Emergency lane temporary opening optimization method and system for high-speed congestion early warning

Abstract

This invention discloses an optimized method and system for the temporary activation of emergency lanes for highway congestion early warning, belonging to the field of intelligent traffic management technology. It obtains accurate traffic flow parameters through video dual-slice preprocessing and manual sampling linear correction, constructs a spatiotemporal density distribution model based on the traffic flow conservation equation, and achieves accurate congestion early warning by combining spatiotemporal Kriging interpolation with an ARMA model that integrates upstream and downstream influences. The optimal activation length and closure timing are calculated based on triple safety constraints, and a cost-accuracy trade-off model is established to optimize monitoring points. This invention, employing the above method and system, effectively solves the problems of blindly opening the entire length of emergency lanes, high risk of congestion in emergency access lanes, and unreasonable closure timing in existing technologies. It can significantly improve road traffic efficiency while ensuring unobstructed rescue channels.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic management technology, and in particular to an optimized method and system for the temporary activation of emergency lanes for highway congestion warning. Background Technology

[0002] With the continuous improvement of my country's expressway network and the sustained growth of motor vehicle ownership, expressway traffic flow has been increasing year by year, and congestion during holidays and peak periods has become increasingly prominent. Emergency lanes, as the "lifeline" of expressways, are mainly used for emergency vehicles such as those involved in engineering rescue, firefighting, and medical assistance, and are normally prohibited from being used by other vehicles. However, during severe congestion, the reasonable and temporary use of emergency lanes to divert traffic can effectively improve road capacity and alleviate traffic pressure.

[0003] Currently, the temporary activation of emergency lanes on highways mainly relies on the on-site experience and judgment of traffic management personnel, which has the following core shortcomings: 1. Delayed and blind decision-making: Emergency lanes are often only opened on the entire section after congestion has already formed, which not only misses the best time to alleviate congestion, but also excessively occupies emergency lanes, posing a safety hazard that prevents emergency vehicles from passing. 2. Lack of refined management: It is impossible to accurately locate congestion areas with a granularity of 200m, so only a "one-size-fits-all" full-section opening mode can be adopted, resulting in excessive time and space occupation of emergency lanes; 3. Lack of safety constraints: No congestion control mechanism has been established for the emergency lane itself, posing a risk that the emergency lane will be blocked after it is opened; 4. Inappropriate timing of closure: The clearance time required for vehicles to safely change lanes was not considered, and sudden closure could easily lead to accidents involving vehicles cutting in. 5. Inadequate surveillance deployment: Most existing video surveillance points are conventional traffic monitoring deployments, which have not been optimized for emergency lane decisions, resulting in large errors in the estimation of traffic flow parameters on key road sections.

[0004] Some invention patents rely solely on real-time traffic flow for decision-making, failing to incorporate congestion warnings and precise regional positioning, and lacking a safety constraint and closure mechanism for emergency lanes. Therefore, there is an urgent need to develop an optimized method for the temporary use of emergency lanes that integrates precise warnings, minimal occupancy control, and safe closure. Summary of the Invention

[0005] The purpose of this invention is to provide an optimized method and system for the temporary activation of emergency lanes for highway congestion warning. It can achieve accurate statistics of traffic flow parameters by combining dual-slice video preprocessing with YOLOv8+Bot-Sort algorithm and manual sampling linear correction. It can achieve refined congestion warning based on traffic flow conservation equation, spatiotemporal Kriging interpolation and ARMA model that integrates upstream and downstream traffic. It establishes a triple safety constraint and lane clearing mechanism and optimizes monitoring points to solve the problems of blind opening of emergency lanes, delayed decision-making and unreasonable closing timing. Under the premise of ensuring the smooth flow of rescue channels, it can effectively improve road traffic efficiency.

[0006] To achieve the above objectives, this invention provides an optimized method for the temporary activation of emergency lanes for highway congestion warning, comprising the following steps: S1. Collect real-time video data from multiple video monitoring points on the target section of the highway. After double-slice preprocessing of the video data, statistical traffic flow parameters of each monitoring point are calculated based on machine vision algorithms, and statistical errors are corrected by manual sampling linear correction algorithm. S2. Based on the traffic flow conservation equation, a spatiotemporal density distribution model of road segments is constructed. Spatiotemporal Kriging interpolation is used to fill in missing data. An ARMA time series model that integrates upstream and downstream influences is used to predict the traffic flow density of each segment of the road segment within a preset time period in the future, so as to realize the location of congestion areas and the early warning of continuous congestion at a preset spatial granularity. S3. Based on the location results of the congested area and the triple safety constraints, calculate the shortest emergency lane activation length and execute the activation decision, and determine the safe closure time based on the lane clearing time. S4. Establish a trade-off model between monitoring accuracy and deployment cost, with the goal of minimizing total deployment cost and the constraint of minimizing the accuracy of parameter estimation for the entire road segment, to optimize the deployment of video surveillance points on the target road segment.

