, 0.0366]

[0063] The factor weight vector of the road section warning misclassification:

[0064] W2=[0.2304,0.1784,0.1361,0.1024,0.0758,0.0546,0.0376,0.0376,0.0376,0.0259,0.0178,0.0178,0.0178,0.0128,0.0097,0.0078]

[0065] S3 determines the fuzzy comment set and determines the score vector of the comment. Determining the fuzzy comment set specifically includes dividing the warning level and the level of each underlying factor into three levels, that is, the fuzzy comment set of the warning level is:

[0066]

[0067] Determine the classification factor level division rules for section warning and the classification factor level division rules for section warning failure, and determine the fuzzy comment set for each factor according to the warning level.

[0068] No warning needed: [1, 0, 0]

[0069] Level 1 warning: [0.1, 0.6, 0.3]

[0070] Level 2 warning: [0, 0, 1]

[0071] Determine the score vector of the comment. Perform quantitative processing on the comment, determine the score interval of the comment, and then determine the score vector of the comment according to the score interval of the comment.

[0072] S = (100, 50, 0)

[0073] S4 establishes a graded warning model for road section danger and conducts multi-cycle safety evaluation.

[0074] The safety level scores of each period of the road section are calculated, and the fuzzy comprehensive evaluation model is built by combining the weight vector determined by the hierarchical analysis method and the fuzzy comment set. The fuzzy evaluation is as follows:

[0075]

[0076] Among them, W1 is the factor weight vector of the road section warning classification, A is the matrix composed of the fuzzy comment set corresponding to the road section warning classification factors, is the factor in the fuzzy evaluation R.

[0077] Based on the score vector of fuzzy evaluation, the comments are quantified to determine the score of the road section safety level, and then the warning comments are determined based on the score range of the comments:

[0078]

[0079] Among them, R is the fuzzy evaluation, is the factor in fuzzy evaluation R, is the transpose of the review score vector.

[0080] Calculate the road section warning score, such as Figure 2 As shown, the road section safety level scores in the first n cycles are taken, and the road section safety warning score is calculated based on the exponential weighted average:

[0081]

[0082]

[0083] …

[0084]

[0085] in, Represents the road safety warning score of the nth cycle, Represents adjustable parameters, Represents the actual road safety level score value of the nth cycle.

[0086] Compare the road section warning score with the road section safety level score and modify the parameters. and When the difference exceeds the error threshold a, the number of cycles n is changed , repeatedly try to modify the number of cycles n and record and If the difference gradually decreases, continue to try to modify the number of cycles; if the difference becomes larger, stop and record the number of cycles n when the error is the smallest.

[0087] Calculate the loss of each factor in each cycle. The loss of each factor of the road section is a score vector based on fuzzy evaluation. Quantify the comments and determine the safety score of each factor of the road section. Subtract the safety score from the full score and multiply it by the weight of the indicator.

[0088]

[0089] in, is the fuzzy comment set corresponding to factor i, is the transpose of the review score vector, It is the weight corresponding to the W2 factor i in the weight vector of the road section warning misclassification factor.

[0090] Determine the overall warning status of the road, and determine the overall warning status of the road by the proportion of the warning status of the road sections that make up the road:

[0091]

[0092]

[0093] Among them, warn_h represents the percentage threshold of the number of high-risk sections on the road, and warn_l represents the percentage threshold of the number of low-risk sections on the road. The final output is the road section warning record table.

[0094] The data requirements of the algorithm in this application consist of vehicle monitoring, traffic operation data, and uploaded event data. Vehicle monitoring includes the number of speeding vehicles, the number of suspected parked vehicles, and the proportion of large vehicles; traffic operation data includes the standard deviation of road speed and the road traffic level; uploaded event data includes possible accidents on the road, such as reversing / reverse driving, road construction, and abnormal slow driving. The functional requirements of the algorithm are to establish a graded warning model for road danger based on the above three aspects of data.

[0095] The specific implementation algorithm is divided into 5 steps: feature extraction and analysis, determining the weight of each factor, calculating the safety level score of each cycle of the road section, calculating the warning score of the road section, and determining the overall warning status of the road. Feature extraction and analysis mainly determines the factors that affect the status of the road section, including all data items in vehicle monitoring, traffic operation and uploaded event data. It should be noted that the number of speeding vehicles needs to be combined with the number of speeding vehicles in vehicle monitoring and the number of abnormal speeding events in the uploaded event data. Abnormal slow driving is the sum of the number of suspected parking in vehicle monitoring and the number of abnormal slow driving events in the uploaded event number. After determining the factors that affect the status of the road section, the weight of each factor is obtained through a single-layer hierarchical analysis method. The three factors with the largest weights are traffic accidents, reversing / reverse driving, and abnormal parking.

