A dynamic identification method for ice and snow low-temperature coupling risk of "two passengers and one dangerous" vehicles in a seasonal freezing region

Abstract

The application relates to a seasonal freezing region 'two passengers and one dangerous' vehicle ice and snow low-temperature coupling risk dynamic identification method, which realizes real-time acquisition of driving parameters of 'two passengers and one dangerous' vehicles based on a Beidou high-precision positioning system; constructs a road surface ice and snow state space-time distribution model by fusing data of road side ice and snow sensors and road surface temperature sensors; collects driving behavior characteristic parameters through a vehicle-mounted CAN bus; establishes a multi-source heterogeneous data space-time coupling model of ice and snow low-temperature-road surface state-driving parameter-driving behavior; constructs a seasonal freezing region exclusive coupling risk identification model based on a dynamic Bayesian network and fuzzy comprehensive judgment, and realizes real-time dynamic identification and early warning of risk grades. The application firstly carries out space-time coupling modeling on a road surface icing state, air temperature, snow thickness and Beidou driving parameters and driving behavior, effectively solves the problems of risk false reports and missed reports of ice and snow road sections in the seasonal freezing region in winter, and fills the technical blank of ice and snow coupling risk identification of 'two passengers and one dangerous' vehicles in the seasonal freezing region.

Description

Technical Field

[0001] This invention relates to the field of road traffic safety and intelligent transportation, specifically to a dynamic identification method for ice and snow low-temperature coupling risks of passenger vehicles and dangerous goods vehicles in seasonally frozen areas. Background Technology

[0002] my country's seasonally frozen zones are widely distributed in the Northeast, Northwest, and North China regions. Winter road icing and snow cover are frequent occurrences, making these areas high-risk for road traffic accidents. "Two-passenger-one-dangerous-goods" vehicles, due to their large passenger capacity or the high risk of transporting goods, often cause serious consequences such as mass casualties or environmental pollution in the event of an accident.

[0003] Existing risk identification methods for passenger vehicles and hazardous material transport vehicles mainly rely on single driving parameters such as speeding and driver fatigue for risk assessment, which has the following shortcomings: First, they do not consider the impact of actual road surface iciness and snow conditions on driving safety, leading to missed risks for vehicles traveling at normal speeds on icy and snowy roads because they are not speeding; Second, they do not couple environmental factors such as temperature and snow depth with vehicle driving conditions, making it difficult to identify complex risks in icy and snowy low-temperature environments; Third, they lack the ability to identify risk characteristics unique to icy and snowy roads, such as increased braking distance and tendency to skid, and cannot accurately reflect the actual risk level in seasonally frozen areas; Fourth, existing methods suffer from serious false alarms and missed alarms on icy and snowy road sections in winter. The high false alarm rate leads to a complacency effect among management personnel, while missed alarms may lead to major accidents.

[0004] Therefore, there is an urgent need for a multi-dimensional coupled risk identification method that can integrate low-temperature environmental parameters of ice and snow, road surface condition parameters, and vehicle driving parameters to solve the safety supervision problem of "passenger and dangerous goods" vehicles operating in winter in seasonally frozen areas. Summary of the Invention

[0005] To address the problems encountered in existing technologies, this invention provides a method for dynamic identification of ice and snow low-temperature coupled risks for passenger and hazardous goods vehicles in seasonally frozen areas. To achieve the aforementioned objectives, the technical solution adopted by this invention is as follows:

[0006] S1. Based on the BeiDou high-precision positioning system, the driving status parameters of "two passengers and one dangerous goods" vehicles are obtained in real time, including latitude and longitude coordinates, instantaneous speed, heading angle, acceleration and braking distance, and the raw data is preprocessed by Kalman filtering.

[0007] S2. By integrating data from roadside ice and snow sensors and road surface temperature sensors, and combining real-time temperature and snowfall information from meteorological stations, a spatiotemporal distribution model of road surface ice and snow conditions is constructed to quantify the attenuation characteristics of road surface friction coefficient.