[0007] Preferably, the double-slice preprocessing step in S1 includes: S11. Time-align the video data from multiple monitoring points and take the common time interval of all videos as the research interval. S12. Generate statistical slices with strict time alignment, and slice them at equal intervals with 5-minute intervals. At the same time, extract 3, 2, and 1 additional slices from the upstream monitoring points to match the characteristics of congestion propagating upstream. S13. Randomly generate test slices for manual sampling verification and error correction model training.

[0008] Preferably, the step in S1 of statistically analyzing traffic flow parameters at each monitoring point using machine vision algorithms is as follows: The YOLOv8 object detection algorithm is used to detect vehicle targets in video frames. The Bot-Sort multi-object tracking algorithm is used to track vehicles that appear stably for more than 5 frames. The number of vehicles passing through the monitoring point in each time slice is counted, and traffic flow is calculated. The formula is: ; in, To calculate the duration, To count the total number of vehicles; The average number of vehicles within a 150m effective detection range in each time slice is counted, and the traffic density is calculated using the following formula: ; in, For interval density, This represents the average number of vehicles per frame. The interval travel speed is calculated by tracking the appearance time of characteristic vehicles at upstream and downstream monitoring points.

[0009] Preferably, the manual sampling linear correction algorithm in S1 is as follows: Multiple test slices were randomly selected for manual vehicle statistics. The mean of the manual statistics was taken as the true value. Linear fitting was performed on the machine vision statistics results for each monitoring point to obtain specific correction coefficients. These correction coefficients were then used to correct the results of all statistical slices. The correction formula is as follows: ; in, This represents the actual number of vehicles. To count the number of vehicles using machine vision. This is the slope correction factor. This is the intercept correction factor.

[0010] Preferably, the spatiotemporal density distribution model of road segments constructed based on the traffic flow conservation equation in S2 is as follows: By the law of conservation of mass: ; in, For traffic flow, This represents the change in traffic flow. For time step, Traffic density, This represents the change in traffic density. The spatial step size; The partial differential equation is simplified to the following formula: ; in, for time Traffic density at the location for time Traffic flow at the location; The road sections are divided into Spatial discretization, time according to Discretize and solve for the initial density distribution using the density recursion formula: ; in, For the first The first time step Traffic density in each segment For the first The first time step Traffic density in each segment For the first The first time step Traffic density in each segment For the first The first time step Traffic flow in each segment, No. The first time step Traffic flow in each segment.

[0011] Preferably, the step in S2 of using spatiotemporal kriging interpolation to complete missing data includes: Calculate the value of the variation function The formula is: ; in, Let the distance vector be between two points. For mathematical expectation; Solving for optimal weights The formula is: ; in, The variation function matrix, For Lagrange multipliers, It is the identity matrix; The formula for estimating the missing point density is: ; in, Point to be estimated Density value at, for time The spatiotemporal weighting coefficient of location.

[0012] Preferably, the step in S2 of predicting traffic flow density using an ARMA time-series model that integrates upstream and downstream influences includes: The formula for calculating the traffic density based on ARMA prediction is as follows: ; in, Traffic density predicted based on ARMA. For the first Autoregressive coefficient of order, For the first Moving average coefficient, The summation index of the autoregressive terms. The subscript for the summation of the moving average term. For constant terms, Let the order be the autoregressive order. The moving average order is... for The residual term at time, This is an inherent residual term of the ARMA model; The final predicted density is obtained by integrating the influence of upstream and downstream nodes. The formula is: ; in, The inertia coefficient of the current node. The influence coefficient of the upstream node. The downstream node influence coefficient. This is a correction term for the fusion of upstream and downstream influences.

[0013] Preferably, S3 specifically includes: The formula for verifying lane balance constraints is: ; in, Lane balance coefficient, , The densities of regular lanes and emergency lanes are respectively. , These are the congestion densities for regular lanes and emergency lanes, respectively. The formula for verifying the road density ratio constraint is: ; The formula for verifying traffic flow constraints on road sections is as follows: ; in, Conversion factor for emergency vehicles; , The traffic flow is divided into regular lanes and emergency lanes, respectively. , The number of regular lanes and emergency lanes are respectively. The proportion of vehicles using the dedicated lane to other vehicles. For ordinary lane traffic capacity; Calculate the shortest emergency lane activation length The formula is: ; in, The length of the congested road section To maintain a safe distance; Calculate safe distance The formula is: ; in, The average speed of the road segment. This is the driver's reaction time, typically taken as 1.5 seconds. The average length of the vehicle itself ensures sufficient space between the front and rear vehicles; The extra safety margin distance added to account for uncertainties (such as sudden braking, unexpected actions of other drivers, etc.), is measured in meters (m) and is usually taken as the distance traveled in one second at the vehicle speed. The safe following distance is reserved to prevent rear-end collisions, and the unit is meters (m).