[0096] The third step starts with the business module of the algorithm. First, calculate the safety level score of each period of the road section. Through the weights of each factor obtained previously and the specific values ​​of the corresponding data items, we can get the safety level score of each factor. For example, if a traffic accident occurs on a road section during a certain period, then the score of the traffic accident factor of the road section during this period is 0. Assuming that the weight of the traffic accident is 0.4, and no other accidents occur on the road section during this period, the operation is smooth, and there are no speeding and parked vehicles, then the safety level score of the road section is 60 points.

[0097] The fourth step is to use the exponential weighted average of the safety level scores of the road section's historical cycle to predict the safety level score of the next cycle, which is called the road section warning score here. Compared with other prediction methods, the advantage of exponential weighted average is that it makes full use of events that have occurred and the current state to reasonably estimate the future state. For example, when using exponential weighted average to predict the temperature of the next day, because the temperature will not change suddenly and drastically, the exponential weighted average can well capture the temperature change trend in the past few days and assign a weight to each historical data. The closer the time, the greater the weight, and the closer to the front, the smaller the weight. Similar to temperature, the trend of changes in road section status will not change drastically in a short period of time, so exponential weighted average can also be used for prediction. After obtaining the warning scores of all sections on the road, the current overall warning status of the road can be given by the proportion of each warning status.

[0098] Example 2

[0099] 1. Functional requirements

[0100] A graded warning model for road section danger is established based on uploaded event data, traffic operation data and vehicle monitoring information.

[0101] 2. Data requirements

[0102] Vehicle Monitoring

[0103] Traffic operation data

[0104] Upload event data

[0105] 3. Algorithm Process

[0106] like Figure 1 Shown

[0107] 4. Calculation process

[0108] 4.1. Determine the weight of each influencing factor

[0109] 4.1.1. Determine the relevant factors of the road section warning score and the relative weights between the factors:

[0110] All uploaded events are uniformly regarded as the same factor and combined with traffic operation factors and vehicle monitoring factors (used to calculate the road section safety warning score, thereby determining the road section's safety warning grade).

[0111] 4.1.2. Factor weight vector of road section warning classification:

[0112] W1 = [0.5139, 0.2801, 0.1003, 0.069, 0.0366]

[0113] 4.1.3. Determine the factors related to the loss of points on a road section and the relative weights between the factors:

[0114] All uploaded events are considered as independent factors and combined with traffic operation factors and vehicle monitoring factors (used to calculate the specific penalty points for each indicator, thereby determining the main warning indicators for the road section).

[0115] a. Traffic accident > b. Reversing / reverse driving > c. Abnormal stoppage > d. Speeding vehicle > e. Abnormal slow driving > f. Pedestrian / non-motor vehicle intrusion > g. Outdoor flame / tunnel flame = h. Outdoor smoke / tunnel smoke = i. Foggy weather > j. Traffic congestion > k. Construction violation = l. Road construction = m. Road surface obstacle > n. Traffic level > o. Large vehicle proportion > p. Vehicle speed standard deviation

[0116] 4.1.3. Factor weight vector for road section warning classification:

[0117] W2=[0.2304,0.1784,0.1361,0.1024,0.0758,0.0546,0.0376,0.0376,0.0376,0.0259,0.0178,0.0178,0.0178,0.0128,0.0097,0.0078]

[0118] Table 1 Relative weights of factors related to loss of points

[0119] 4.2. Determining the fuzzy review set

[0120] Considering the feasibility of the early warning model, the early warning level and the level of each underlying factor are divided into three levels, that is, the fuzzy comment set of the early warning level is V = {no early warning, first-level early warning, second-level early warning}

[0121] 4.2.1. Rules for classifying factors of road section warning:

[0122] Table 2 Rules for classifying factors of road section warning

[0123] 4.2.2. Rules for classifying road section warning loss classification factors:

[0124] Table 3 Rules for classifying road section warning loss classification factors

[0125] 4.2.3. Determine the fuzzy comment set of each factor according to the warning level:

[0126] No warning needed: [1, 0, 0]

[0127] Level 1 warning: [0.1, 0.6, 0.3]

[0128] Level 2 warning: [0, 0, 1]

[0129] 4.3. Calculating the road section warning score

[0130] 4.3.1. Fuzzy evaluation modeling

[0131] Combining the weight vector determined by the hierarchical analysis method with the fuzzy comment set, a fuzzy comprehensive evaluation model is built. The fuzzy evaluation is as follows:

[0132] R = W1 * A = (r1, r2, r3)

[0133] Among them, W1 is the factor weight vector of the road section warning classification, and A is the matrix composed of the fuzzy comment set corresponding to the road section warning classification factors.

[0134] 4.3.2. Calculate the road section safety level score

[0135] Based on the score vector of fuzzy evaluation, the comments are quantified to determine the score of the road section safety level, and then the warning comments are determined based on the score range of the comments:

[0136]

[0137] Among them, R is the fuzzy evaluation, Among them, R is the fuzzy evaluation, is the factor in fuzzy evaluation R, is the transpose of the review score vector.