[0008] S3. Collect driving behavior characteristic parameters through the vehicle CAN bus and extract key indicators such as emergency braking frequency, steering wheel angle change rate, lane departure, yaw rate and lateral acceleration.

[0009] S4. Establish a spatiotemporal coupling model of multi-source heterogeneous data, including ice and snow low-temperature environment parameters, road surface condition parameters, BeiDou driving parameters, and driving behavior parameters, and achieve multi-dimensional data fusion through timestamp alignment and spatial matching.

[0010] S5. Based on dynamic Bayesian networks and fuzzy comprehensive evaluation methods, a dynamic identification model for the coupled risk of ice and snow low temperature in seasonally frozen areas is constructed, which outputs the risk level and triggers corresponding early warning strategies.

[0011] Furthermore, the specific steps of step S1 are as follows:

[0012] S1-1, Acquisition of BeiDou high-precision positioning data. Utilizing the BeiDou-3 system's RTK positioning mode, the vehicle's geocentric and geofixed coordinates are acquired in real time. The positioning accuracy is better than 0.1m, and the update frequency is not less than 10Hz;

[0013] S1-2. Coordinate Transformation. Converting Earth-centered and Earth-fixed coordinates to geodetic coordinates. :

[0014]

[0015]

[0016] in, , , For the Earth's semi-major axis, For the Earth's minor axis, For the first eccentricity, The second eccentricity;

[0017] S1-3, Calculation of Driving Speed ​​and Acceleration. Based on the position information from two consecutive epochs, calculate the instantaneous velocity. :

[0018]

[0019] in, , , The average radius of the Earth Sampling interval. Longitudinal acceleration. Obtained from velocity difference:

[0020]

[0021] S1-4, Braking Distance Calculation. When vehicle longitudinal acceleration is detected... ( When the braking acceleration threshold is reached, it is determined to be a braking behavior, and the braking distance is... for:

[0022]

[0023] in, The braking start time, The moment the vehicle stops;

[0024] S1-5, Kalman filter preprocessing. Establish the vehicle motion state equation and observation equation, and perform filtering on the raw BeiDou positioning data:

[0025]

[0026]

[0027] in, For state vectors, Here is the state transition matrix. For the observation matrix, and The process noise and observation noise are respectively assumed to be zero-mean Gaussian white noise.

[0028] Furthermore, the specific steps of step S2 are as follows:

[0029] S2-1, Road Surface Temperature Acquisition and Freeze Determination. Road surface temperature is acquired using a road surface temperature sensor. relative humidity is obtained by combining road surface humidity sensors. The road surface is considered to be in an icy state when the following conditions are met:

[0030]

[0031] in, The road surface freezing temperature threshold (taken as -2°C) is used. The humidity threshold is set to 85%.

[0032] S2-2, Snow depth monitoring. Snow depth is obtained through roadside ice and snow sensors. And establish a snow cover level classification model:

[0033]

[0034] S2-3, Road Surface Friction Coefficient Attenuation Model. A road surface friction coefficient model is established under the dual influence of temperature and snow accumulation:

[0035]

[0036] in, This is the standard friction coefficient for dry road surfaces. Temperature decay coefficient. for:

[0037]

[0038] Snow attenuation coefficient for:

[0039]

[0040] when hour ;when mm ;when mm This conforms to the physical law that the thicker the snow, the lower the coefficient of friction;

[0041] S2-4. Spatiotemporal distribution model of road surface condition. Based on the sensor network deployed along the route, the spatiotemporal distribution field of road surface ice and snow condition is constructed using ordinary kriging interpolation:

[0042]

[0043] in, The pavement condition value of the point to be estimated. For the first Observations from each sensor The Kriging weights satisfy the unbiased constraint. The weighting coefficients are obtained by solving the Kriging equations.