[0014] Calculate lane clearing time : ; The length of the emergency lane in use, Lane changing frequency; ; in, The distance required for a single vehicle to change lanes. The average time required to complete one lane change; The lane clearing time can be derived as follows: .

[0015] Preferably, S4 specifically includes: With the goal of minimizing the total deployment cost and constrained by minimizing the parameter estimation accuracy across the entire road segment, the optimal number and location of new monitoring points are determined, leading to the establishment of a trade-off model between monitoring accuracy and deployment cost: ; in, Total deployment cost, To increase the number of monitoring points, Cost per monitoring point For the first The accuracy of parameter estimation for each road segment. The minimum accuracy threshold, The error coefficient per unit distance. For the first The distance of each road segment from the nearest monitoring point; The total length of the target road segment. For the length of a single segment, To estimate the accuracy of the parameters.

[0016] This invention also provides an optimization system for the temporary activation of emergency lanes for highway congestion warning, which is used to execute the above-mentioned optimization method for the temporary activation of emergency lanes for highway congestion warning, including a video acquisition module, an edge computing module, a cloud warning module, a control center module, and a monitoring optimization module; The video acquisition module collects real-time video data from multiple monitoring points along the target section of the highway and transmits the real-time video data to the edge computing module. The edge computing module performs video double-slice preprocessing, runs the YOLOv8+Bot-Sort algorithm to count vehicles, loads pre-trained correction coefficients to calibrate errors, calculates traffic flow parameters, and uploads the standardized traffic flow parameters to the cloud early warning module; The cloud-based early warning module receives traffic flow data, constructs a spatiotemporal density distribution model of road segments based on the traffic flow conservation equation, uses spatiotemporal Kriging interpolation to fill in missing data, and adopts an ARMA time series model that integrates upstream and downstream influences to predict traffic flow density. When a threshold is triggered, an early warning is generated and the warning information is pushed to the control center module and the historical operation data is synchronized to the monitoring and optimization module. The control center module receives early warnings, automatically verifies triple safety constraints, calculates the shortest activation length and clearing time, monitors traffic flow status in real time, and issues control instructions to on-site control equipment. The monitoring optimization module calculates the parameter estimation error for the entire road segment, substitutes it into a model that balances monitoring accuracy and deployment cost to solve for the optimal monitoring points, generates an upgrade plan, and outputs the monitoring point optimization plan to the operation and maintenance interface of the control center module.

[0017] Therefore, the present invention employs the above-mentioned optimized method and system for temporary activation of emergency lanes for highway congestion early warning, which has the following beneficial effects: (1) By adopting a double-slice preprocessing and manual sampling linear correction mechanism, the statistical error of traffic flow parameters is controlled within 17%, and the data accuracy is significantly improved; (2) By integrating the traffic flow conservation equation with spatiotemporal Kriging interpolation, the precise location of congestion areas with a granularity of 200m is achieved. Combined with the ARMA model that integrates upstream and downstream influences, an early warning of 10 minutes is achieved with an accuracy rate of 92%. (3) Based on the principle of minimum occupancy and triple safety constraints, while ensuring the smooth flow of emergency access, the road traffic efficiency is increased by more than 25%; (4) Establish a monitoring and optimization model that balances cost and accuracy, so as to significantly reduce hardware deployment costs while ensuring decision-making accuracy.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] Figure 1 This is a flowchart of the optimized method for temporary activation of emergency lanes for highway congestion early warning according to the present invention; Figure 2 This is a schematic diagram of a highway cross-section according to an embodiment of the present invention; Figure 3 This is an example diagram of road space discretization according to an embodiment of the present invention; Figure 4 This is a schematic diagram of spatiotemporal kriging interpolation according to an embodiment of the present invention; Figure 5 This is a schematic diagram of traffic flow density prediction based on the ARMA model according to an embodiment of the present invention; Figure 6 This is a flowchart illustrating the emergency lane activation decision-making process according to an embodiment of the present invention. Figure 7 This is a schematic diagram illustrating the upstream and downstream estimation accuracy of a single monitoring point according to an embodiment of the present invention. Figure 8 This is a distribution chart of parameter estimation accuracy without newly added monitoring points in an embodiment of the present invention; Figure 9 This is an accuracy distribution chart for different numbers of newly added monitoring points in an embodiment of the present invention; (a) Add 1 monitoring point; (b) Two new monitoring points were added; (c) Three new monitoring points were added; Figure 10 This is a graph showing the relationship between the minimum estimation accuracy and the number of newly added monitoring points in an embodiment of the present invention. (a) Unit error e = 0.05; (b) Unit error e = 0.075; (c) Unit error e = 0.1. Detailed Implementation

[0020] Example This embodiment takes a 5000m long highway section as the research object. The section includes two driving lanes and one emergency lane, with no entrances or exits. There are four existing video surveillance points, numbered GS01, GS02, GS03 and GS04. The distance between GS01 and GS02 is 1000m, the distance between GS02 and GS03 is 1000m, and the distance between GS03 and GS04 is 3000m.