[0138] 4.3.3. Calculating the road section safety warning score

[0139] Take the road section safety level scores in the first n cycles (5 minutes for one cycle) and calculate the road section safety warning score based on the exponential weighted average:

[0140]

[0141]

[0142] …

[0143]

[0144] in, Represents the road safety warning score of the nth cycle, represents adjustable parameters, Represents the actual road safety level score value of the nth cycle.

[0145] 4.3.4. Compare the road section warning score with the road section safety level score and modify the parameters

[0146] Compare the road section warning score with the road section safety level score and modify the parameters. and When the difference exceeds the error threshold a, the number of cycles n is changed , repeatedly try to modify the number of cycles n and record and If the difference gradually decreases, continue to try to modify the number of cycles; if the difference becomes larger, stop and record the number of cycles n when the error is the smallest.

[0147] 4.3.5. Determine the road section safety warning level

[0148] according to The value of is combined with the score vector of the comments to determine the risk warning level of the road section.

[0149] 4.3.6. Quantify the comments and determine the score range of the comments:

[0150] 4.4. Calculate the loss of each factor in each period

[0151] 4.4.1. Calculate the loss of each factor in each cycle in the road section warning loss classification

[0152] The loss score of each factor of the road section is a score vector based on fuzzy evaluation. Pinghu is quantitatively processed to determine the safety score of each factor of the road section. The safety score is subtracted from the full score and then multiplied by the weight of the indicator.

[0153]

[0154] in, is the fuzzy comment set corresponding to factor i, is the transpose of the review score vector, It is the weight corresponding to the W2 factor i in the weight vector of the road section warning misclassification factor.

[0155] 4.4.2. Calculation of warning loss scores for each factor

[0156] According to the factor loss within n cycles, the warning loss score of each factor of the road section is calculated based on the exponential weighted average. The calculation process is similar to the road section safety warning score.

[0157] 4.4.3. Determine the main warning factors of the road section

[0158] The main warning factor is the factor with the highest warning loss score among the classification factors during the warning, which is the main reason for warnings at all levels of the road section.

[0159] 4.5. Determine the overall road warning status

[0160] The overall warning status of a road is determined by the proportion of warning status of the road sections that make up the road.

[0161]

[0162]

[0163] Among them, warn_h represents the threshold of the proportion of high-risk sections on the road, and warn_l represents the threshold of the proportion of low-risk sections on the road.

[0164] 5. Implementation process

[0165] 5.1. Demand data processing

[0166] (1) Initialize basic information:

[0167] Traverse basic information such as roads and road sections.

[0168] (2) Obtaining and processing vehicle passing data:

[0169] The vehicle passing data within the cycle is obtained by obtaining the most recent cycle period with the current time as the end time according to the cron expression.

[0170] Merge the previously saved rollback vehicle data into the data set of the current cycle, and clear the rollback data set in preparation for re-accumulation.

[0171] (3) Calculate the number of vehicles on the road:

[0172] If the vehicle passing data set is not empty, extract the license plate number of each vehicle, remove duplicates and calculate the total number of vehicles on the way.

[0173] (4) Create road records:

[0174] Create an object, fill in relevant information such as the current date, number of vehicles in transit, unique identifier (uuid), timestamp, etc., and add it to the list of vehicles in transit record collection.

[0175] (5) Grouping vehicle passing data by road section:

[0176] Use stream operations to group vehicle passing data by road section (checkpoint number + direction) to facilitate subsequent analysis by road section.

[0177] 5.2. Algorithm Implementation

[0178] 5.2.1. Analyze road section indicators

[0179] (1) Traverse all road segment information and construct unique keys for the starting point and end point of each road segment.

[0180] (2) Check whether there is any vehicle data matching the start and end points of the road section

[0181] (3) If matching data exists, data analysis is performed, including but not limited to average speed analysis of road sections and early warning information analysis:

[0182] 5.2.1.1. Calculation logic for vehicles passing through expressways

[0183] Processing of vehicle passing data analysis on expressways aims to provide functions such as vehicle speed, vehicle classification statistics and abnormal driving path detection.

[0184] (1) Data verification and initialization:

[0185] First, check whether the upstream checkpoint vehicle passing data and the downstream checkpoint vehicle passing data are empty. If either is empty, return directly.

[0186] Initialize a data set to store the passing analysis results of each vehicle.

[0187] (2) Upstream checkpoint data grouping:

[0188] Use stream operations to group upstream vehicle data by license plate number and vehicle type combination to facilitate subsequent matching with downstream data.

[0189] (3) Vehicle type counting and lane change detection:

[0190] Initialize three counters: the number of large vehicles, the number of small vehicles, and the number of vehicles changing lanes.