[0044] Furthermore, the specific steps of step S3 are as follows:

[0045] S3-1, CAN Bus Data Acquisition. Connect to the vehicle's CAN bus via the OBD-II interface to read the following parameters in real time: brake pedal opening. Steering wheel angle yaw rate Wheel speed of each wheel ( and engine speed The sampling frequency is not less than 50Hz;

[0046] S3-2, Emergency Braking Behavior Recognition. When the braking deceleration exceeds a set threshold and the brake pedal opening exceeds the threshold, it is determined as an emergency braking event.

[0047]

[0048] in, m / s² is the deceleration threshold for emergency braking on icy and snowy roads. This represents the brake pedal opening threshold. Within the slip time window... The frequency of emergency braking within 300 seconds is as follows:

[0049]

[0050] S3-3, Lane Departure Calculation. Based on the BeiDou positioning trajectory and the lane centerline on a high-precision map, calculate the vertical distance from the vehicle's current position to the nearest lane centerline. :

[0051]

[0052] in, The vehicle's current location. The coordinates of the nearest projected point on the lane centerline. The heading angle of the lane centerline at that point. is the average radius of the Earth. When ( A distance of 0.6m is considered a lane departure event.

[0053] S3-4, Sideslip Risk Indicators. Based on the deviation between the actual and ideal yaw rates and lateral acceleration, a normalized sideslip risk index is constructed. :

[0054]

[0055] The ideal yaw rate is calculated based on a linear two-degree-of-freedom vehicle model:

[0056]

[0057] This refers to the vehicle's wheelbase. For insufficient steering gradient, It is lateral acceleration. and These are the upper safety limits for yaw rate and lateral acceleration, respectively.

[0058] S3-5, Comprehensive Feature Vector of Driving Behavior. The above parameters are used to construct a driving behavior feature vector:

[0059]

[0060] in, The rate of change of steering wheel angle. This represents the standard deviation of velocity within the time window.

[0061] Furthermore, the specific steps of step S4 are as follows:

[0062] S4-1. Multi-source data time alignment. Using the BeiDou Time (BDT) timestamp of the BeiDou positioning data as the reference clock, time alignment is performed on the road sensor data and CAN bus data respectively:

[0063]

[0064]

[0065] in, and These are the communication transmission delay compensation amounts for the road surface sensor and the CAN bus, respectively.

[0066] S4-2, Spatial Matching. Spatial matching is performed between the vehicle's real-time BeiDou positioning and the positions of various roadside sensors. When the vehicle is in contact with the... When the ground distance of a sensor is less than the matching radius, the data from that sensor is associated with the vehicle.

[0067]

[0068] in, The distance is spherical. m is the spatial matching radius;

[0069] S4-3, Construction of Multidimensional Coupled Feature Vectors. The spatiotemporally aligned multi-source data is fused into a 12-dimensional coupled feature vector:

[0070]

[0071] S4-4, Feature Normalization. The Min-Max normalization method is used to map the features of each dimension to... Interval:

[0072]

[0073] in, For the first Original values ​​of dimensional features and These are the minimum and maximum values ​​of the feature in the historical samples, respectively.

[0074] Furthermore, the specific steps of step S5 are as follows:

[0075] S5-1, Fuzzy Quantization of Risk Factors. A triangular fuzzy membership function is established for each dimension of the factors in the normalized coupled feature vector. Taking the road surface friction coefficient as an example... For example, three fuzzy sets are established: "low" (high risk), "medium", and "high" (low risk):

[0076]

[0077]

[0078]

[0079] The above three membership functions satisfy the following conditions on the entire domain: This ensures the completeness of the fuzzy classification.

[0080] S5-2, Dynamic Bayesian Network Construction. Establishment based on risk level... (Values) A dynamic Bayesian network with target nodes (corresponding to level four risk). The network contains three layers of evidence nodes: the environment layer (...). ), train level ( ) and driving behavior layer ( Time slice The posterior probability of the risk is calculated using Bayes' theorem:

[0081]

[0082] in, Let be the likelihood probability of the current evidence. The risk level represents the state transition probability, and the transition probability matrix is ​​obtained through statistical analysis of historical accident data.