[0021] like Figure 1 As shown, this embodiment discloses an optimized method for the temporary activation of emergency lanes for highway congestion warning, the steps of which include: S1. Collect real-time video data from multiple video surveillance points along the target section of the highway. GS01.mp4 covers the time period from 11:41:03 to 16:14:31, GS02.mp4 covers the time period from 11:52:27 to 15:28:20, GS03.mp4 covers the time period from 11:35:43 to 16:09:12, and GS04.mp4 covers the time period from 12:56:47 to 15:20:03. Perform double-slice preprocessing on the video data, specifically including: S11. Time alignment is performed on the video data from multiple monitoring points, and the common time interval of all videos, 12:56:47-15:20:03, is taken as the research interval. S12. Generate statistical slices with strictly aligned time, and slice them at equal intervals with 5-minute intervals. At the same time, GS01.mp4 is truncated 3 slices forward, GS02.mp4 is truncated 2 slices forward, and GS03.mp4 is truncated 1 slice forward to match the characteristics of congestion propagating upstream. S13. Randomly generate test slices for manual sampling verification and error correction model training.

[0022] Traffic flow parameters at each monitoring point are statistically analyzed based on machine vision algorithms, specifically: The YOLOv8 target detection algorithm is used to detect vehicle targets in video frames. The Bot-Sort multi-target tracking algorithm is used to track vehicles that appear stably for more than 5 frames, eliminating temporary noise and duplicate counts.

[0023] Ten time slices were randomly selected for manual statistical analysis. The mean of the manual results was taken as the true value. Linear fitting was then performed on the machine vision statistical results to obtain the correction coefficients for each monitoring point. Monitoring point GS01: a=0.46984, b=132.62675, average error 5.01%; GS02 monitoring point: a=0.628, b=45.73556, average error 17.15%; GS03 monitoring point: a=1.02987, b=-1.73435, average error 12.79%; GS04 monitoring point: Due to the large statistical error of machine vision, manual statistics are used.

[0024] Count the number of vehicles passing through the monitoring points in each time slice and calculate the traffic flow. The formula is: ; in, To calculate the duration, To count the total number of vehicles; The average number of vehicles within a 150m effective detection range in each time slice is counted, and the traffic density is calculated using the following formula: ; in, For interval density, This represents the average number of vehicles per frame. The interval travel speed is calculated by tracking the appearance time of characteristic vehicles at upstream and downstream monitoring points.

[0025] The statistical error is corrected using a linear correction algorithm based on manual sampling, specifically as follows: To address inherent errors such as occlusion of monitoring viewpoints, missed detection of small targets, and ID jumps, a manual sampling linear correction mechanism is introduced. Multiple test slices are randomly selected for manual vehicle statistics. The mean of the manual statistics is taken as the true value. Linear fitting is performed on the machine vision statistics results for each monitoring point to obtain a specific correction coefficient. This correction coefficient is then used to correct the results of all statistical slices. The correction formula is as follows: ; in, This represents the actual number of vehicles. To count the number of vehicles using machine vision. This is the slope correction factor. This is the intercept correction factor.

[0026] S2. Construct a spatiotemporal density distribution model for road segments based on the traffic flow conservation equation, specifically: By the law of conservation of mass: ; in, For traffic flow, This represents the change in traffic flow. For time step, Traffic density, This represents the change in traffic density. The spatial step size; The partial differential equation is simplified to the following formula: ; in, for time Traffic density at the location for time Traffic flow at the location; To achieve proactive congestion prediction and precise location, the 3000m section between GS03 and GS04 will be divided into... Spatial discretization, time according to Discretization, as shown in the schematic diagram of the highway cross-section in this embodiment, as follows: Figure 2 As shown.

[0027] Solve for the initial density distribution using the density recursion formula: ; in, For the first The first time step Traffic density in each segment For the first The first time step Traffic density in each segment For the first The first time step Traffic density in each segment For the first The first time step Traffic flow in each segment, No. The first time step Traffic flow in each segment.

[0028] This embodiment presents an example diagram of road space discretization, as shown below. Figure 3 As shown.

[0029] A 3×3 sliding window is used to complete missing data using spatiotemporal kriging interpolation, specifically including: Calculate the value of the variation function The formula is: ; in, Let the distance vector be between two points. For mathematical expectation; Solving for optimal weights The formula is: ; in, The variation function matrix, For Lagrange multipliers, It is the identity matrix; The formula for estimating the missing point density is: ; in, Point to be estimated Density value at, for time The spatiotemporal weighting coefficients of location. A schematic diagram of spatiotemporal kriging interpolation in this embodiment is shown below. Figure 4 As shown.