[0191] Traverse the downstream vehicle passing data and perform the following operations for each record:

[0192] Generate a unique identifier of the license plate number plus the vehicle type.

[0193] Determine whether there is a vehicle with this identifier in the upstream data. If not, skip the current loop.

[0194] Calculate the time difference between the vehicle passing through the two checkpoints and convert it into seconds.

[0195] If the time difference is negative, it means the data may be wrong and the record is skipped.

[0196] Increment the corresponding counter according to the vehicle type.

[0197] Check and record if the vehicle changes lanes while in motion (i.e., the entry and exit lane numbers are different).

[0198] Calculate and record the average speed of the vehicle on the road section.

[0199] Encapsulate the processed data into an object and store it in a data collection.

[0200] Remove paired data items from the upstream and downstream vehicle passing data sets to reduce comparisons in subsequent loops.

[0201] (4) Data integrity check:

[0202] After the traversal is completed, if the data set is still empty, it means that no vehicle has been successfully paired, and an exception is thrown to indicate that there is no paired vehicle passing record on this highway section.

[0203] (5) Organize and return the analysis results:

[0204] The unpaired vehicle data is marked as needing to be rolled back (although this step is commented out in the code, in theory this part of the data should be processed for further analysis or subsequent processing).

[0205] Create and populate a result object, including:

[0206] Pairs of passing vehicles and their average speed are mapped into the data set.

[0207] Unpaired vehicle passing data list rollback collection.

[0208] The number of large and small vehicles and the number of vehicles changing lanes.

[0209] 5.2.1.2. Analysis of road section indicators and speed warning calculation logic

[0210] Through complex logical processing, the traffic conditions of roads are quantified (such as standard deviation, maximum speed and minimum speed), and the traffic status levels are divided according to the traffic flow characteristics. The necessary early warning mechanisms are triggered to provide detailed data support for traffic management and planning.

[0211] (1) Calculate the standard deviation of vehicle speed:

[0212] Traverse all vehicle sets and calculate the sum of squared deviations of each vehicle's speed from the average speed.

[0213] Divide the sum of squared deviations by the number of vehicles and take the square root to get the standard deviation of vehicle speed.

[0214] (2) Get the maximum and minimum speed:

[0215] Sort the collect set by speed, and the sorting basis is the speed of the vehicle.

[0216] Get the speed values ​​of the first and last elements after sorting and fill them back into the object properties.

[0217] (3) Calculate the average travel time and section tolerance time:

[0218] Convert the average speed to meters per second (multiply by the free stream speed).

[0219] Calculate the average travel time: segment length divided by average speed.

[0220] Calculate dynamic window time range (minimum and maximum allowed time).

[0221] Calculate the segment tolerance time: average travel time plus the minimum allowed time minus the maximum allowed time.

[0222] (4) Calculate the travel time ratio:

[0223] Convert the average speed to meters per second.

[0224] Logical: Compares the average travel time of a segment with the free-flow travel time and determines the travel time ratio. If the average travel time is longer, the ratio is equal to the average travel time divided by the free-flow travel time; otherwise, it defaults to 1, indicating free-flow.

[0225] (5) Determine traffic conditions and set warning information:

[0226] Judging traffic conditions based on average speed:

[0227] If the average speed is greater than the smooth threshold, it is set to a smooth state.

[0228] If it is between the slow-moving threshold and the smooth-moving threshold, it is set to the slow-moving state and generates slow-moving event information.

[0229] If it is between the congestion threshold and the slow-moving threshold, it is set to a light congestion state and congestion event information is generated.

[0230] Otherwise, it is set to a severe congestion state and generates severe congestion event information.

[0231] Generate an event information object, including event type, name, timestamp, etc., and add it to the warning information queue.

[0232] (6) Set the number of overspeed and underspeed vehicles:

[0233] According to the input parameters and the pre-counted overspeed and lowspeed vehicle mapping, the number of overspeed vehicles and lowspeed vehicles on the road section is set to the object.

[0234] (7) Encapsulate the result and return it:

[0235] All calculation and analysis results are encapsulated into an object, and this object is returned at the end of the function as the comprehensive output of the function processing the road segment data.

[0236] 5.2.1.3. Calculation of warning comments

[0237] Get the monitoring weight in the traffic cycle configuration, analyze and process the given road section indicator data, calculate the scores of different dimensions (road section safety, traffic operation, vehicle monitoring, environmental impact) and the comprehensive total score, and judge the warning level based on this, and finally encapsulate this information into an object and return it.

[0238] 5.2.1.3.1. Road safety dimension

[0239] The attribution degree matrix of the road safety dimension is calculated based on the basic information of the road section.

[0240] (1) Read configuration and parameter variables:

[0241] Get the number of lanes, turning radius, absolute value of slope, defect rate, whether it is a bridge or tunnel, and accident rate from the input road section basic information object.