[0083] S5-3. Comprehensive Assessment of Coupling Risk. A coupling risk index for ice and snow at low temperatures is constructed by combining fuzzy assessment and Bayesian inference results. :

[0084]

[0085] in, Scoring environmental risk factors To score the driving risk factors, Scoring of driving behavior risk factors This is a multi-factor coupling enhancement term. Weighting coefficients. Determined using the Analytic Hierarchy Process (AHP), satisfying... .

[0086] The coupling enhancement term is calculated as follows:

[0087]

[0088] in, For coupling enhancement coefficient, and These are the upper limit normalized reference values ​​for emergency braking frequency and sideslip index, respectively. This item applies only when the road surface is icy ( It is activated when there is sudden braking or skidding, reflecting the synergistic risk amplification effect when icy and snowy roads and abnormal driving behavior occur simultaneously.

[0089] S5-4, Risk Level Classification and Early Warning. Based on the coupled risk index. The risk level is divided into four levels:

[0090]

[0091] When the risk level reaches Level III or above, the system will simultaneously push early warning information to the transportation company's monitoring platform, driver's terminal, and traffic management department.

[0092] This invention is the first to perform spatiotemporal coupling modeling of road surface icing, temperature, snow thickness, BeiDou driving parameters, and driving behavior, effectively solving the problem of false alarms and missed alarms of risks on icy and snowy road sections in seasonally frozen areas during winter, and filling the technical gap in the identification of ice and snow coupled risks of "passenger and dangerous goods" vehicles in seasonally frozen areas. Attached Figure Description

[0093] A more complete understanding of the invention and a clearer appreciation of its accompanying advantages and technical features will be achieved by referring to the accompanying drawings and the detailed description that follows. The drawings are described below:

[0094] Figure 1 This is a schematic diagram of the overall process of the present invention. Detailed Implementation

[0095] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0096] To address the problems existing in the prior art, this invention provides a method for dynamic identification of ice and snow low-temperature coupling risks for passenger and hazardous goods vehicles in seasonally frozen areas. The invention will be described in detail below with reference to the accompanying drawings.

[0097] like Figure 1 As shown in the figure, the method for dynamic identification of ice and snow low-temperature coupling risk of "passenger and dangerous goods" vehicles in seasonally frozen areas provided by the present invention includes the following steps:

[0098] S1. Based on the BeiDou high-precision positioning system, the driving status parameters of "two passengers and one dangerous goods" vehicles are obtained in real time, including latitude and longitude coordinates, instantaneous speed, heading angle, acceleration and braking distance, and the raw data is preprocessed by Kalman filtering.

[0099] S2. By integrating data from roadside ice and snow sensors and road surface temperature sensors, and combining real-time temperature and snowfall information from meteorological stations, a spatiotemporal distribution model of road surface ice and snow conditions is constructed to quantify the attenuation characteristics of road surface friction coefficient.

[0100] S3. Collect driving behavior characteristic parameters through the vehicle CAN bus and extract key indicators such as emergency braking frequency, steering wheel angle change rate, lane departure, yaw rate and lateral acceleration.

[0101] S4. Establish a spatiotemporal coupling model of multi-source heterogeneous data, including ice and snow low-temperature environment parameters, road surface condition parameters, BeiDou driving parameters, and driving behavior parameters, and achieve multi-dimensional data fusion through timestamp alignment and spatial matching.

[0102] S5. Based on dynamic Bayesian networks and fuzzy comprehensive evaluation methods, a dynamic identification model for the coupled risk of ice and snow low temperature in seasonally frozen areas is constructed, which outputs the risk level and triggers corresponding early warning strategies.