[0030] An ARMA time-series model incorporating upstream and downstream influences is used to predict traffic density in each segment of the road within a preset time period, specifically including: The formula for calculating the traffic density based on ARMA prediction is as follows: ; in, Traffic density predicted based on ARMA. For the first Autoregressive coefficient of order, For the first Moving average coefficient, The summation index of the autoregressive terms. The subscript for the summation of the moving average term. For constant terms, Let the order be the autoregressive order. The moving average order is... for The residual term at time, This is an inherent residual term of the ARMA model; The final predicted density is obtained by integrating the influence of upstream and downstream nodes. The formula is: ; in, The inertia coefficient of the current node. The influence coefficient of the upstream node. The downstream node influence coefficient. This includes a correction term to account for the fusion of upstream and downstream influences. A schematic diagram of traffic density prediction based on the ARMA model is shown below. Figure 5 As shown.

[0031] The congestion density threshold is set at 66 veh / km. When the density of three consecutive time slices in the 15th segment (2800-3000m downstream of GS03) exceeds 66 veh / km during the predicted period of 13:50-14:00, a continuous congestion warning is issued 10 minutes in advance, with a warning accuracy rate of 92%. This achieves congestion area location and continuous congestion warning with a preset spatial granularity.

[0032] S3. Based on the early warning results, the congested area is determined to be 2600-3000m downstream of GS03, with a congested section length of [missing information]. =400m. Based on the congestion area location results and triple safety constraints, the shortest emergency lane activation length is calculated and activation decisions are made. The safe closure timing is determined based on the lane clearing time; specifically including: Ordinary lane density =62veh / km, predicted density of emergency lanes =18veh / km, congestion density of ordinary lanes =66veh / km, emergency lane congestion density =66veh / km; ordinary lane flow rate =2800veh / h, emergency lane emergency vehicle flow rate =5veh / h, number of ordinary lanes =2, Number of emergency lanes =1. The percentage of vehicles using the dedicated lane relative to other vehicles using the lane is set at 0.8. This is for the capacity of ordinary lanes.

[0033] The formula for verifying lane balance constraints is: in, The lane balance coefficient is set to 0.2, and the constraint conditions are verified to be met. , The densities of regular lanes and emergency lanes are respectively. , These are the congestion densities for regular lanes and emergency lanes, respectively. The formula for verifying the road density ratio constraint is: ; Verify that the constraints are met.

[0034] The formula for verifying traffic flow constraints on road sections is as follows: ; in, The conversion factor for emergency vehicles is set to 1.5; the constraints are verified to be met.

[0035] Calculate the shortest emergency lane activation length The formula is: ; in, The length of the congested road section To maintain a safe distance; Calculate safe distance The formula is: ; in, =10km / h is the average speed of the road segment This represents the driver's reaction time, set to 1.5 seconds. =5m is the average length of the vehicle itself, ensuring sufficient space between the front and rear vehicles; =2.8m is an additional safety margin distance added to account for uncertainties (such as sudden braking, unexpected actions of other drivers, etc.), and the unit is meters; =2m is the safety following buffer distance reserved to prevent rear-end collisions. Substituting this into the calculation, the one-way safety distance is 14m.

[0036] Therefore, the shortest emergency lane activation length is: .

[0037] Calculate lane clearing time : ; The length of the emergency lane in use, Lane changing frequency; ; in, This represents the distance required for a single vehicle to change lanes, and is set to 100m / veh. The average time required to complete one lane change is set to 4 seconds. The lane clearing time can be derived as follows: ; Substituting the values ​​into the calculation, the lane clearing time is 7.2 minutes.

[0038] If all three constraints are met, the emergency lane can be activated. The decision-making process for activating the emergency lane is as follows: Figure 6 As shown. After activation, traffic flow is monitored in real time. At 14:20, the density of ordinary lanes dropped below 40 veh / km, and the emergency lane closure procedure was initiated. After waiting 7.2 minutes for the lanes to clear, the emergency lane was officially closed at 14:27.

[0039] The simulation was verified using Vissim, with the scenario set as passenger cars: large trucks: buses = 15:1:1 and the probability of congestion and lane-changing being 10%. The results showed that after the emergency lane was activated, the average travel time for vehicles on the GS03-GS04 section was reduced from 45 minutes to 33 minutes, and the traffic efficiency was improved by 26.7%.

[0040] S4. Establish a trade-off model between monitoring accuracy and deployment cost, with the goal of minimizing total deployment cost and the constraint of minimizing parameter estimation accuracy across the entire road segment, to optimize the deployment of video surveillance points on the target road segment; specifically including: setting the cost of a single monitoring point. =10,000 yuan, minimum accuracy threshold =90%, unit distance error coefficient =0.05 / 200m.