[0242] Get the preset values ​​for edge monitoring, including the maximum, minimum, and default values ​​of various security attributes.

[0243] (2) Create the result array:

[0244] Based on the comparison results, a 6x3 two-dimensional array is constructed, where each row represents a road feature (number of lanes, turning radius, absolute value of slope, defect rate, bridge and tunnel signs, accident rate), and each column corresponds to a score of the attribute level. This matrix comprehensively reflects the safety status of various indicators of the input road surface.

[0245] (3) Calculate the degree of belonging of each dimension:

[0246] For each security attribute, conditional judgment is performed, the attribute value is compared with the preset threshold, the attribute is graded according to the comparison result, and these attribute levels are mapped to array form for storage.

[0247] Number of lanes: The number is compared with the configured maximum and minimum values.

[0248] Turning radius: Compare its size with the preset maximum and minimum turning radii.

[0249] Slope, defect rate and accident rate: The level is also determined by comparing the numerical value with the threshold.

[0250] Bridge and tunnel markings: If they are consistent with the preset bridge and tunnel markings, they will be specially treated.

[0251] (4) Summary of results:

[0252] The security attribution result of each attribute is stored in a two-dimensional array edge. Each row of the array corresponds to a security dimension, and the three values ​​in each row represent the attribution level details under the dimension.

[0253] (5) Return result:

[0254] Finally, a two-dimensional array containing six different dimensions of safety attribution is returned, which fully reflects the comprehensive evaluation results of each safety dimension of the road surface.

[0255] Through the above steps, according to the preset monitoring configuration, an attribution matrix reflecting the risk or safety level of each safety dimension of the road surface is calculated and returned, which can provide data support for subsequent road safety assessment and management.

[0256] 5.2.1.3.2. Traffic operation dimension

[0257] The attribution degree matrix of the traffic operation dimension is calculated based on the road section index information.

[0258] (1) Read configuration and parameter variables:

[0259] Get traffic operation safety warning threshold and default value configuration information, which includes preset parameters such as maximum and minimum values ​​of free flow speed ratio, standard value, saturation level and other traffic operation related information.

[0260] Initialize a 3x3 double-dimensional array traffic to store the calculated attribution results.

[0261] (2) Processing free flow velocity ratio (avgFreeFlowVelocityRtio):

[0262] According to the comparison between the free flow speed ratio and the maximum and minimum values ​​in the configuration, and the comparison between avgFreeFlowVelocityRtio and the preset maximum and minimum values, the traffic state's attribution level is determined, and the attribution degree array of the corresponding level is obtained from the configuration and filled into the first row of traffic. There are three attribution levels: AttributionegreeNo (not up to standard), AttributionegreeOne (average), AttributionegreeTwo (excellent), and the corresponding attribution degree values ​​are stored in traffic[0].

[0263] (3) Default settings for warning status:

[0264] Directly set traffic[1] to the default state vector {1, 0, 0}, indicating that the warning level is not considered or is set to normal by default.

[0265] (4) Processing saturation (saturationLevel):

[0266] Determine whether the saturation in the input is empty. If it is empty, use the default value in the configuration.

[0267] The saturation is compared with the maximum and minimum values ​​in the configuration to determine its attribute level, and the result is stored in traffic[2].

[0268] (5) Return result:

[0269] After completing the processing of all the above dimensions, the two-dimensional array traffic containing the degree of ownership information of the traffic operation dimensions is finally returned.

[0270] This function comprehensively analyzes two key indicators of traffic data (free flow speed ratio and saturation), and classifies traffic status into several predefined levels based on these indicators, providing a quantitative reference for traffic management or analysis.

[0271] 5.2.1.3.3. Vehicle monitoring dimension

[0272] The attribution degree matrix of the vehicle dimension is calculated based on the road segment index information.

[0273] It comprehensively considers various vehicle operating conditions, systematically evaluates the traffic conditions of specific road sections through set thresholds, and expresses them in numerical form, which can serve as the data basis for further analysis or decision support.

[0274] (1) Get configuration and parameters:

[0275] Get the vehicle monitoring value and the attribution threshold and default value of the monitoring level from the global traffic flow configuration.

[0276] Get the values ​​of various indicators in the object, such as the proportion of large vehicles, the number of speeding vehicles, the number of low-speed vehicles, the number of lane changes, and the lane change rate.

[0277] (2) Calculate the degree of belonging of each dimension:

[0278] Large vehicle ratio: Determine the attribution level (none, level 1, or level 2) based on the comparison between the large vehicle ratio and the configured minimum and maximum values. Determine the attribution level (0, 1, 2) based on the comparison between the large vehicle ratio and the preset minimum and maximum values, and convert the attribution value of the corresponding level into a double-precision array to fill in the first row of the result two-dimensional array.