[0103] The specific method for step S1 is as follows:

[0104] S1-1, Acquisition of BeiDou high-precision positioning data. Utilizing the BeiDou-3 system's RTK positioning mode, the vehicle's geocentric and geofixed coordinates are acquired in real time. The positioning accuracy is better than 0.1m, and the update frequency is not less than 10Hz;

[0105] S1-2. Coordinate Transformation. Converting Earth-centered and Earth-fixed coordinates to geodetic coordinates. :

[0106]

[0107]

[0108] in, , , For the Earth's semi-major axis, For the Earth's minor axis, For the first eccentricity, The second eccentricity;

[0109] S1-3, Calculation of Driving Speed ​​and Acceleration. Based on the position information from two consecutive epochs, calculate the instantaneous velocity. :

[0110]

[0111] in, , , The average radius of the Earth Sampling interval. Longitudinal acceleration. Obtained from velocity difference:

[0112]

[0113] S1-4, Braking Distance Calculation. When vehicle longitudinal acceleration is detected... ( When the braking acceleration threshold is reached, it is determined to be a braking behavior, and the braking distance is... for:

[0114]

[0115] in, The braking start time, The moment the vehicle stops;

[0116] S1-5, Kalman filter preprocessing. Establish the vehicle motion state equation and observation equation, and perform filtering on the raw BeiDou positioning data:

[0117]

[0118]

[0119] in, For state vectors, Here is the state transition matrix. For the observation matrix, and The process noise and observation noise are respectively assumed to be zero-mean Gaussian white noise.

[0120] The specific method for step S2 is as follows:

[0121] S2-1, Road Surface Temperature Acquisition and Freeze Determination. Road surface temperature is acquired using a road surface temperature sensor. relative humidity is obtained by combining road surface humidity sensors. The road surface is considered to be in an icy state when the following conditions are met:

[0122]

[0123] in, The road surface freezing temperature threshold (taken as -2°C) is used. The humidity threshold is set to 85%.

[0124] S2-2, Snow depth monitoring. Snow depth is obtained through roadside ice and snow sensors. And establish a snow cover level classification model:

[0125]

[0126] S2-3, Road Surface Friction Coefficient Attenuation Model. A road surface friction coefficient model is established under the dual influence of temperature and snow accumulation:

[0127]

[0128] in, This is the standard friction coefficient for dry road surfaces. Temperature decay coefficient. for:

[0129]

[0130] Snow attenuation coefficient for:

[0131]

[0132] when hour ;when mm ;when mm This conforms to the physical law that the thicker the snow, the lower the coefficient of friction;

[0133] S2-4. Spatiotemporal distribution model of road surface condition. Based on the sensor network deployed along the route, the spatiotemporal distribution field of road surface ice and snow condition is constructed using ordinary kriging interpolation:

[0134]

[0135] in, The pavement condition value of the point to be estimated. For the first Observations from each sensor The Kriging weights satisfy the unbiased constraint. The weighting coefficients are obtained by solving the Kriging equations.

[0136] The specific method for step S3 is as follows:

[0137] S3-1, CAN Bus Data Acquisition. Connect to the vehicle's CAN bus via the OBD-II interface to read the following parameters in real time: brake pedal opening. Steering wheel angle yaw rate Wheel speed of each wheel ( and engine speed The sampling frequency is not less than 50Hz;

[0138] S3-2, Emergency Braking Behavior Recognition. When the braking deceleration exceeds a set threshold and the brake pedal opening exceeds the threshold, it is determined as an emergency braking event.

[0139]

[0140] in, m / s² is the deceleration threshold for emergency braking on icy and snowy roads. This represents the brake pedal opening threshold. Within the slip time window... The frequency of emergency braking within 300 seconds is as follows:

[0141]

[0142] S3-3, Lane Departure Calculation. Based on the BeiDou positioning trajectory and the lane centerline on a high-precision map, calculate the vertical distance from the vehicle's current position to the nearest lane centerline. :

[0143]

[0144] in, The vehicle's current location. The coordinates of the nearest projected point on the lane centerline. The heading angle of the lane centerline at that point. is the average radius of the Earth. When ( A distance of 0.6m is considered a lane departure event.