[0041] With the goal of minimizing the total deployment cost and constrained by minimizing the parameter estimation accuracy across the entire road segment, the optimal number and location of new monitoring points are determined, leading to the establishment of a trade-off model between monitoring accuracy and deployment cost: ; in, Total deployment cost, To increase the number of monitoring points, Cost per monitoring point For the first The accuracy of parameter estimation for each road segment. The minimum accuracy threshold, The error coefficient per unit distance. For the first The distance of each road segment from the nearest monitoring point; The total length of the target road segment. For the length of a single segment, To estimate the accuracy of the parameters.

[0042] Quantitative analysis reveals a linear increase in traffic flow parameter estimation error with distance from monitoring points. The distribution of estimation accuracy upstream and downstream of a single monitoring point is shown below. Figure 7 As shown; the accuracy distribution without additional monitoring points exhibits a "V" shape, as... Figure 8 As shown in the figure. With the goal of minimizing the total deployment cost and the constraint of achieving a minimum parameter estimation accuracy of ≥90% across the entire road segment, the optimal number and location of new monitoring points are determined. For a 3km downstream segment, the optimal solution is to uniformly add 3 monitoring points, evenly distributed at 750m, 1500m, and 2250m downstream of GS03. The accuracy distribution at this point is as follows. Figure 9 As shown in the figure, this solution can improve the minimum estimation accuracy of the entire road segment from 70% to 90%, while reducing the total deployment cost by more than 80% compared to dense deployment across the entire road segment, with a total deployment cost of 30,000 yuan. The minimum accuracy variation curves under different numbers of monitoring points are shown in the figure. Figure 10 As shown.

[0043] This invention also discloses an emergency lane temporary activation optimization system for highway congestion warning, used to execute the above-mentioned emergency lane temporary activation optimization method for highway congestion warning, including a video acquisition module, an edge computing module, a cloud warning module, a control center module, and a monitoring optimization module; The video acquisition module collects real-time video data from multiple monitoring points along the target section of the highway and transmits the real-time video data to the edge computing module. In this embodiment, the video acquisition module deploys four 1280×720 / 25fps high-definition cameras, which are installed on the top of the monitoring poles along the road. They are equipped with local storage and infrared illumination and are directly connected to the edge computing module via gigabit fiber optic cable. The module can collect video 24 / 7 and cache 7 days of data locally for backtracking. The compressed video stream is then transmitted to the edge for processing.

[0044] The edge computing module performs video double-slice preprocessing, runs the YOLOv8+Bot-Sort algorithm to count vehicles, loads pre-trained correction coefficients to calibrate errors, calculates traffic flow parameters, and uploads the standardized traffic flow parameters to the cloud early warning module. In this embodiment, the edge computing module is equipped with an industrial-grade edge server and installed in a gantry rack. It connects to the cloud through a 5G industrial gateway, adopts the YOLOv8n lightweight model, uploads standardized JSON data every 5 minutes, caches it locally when the network is down, and automatically re-uploads it after the network is restored.

[0045] The cloud-based early warning module receives traffic flow data, constructs a spatiotemporal density distribution model of road segments based on the traffic flow conservation equation, uses spatiotemporal Kriging interpolation to complete missing data, and employs an ARMA time-series model that integrates upstream and downstream influences to predict traffic flow density. When a threshold is triggered, an early warning is generated and pushed to the control center module. Historical operational data is synchronized to the monitoring and optimization module. In this embodiment, the cloud-based early warning module is deployed using a cloud server cluster, configured with load balancing and disaster recovery backup, establishes a long TCP connection with the edge, pushes early warnings via the MQTT protocol, and uses an ARMA model to predict traffic flow density for the next 10 minutes, generating accurate early warning information including congestion location and duration.

[0046] The control center module receives early warnings, automatically verifies the triple safety constraints, calculates the shortest activation length and clearing time, monitors traffic flow status in real time, and issues control instructions to on-site control equipment. In this embodiment, the control center module is deployed in the operation management center, using a local server cluster equipped with dual-screen terminals and audible and visual alarms. It connects to on-site variable message signs, guidance lights, and law enforcement terminals through a high-speed dedicated communication network, automatically verifies the triple constraints of lane balance, road density ratio, and road segment traffic flow, and records control operation logs in real time.

[0047] The monitoring optimization module statistically analyzes the parameter estimation error across the entire road segment, substitutes it into a model that balances monitoring accuracy and deployment cost to solve for the optimal monitoring points, generates an upgrade plan, and outputs the optimized monitoring point plan to the operation and maintenance interface of the control center module. In this embodiment, the monitoring optimization module and the cloud-based early warning module share cloud resources, statistically analyze the parameter estimation error across the entire road segment monthly, and use the minimum total deployment cost as the objective and the minimum parameter estimation accuracy across the entire road segment as a constraint to solve for the optimal points, outputting a visualized upgrade plan to provide a decision-making basis for the upgrading and transformation of the monitoring system.