[0279] Number of speeding vehicles: Similarly, the degree of belonging is determined by comparing the number of speeding vehicles with the corresponding threshold. After the comparison, the degree of belonging is determined and the second row of the result array is updated.

[0280] Number of low-speed vehicles: Determine the degree of ownership based on the value of low speed / illegal parking. Determine the degree of ownership based on the value range and update the third row of the result array.

[0281] Lane change rate: Finally, the ownership level is determined based on the proportion of vehicles changing lanes. The lane change frequency is checked for null values ​​and ranges, and the last row of the ownership array is filled in according to the situation.

[0282] (3) Construct the result array:

[0283] For each dimension, fill a 4-row, 3-column two-dimensional array vehicle with the values ​​corresponding to the degree of attribution.

[0284] (4) Return result:

[0285] The function returns this two-dimensional array, where each row represents a different vehicle behavior metric, and each column may correspond to a different aspect of that metric (although the specific details are not shown in the code).

[0286] After the above processing, a two-dimensional array of 4 rows and 3 columns is obtained. Each row represents the attribution coefficient of a dimension. The coefficient reflects the running status of the vehicle in the corresponding dimension, thereby assisting in analyzing the traffic conditions of the road section.

[0287] 5.2.1.3.4. Environmental impact dimension

[0288] The attribution matrix of the impact of weather conditions on the road section is calculated based on the basic information of the road section.

[0289] The attribution matrix of the impact of comprehensive key weather parameters can be used as a data basis for further analysis or decision support, such as predicting traffic flow, arranging maintenance measures, etc.

[0290] (1) Parameter acquisition and default value processing:

[0291] The function receives a parameter of the weather warning threshold and default value type. The object contains the real-time weather observation data of the road section, including rainfall, snowfall, visibility and wind speed.

[0292] For each of these parameters, if the value passed in is empty, the function will use the default value preset in the configuration object Traffic Index Configuration Class instead. This ensures that even if some observations are missing, there is still basic data for evaluation.

[0293] (2) Logic of attribution evaluation:

[0294] For each weather factor, the function will make the following judgments to determine its attribute level and store the result in the corresponding position of the two-dimensional array:

[0295] Rainfall: It is graded based on its comparison with the configured minimum and maximum rainfall.

[0296] Snowfall: Same logic as rainfall processing, similar attribution judgment is made based on snowfall.

[0297] Visibility: When visibility reaches the maximum value, it is classified as "no impact", when it is lower than the minimum value, it is classified as "serious impact", and the rest is "mild impact". The judgment standard here is the visibility. The lower the visibility, the greater the impact may be.

[0298] Wind speed: Same logic as rainfall and snowfall, high wind speeds can mean higher risk or impact.

[0299] (3) Attribution mapping:

[0300] The attribution level is converted into specific numerical arrays for storage. These arrays correspond to four dimensions in the two-dimensional array weather, and each dimension represents the attribution evaluation result of a weather factor.

[0301] (4) Return result:

[0302] Finally, the function returns the four-row and three-column two-dimensional array weather. Each row of the array corresponds to three attribute level values ​​of a weather factor (precipitation, snowfall, visibility, wind speed), which can be used to further analyze the impact of road weather on traffic conditions.

[0303] 5.2.1.3.5. Performing matrix operations

[0304] Construct a two-level fuzzy evaluation matrix: multiply the degree of belonging of each dimension by the corresponding weight matrix to obtain R1-R4, which represent the weighted scores of different dimensions.

[0305] Constructing first-level comments: Further integrate the results of R1-R4 to form a higher-level evaluation matrix vector.

[0306] Calculate the comprehensive score: Combine the first-level evaluation matrix and higher-level weights to calculate the final score R.

[0307] 5.2.1.3.6. Calculating the reciprocal factor and score

[0308] For each dimension and composite score, calculate the inverse factor of its weight.

[0309] The scores of each dimension are calculated using the inverse and weight, including the scores of road safety, traffic operation, vehicles, weather and the comprehensive total score.

[0310] 5.2.1.3.7. Determine the warning level

[0311] Determine the warning level of the road segment based on the total score and return a numerical object.

[0312] The entire function implements a set of conversion processes from raw road section indicator data to comprehensive early warning assessment information, involving data processing in multiple dimensions and complex mathematical operations, providing road section safety early warning assessment information for highway management, helping decision makers understand road section conditions and take corresponding measures.

[0313] 5.2.2. Analyzing overall road indicators

[0314] Analyze comprehensive indicators of highways, including but not limited to average travel speed, travel time ratio, delay time ratio, etc.

[0315] (1) Initialization and summation VKT (Vehicle Kilometers Traveled):

[0316] The function receives a list of road segment indicators as a parameter.

[0317] Safely accumulate the VKT values ​​of all road segments to ensure the atomicity of operations in a multi-threaded environment.