[0048] Therefore, this invention adopts the above-mentioned optimization method and system for the temporary activation of emergency lanes for highway congestion early warning. It can achieve accurate statistics of traffic flow parameters by combining dual-slice video preprocessing with YOLOv8+Bot-Sort algorithm and manual sampling linear correction. It can achieve refined congestion early warning based on traffic flow conservation equation, spatiotemporal Kriging interpolation and ARMA model that integrates upstream and downstream. It establishes a triple safety constraint and lane clearing mechanism and optimizes monitoring points to solve the problems of blind opening of emergency lanes, delayed decision-making and unreasonable closing timing. Under the premise of ensuring the smooth flow of rescue channels, it can effectively improve road traffic efficiency.

[0049] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An optimized method for the temporary activation of emergency lanes in response to highway congestion warnings, characterized by the following steps: include: S1. Collect real-time video data from multiple video monitoring points on the target section of the highway. After double-slice preprocessing of the video data, statistical traffic flow parameters of each monitoring point are calculated based on machine vision algorithms, and statistical errors are corrected by manual sampling linear correction algorithm. S2. Based on the traffic flow conservation equation, a spatiotemporal density distribution model of road segments is constructed. Spatiotemporal Kriging interpolation is used to fill in missing data. An ARMA time series model that integrates upstream and downstream influences is used to predict the traffic flow density of each segment of the road segment within a preset time period in the future, so as to realize the location of congestion areas and the early warning of continuous congestion at a preset spatial granularity. S3. Based on the location results of the congested area and the triple safety constraints, calculate the shortest emergency lane activation length and execute the activation decision, and determine the safe closure time based on the lane clearing time. S4. Establish a trade-off model between monitoring accuracy and deployment cost, with the goal of minimizing total deployment cost and the constraint of minimizing the accuracy of parameter estimation for the entire road segment, to optimize the deployment of video surveillance points on the target road segment.

2. The optimized method for temporary activation of emergency lanes for highway congestion early warning as described in claim 1, characterized in that, The steps for double-slice preprocessing in S1 include: S11. Time-align the video data from multiple monitoring points and take the common time interval of all videos as the research interval. S12. Generate statistical slices with strict time alignment, and slice them at equal intervals with 5-minute intervals. At the same time, extract 3, 2, and 1 additional slices from the upstream monitoring points to match the characteristics of congestion propagating upstream. S13. Randomly generate test slices for manual sampling verification and error correction model training.

3. The optimized method for temporary activation of emergency lanes for highway congestion early warning as described in claim 1, characterized in that, The specific steps in S1 for calculating traffic flow parameters at each monitoring point based on machine vision algorithms are as follows: The YOLOv8 object detection algorithm is used to detect vehicle targets in video frames. The Bot-Sort multi-object tracking algorithm is used to track vehicles that appear stably for more than 5 frames. The number of vehicles passing through the monitoring point in each time slice is counted, and traffic flow is calculated. The formula is: ； in, To calculate the duration, To count the total number of vehicles; The average number of vehicles within a 150m effective detection range in each time slice is counted, and the traffic density is calculated using the following formula: ； in, For interval density, This represents the average number of vehicles per frame. The interval travel speed is calculated by tracking the appearance time of characteristic vehicles at upstream and downstream monitoring points.

4. The optimized method for temporary activation of emergency lanes for highway congestion early warning as described in claim 1, characterized in that, The specific linear correction algorithm for manual sampling in S1 is as follows: Multiple test slices were randomly selected for manual vehicle statistics. The mean of the manual statistics was taken as the true value. Linear fitting was performed on the machine vision statistics results for each monitoring point to obtain specific correction coefficients. These correction coefficients were then used to correct the results of all statistical slices. The correction formula is as follows: ； in, This represents the actual number of vehicles. To count the number of vehicles using machine vision. This is the slope correction factor. This is the intercept correction factor.

5. The optimized method for temporary activation of emergency lanes for highway congestion early warning as described in claim 1, characterized in that, The specific spatiotemporal density distribution model of road segments constructed based on the traffic flow conservation equation in S2 is as follows: By the law of conservation of mass: ； in, For traffic flow, This represents the change in traffic flow. For time step, Traffic density, This represents the change in traffic density. The spatial step size; Simplifying, we obtain the partial differential equation, the formula is: ； in, for time Traffic density at the location for time Traffic flow at the location; The road sections are divided into Spatial discretization, time according to Discretize and solve for the initial density distribution using the density recursion formula: ； in, For the first The first time step Traffic density in each segment For the first The first time step Traffic density in each segment For the first The first time step Traffic density in each segment For the first The first time step Traffic flow in each segment, No. The first time step Traffic flow in each segment.

6. The optimized method for temporary activation of emergency lanes for highway congestion early warning as described in claim 5, characterized in that, The steps in S2 to complete missing data using spatiotemporal kriging interpolation include: Calculate the value of the variation function The formula is: ； in, Let the distance vector be between two points. For mathematical expectation; Solving for optimal weights The formula is: ； in, The variation function matrix, For Lagrange multipliers, It is the identity matrix; The formula for estimating the missing point density is: ； in, Point to be estimated Density value at that location for time The spatiotemporal weighting coefficient of location.

7. The optimized method for temporary activation of emergency lanes for highway congestion early warning as described in claim 6, characterized in that, The steps in S2 for predicting traffic flow density using an ARMA time-series model that integrates upstream and downstream influences include: The formula for calculating the traffic density based on ARMA prediction is as follows: ； in, Traffic density predicted based on ARMA. For the first Autoregressive coefficient of order, For the first Moving average coefficient, The summation index of the autoregressive terms. The subscript for the summation of the moving average term. For constant terms, Let the order be the autoregressive order. The moving average order is... for The residual term at time, This is an inherent residual term of the ARMA model; The final predicted density is obtained by integrating the influence of upstream and downstream nodes. The formula is: ； in, The inertia coefficient of the current node. The influence coefficient of the upstream node. The downstream node influence coefficient. This is a correction term for the fusion of upstream and downstream influences.

8. The optimized method for temporary activation of emergency lanes for highway congestion early warning as described in claim 1, characterized in that, S3 specifically includes: The formula for verifying lane balance constraints is: ； in, Lane balance coefficient, , The density of regular lanes and emergency lanes are respectively. , These are the congestion densities for regular lanes and emergency lanes, respectively. The formula for verifying the road density ratio constraint is: ； The formula for verifying traffic flow constraints on road sections is as follows: ； in, Conversion factor for emergency vehicles; , The traffic flow is divided into regular lanes and emergency lanes, respectively. , The number of regular lanes and emergency lanes are respectively. The proportion of vehicles using the dedicated lane to other vehicles. For ordinary lane traffic capacity; Calculate the shortest emergency lane activation length The formula is: ； in, The length of the congested road section To maintain a safe distance; Calculate safe distance The formula is: ； in, The average speed of the road segment. For driver reaction time, The average length of the vehicle itself. Additional safety margin distance added to account for uncertainties; A safe following distance is provided to prevent rear-end collisions; Calculate lane clearing time : ； in, The distance required for a single vehicle to change lanes. The average time required to complete one lane change.

9. The optimized method for temporary activation of emergency lanes for highway congestion early warning as described in claim 1, characterized in that, S4 specifically includes: With the goal of minimizing the total deployment cost and constrained by minimizing the parameter estimation accuracy across the entire road segment, the optimal number and location of new monitoring points are determined, leading to the establishment of a trade-off model between monitoring accuracy and deployment cost: ； in, Total deployment cost To increase the number of monitoring points, Cost per monitoring point For the first The accuracy of parameter estimation for each road segment. The minimum accuracy threshold, The error coefficient per unit distance. For the first The distance of each road segment from the nearest monitoring point; The total length of the target road segment. For the length of a single segment, To estimate the accuracy of the parameters.

10. An optimization system for the temporary activation of emergency lanes for highway congestion warning, used to execute the optimization method for the temporary activation of emergency lanes for highway congestion warning as described in any one of claims 1-9, characterized in that: It includes a video acquisition module, an edge computing module, a cloud-based early warning module, a control center module, and a monitoring optimization module; The video acquisition module collects real-time video data from multiple monitoring points along the target section of the highway and transmits the real-time video data to the edge computing module. The edge computing module performs video double-slice preprocessing, runs the YOLOv8+Bot-Sort algorithm to count vehicles, loads pre-trained correction coefficients to calibrate errors, calculates traffic flow parameters, and uploads the standardized traffic flow parameters to the cloud early warning module; The cloud-based early warning module receives traffic flow data, constructs a spatiotemporal density distribution model of road segments based on the traffic flow conservation equation, uses spatiotemporal Kriging interpolation to fill in missing data, and adopts an ARMA time series model that integrates upstream and downstream influences to predict traffic flow density. When a threshold is triggered, an early warning is generated and the warning information is pushed to the control center module and the historical operation data is synchronized to the monitoring and optimization module. The control center module receives early warnings, automatically verifies triple safety constraints, calculates the shortest activation length and clearing time, monitors traffic flow status in real time, and issues control instructions to on-site control equipment. The monitoring optimization module calculates the parameter estimation error for the entire road segment, substitutes it into a model that balances monitoring accuracy and deployment cost to solve for the optimal monitoring points, generates an upgrade plan, and outputs the monitoring point optimization plan to the operation and maintenance interface of the control center module.

Images