[0318] (2) Calculate the average travel speed:

[0319] Traverse each road section, and multiply the average speed of the road section by the proportion of its VKT to the total VKT to get the average travel speed of the entire road. This process takes into account the impact of the traffic flow of the road section on the average speed.

[0320] (3) Calculation of travel time ratio and delay time ratio:

[0321] The weighted average of the travel time ratios (i.e., the ratio of the actual travel time to the theoretical travel time under free flow conditions) of each road section is calculated, where the weight is the VKT share of the road section.

[0322] The delay time ratio is calculated in the same way, but it is necessary to additionally check whether the delay time ratio of the road section and VKT are empty.

[0323] (4) Statistics of congestion length:

[0324] Determine whether a road section is congested (not basically unobstructed or unobstructed), accumulate the lengths of these congested sections, and finally calculate the ratio of the congested length to the total length (converted to kilometers) to assess the degree of road congestion.

[0325] (5) Calculate the average speed of the truck:

[0326] Similar to the average trip speed calculation method, but calculated for the average speed of larger vehicles.

[0327] (6) Calculation of traffic status ratio:

[0328] Initialize a mapping to record the proportions of various traffic flow states, and traverse all state values ​​to set them to the initial value 0.

[0329] The road segment collection is grouped, segments are aggregated according to traffic levels, and then the proportion of the total length of each level of segments to the total length is calculated.

[0330] Store the calculated proportions of each level into the previously initialized map for subsequent viewing and use.

[0331] (7) Output comprehensive indicator information:

[0332] The above calculated indicators (such as average speed, travel time ratio, delay time ratio, proportion of each traffic state, congestion length, etc.) are set to a new road indicator information object.

[0333] Returns this object, which contains the comprehensive evaluation indicators of the entire highway.

[0334] 5.2.3. Further analysis of the incident

[0335] This function is mainly used to process highway data, including initializing basic information, obtaining and processing vehicle data, and analyzing road sections and road indicators. Its core purpose is to identify and warn of potential road problems by analyzing historical traffic data.

[0336] (1) Parameter analysis and time calculation

[0337] Input parameters: accepts cron expressions (used to determine the time rules for periodic task execution), Elasticsearch client instances, integer variables T (representing the number of cycles analyzed), integer variables cycle (current cycle identifier), and string uuid (unique identifier).

[0338] Calculate the start and end time: Calculate the current end time of the execution cycle and the corresponding cycle time length based on the cron expression, and then deduce the start time of the analysis.

[0339] (2) Indexing and data acquisition

[0340] The road indicator information and event information are extracted according to the above index array and time range.

[0341] (3) Data preprocessing and grouping

[0342] Group by period: First, use stream operations to group the extracted data according to their respective periods to form two hash sets, with the key being the period and the value being a list of all road indicators or event information within the period.

[0343] (4) Fusion analysis of road indicators and events

[0344] Traversal cycle: traverse the specified T cycles from the current cycle in reverse order.

[0345] Integrate road indicators: For each cycle, if there is road indicator data, traverse this data, create a road section warning indicator object and fill in relevant information, including road section ID, proportion of large vehicles, number of low-speed vehicles, traffic level, etc.

[0346] Processing event data: Similarly, the event data is processed by first grouping them according to edge IDs, then breaking them down according to event types, and counting the number of events of each type.

[0347] (5) Result aggregation and storage

[0348] Trend information list generation: further organize the data in the hash set through functions, which may include sorting, filtering and other operations.

[0349] Road trend index construction: Generate higher-level road trend index information based on the above list by analyzing warning roads and road network indicators, which may involve the aggregation or analysis of trends on individual road sections.

[0350] Data persistence: The processed trend information and road trend index are written back to the Elasticsearch database to achieve persistent storage of data.

[0351] Additional processing: Calling functions to perform some additional event processing or conversion that is not explicitly stated in the main flow.

[0352] This function aims to generate road warning information to support traffic management and decision-making by collecting, grouping, and fusing road indicators and event data in a specified time period.

[0353] 5.2.4. Saving Data

[0354] Finally, the processed data is written back to the database to achieve persistent storage of data.

[0355] Insert the vehicle-in-transit records, road section indicators, highway indicators, and warning information scores into Elasticsearch.

[0356] The entire process involves complex data processing logic, reflecting the understanding and application of big data processing and early warning system development.

[0357] The specific embodiments described herein are merely examples of the spirit of the present invention. The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that technicians in the technical field to which the present invention belongs can make various modifications or supplements to the described specific embodiments or replace them in a similar manner, but will not deviate from the spirit of the present invention or exceed the scope defined in the attached claims. For ordinary technicians in this field, multiple deformations and improvements can also be made without departing from the concept of the present invention. Therefore, the scope of protection of the patent of the present invention shall be based on the attached claims.

Claims

1. A three-color early warning method for expressway sections, characterized in that: The following steps are involved: S1 collects and uploads event data, traffic operation data and vehicle monitoring information; S2 constructs a judgment matrix and calculates the corresponding eigenvectors of each judgment matrix, which is the factor weight vector, determines the relevant factors of the road section warning score and the relative weights between the factors, and treats all uploaded events as the same factor. Combined with traffic operation factors and vehicle monitoring factors, the road section safety warning score is calculated to determine the safety warning grade of the road section, and the specific deduction of each indicator is calculated to determine the main warning indicator of the road section; S3 determines the fuzzy comment set and the score vector of the comment, calculates the safety level score of each period of the road section, combines the weight vector determined by the hierarchical analysis method with the fuzzy comment set, conducts fuzzy comprehensive evaluation modeling, and quantitatively processes the comments based on the score vector of the fuzzy evaluation to determine the score of the road section safety level; S4 establishes a hierarchical warning model for road section danger and conducts multi-cycle safety evaluation; Calculate the road section warning score and take the road section safety level score θ in the first n cycles n , calculate the road section safety warning score v based on exponential weighted average n , compare v n With θ n , modify the parameter β, when v n With θ n When the difference exceeds the error threshold a, change β by modifying the number of cycles n, repeat the modification and record v n With θ n If the difference gradually decreases, continue to try to modify the number of cycles; if the difference becomes larger, stop and record the number of cycles n when the error is the smallest.

2. A three-color early warning method for a highway section according to claim 1, characterized in that: The step S2 constructs a judgment matrix and introduces a matrix judgment scale so that the importance of each element in the matrix can be quantitatively displayed: i / x j The importance of the number is from 1 to 12, the larger the number, the more x i / x j The more important.

3. A three-color early warning method for a highway section according to claim 1, characterized in that: The step S3 of determining the fuzzy comment set specifically includes dividing the warning level and the level of each underlying factor into three levels, that is, the fuzzy comment set of the warning level is: V={no warning required, level 1 warning, level 2 warning} Determine the classification factor level division rules for section warning and the classification factor level division rules for section warning failure, and determine the fuzzy comment set for each factor according to the warning level.

4. A three-color early warning method for a highway section according to claim 1 or 3, characterized in that: The step S3 determines the score vector of the comment to perform quantitative processing on the comment, determines the score interval of the comment, and then determines the score vector of the comment according to the score interval of the comment.

5. A three-color early warning method for a highway section according to claim 1, 2 or 3, characterized in that: The step S4 includes calculating the safety level score of each period of the road section, combining the weight vector determined by the hierarchical analysis method with the fuzzy comment set, and performing fuzzy comprehensive evaluation modeling. The fuzzy evaluation is as follows: R=W1*A=(r1,r2,r3) Among them, W1 is the factor weight vector of the road section warning classification, A is the matrix composed of the fuzzy comment set corresponding to the road section warning classification factors, and r1, r2, and r3 are the factors in the fuzzy evaluation R.

6. A three-color early warning method for a highway section according to claim 5, characterized in that: Based on the score vector of fuzzy evaluation, the comments are quantified to determine the score of the road section safety level, and then the warning comments are determined based on the score range of the comments: Among them, R is the fuzzy evaluation, r1, r2, r3 are the factors in the fuzzy evaluation R, S T is the transpose of the review score vector.

7. A three-color early warning method for a highway section according to claim 1, 2 or 3, characterized in that: The step S4 includes calculating the road section warning score, taking the road section safety level scores in the first n cycles, and calculating the road section safety warning score based on the exponential weighted average: v1=θ1 v2=β×v1+(1-β)×θ2 … v n =β×v n-1 +(1-β)×θ n Among them, v n Represents the road safety warning score of the nth cycle, represents the adjustable parameter, θ n Represents the actual road safety level score value of the nth cycle.

8. A three-color early warning method for a highway section according to claim 1, 2 or 3, characterized in that: The step S4 includes calculating the loss score of each factor in each cycle. The loss score of each factor of the road section is a score vector based on fuzzy evaluation. The comments are quantified to determine the safety score of each factor of the road section. The safety score is subtracted from the full score and then multiplied by the indicator weight. Score i =(100-a i ×S T )*w i Among them, a i is the fuzzy comment set corresponding to factor i, S T is the transpose of the review score vector, w i It is the weight corresponding to the W2 factor i in the weight vector of the road section warning misclassification factor.

9. A three-color early warning method for a highway section according to claim 1, 2 or 3, characterized in that: The step S4 includes determining the overall warning state of the road, and determining the overall warning state of the road by the proportion of warning states of the road sections constituting the road: Among them, warn_h represents the threshold of the proportion of high-risk sections on the road, and warn_l represents the threshold of the proportion of low-risk sections on the road.

Images