[0145] S3-4, Sideslip Risk Indicators. Based on the deviation between the actual and ideal yaw rates and lateral acceleration, a normalized sideslip risk index is constructed. :

[0146]

[0147] The ideal yaw rate is calculated based on a linear two-degree-of-freedom vehicle model:

[0148]

[0149] This refers to the vehicle's wheelbase. For insufficient steering gradient, It is lateral acceleration. and These are the upper safety limits for yaw rate and lateral acceleration, respectively.

[0150] S3-5, Comprehensive Feature Vector of Driving Behavior. The above parameters are used to construct a driving behavior feature vector:

[0151]

[0152] in, The rate of change of steering wheel angle. This represents the standard deviation of velocity within the time window.

[0153] The specific method for step S4 is as follows:

[0154] S4-1. Multi-source data time alignment. Using the BeiDou Time (BDT) timestamp of the BeiDou positioning data as the reference clock, time alignment is performed on the road sensor data and CAN bus data respectively:

[0155]

[0156]

[0157] in, and These are the communication transmission delay compensation amounts for the road surface sensor and the CAN bus, respectively.

[0158] S4-2, Spatial Matching. Spatial matching is performed between the vehicle's real-time BeiDou positioning and the positions of various roadside sensors. When the vehicle is in contact with the... When the ground distance of a sensor is less than the matching radius, the data from that sensor is associated with the vehicle.

[0159]

[0160] in, The distance is spherical. m is the spatial matching radius;

[0161] S4-3, Construction of Multidimensional Coupled Feature Vectors. The spatiotemporally aligned multi-source data is fused into a 12-dimensional coupled feature vector:

[0162]

[0163] S4-4, Feature Normalization. The Min-Max normalization method is used to map the features of each dimension to... Interval:

[0164]

[0165] in, For the first Original values ​​of dimensional features and These are the minimum and maximum values ​​of the feature in the historical samples, respectively.

[0166] The specific method for step S5 is as follows:

[0167] S5-1, Fuzzy Quantization of Risk Factors. A triangular fuzzy membership function is established for each dimension of the factors in the normalized coupled feature vector. Taking the road surface friction coefficient as an example... For example, three fuzzy sets are established: "low" (high risk), "medium", and "high" (low risk):

[0168]

[0169]

[0170]

[0171] The above three membership functions satisfy the following conditions on the entire domain: This ensures the completeness of the fuzzy classification.

[0172] S5-2, Dynamic Bayesian Network Construction. Establishment based on risk level... (Values) A dynamic Bayesian network with target nodes (corresponding to level four risk). The network contains three layers of evidence nodes: the environment layer (...). ), train level ( ) and driving behavior layer ( Time slice The posterior probability of the risk is calculated using Bayes' theorem:

[0173]

[0174] in, Let be the likelihood probability of the current evidence. The risk level represents the state transition probability, and the transition probability matrix is ​​obtained through statistical analysis of historical accident data.

[0175] S5-3. Comprehensive Assessment of Coupling Risk. A coupling risk index for ice and snow at low temperatures is constructed by combining fuzzy assessment and Bayesian inference results. :

[0176]

[0177] in, Scoring environmental risk factors To score the driving risk factors, Scoring of driving behavior risk factors This is a multi-factor coupling enhancement term. Weighting coefficients. Determined using the Analytic Hierarchy Process (AHP), satisfying... .

[0178] The coupling enhancement term is calculated as follows:

[0179]

[0180] in, For coupling enhancement coefficient, and These are the upper limit normalized reference values ​​for emergency braking frequency and sideslip index, respectively. This item applies only when the road surface is icy ( It is activated when there is sudden braking or skidding, reflecting the synergistic risk amplification effect when icy and snowy roads and abnormal driving behavior occur simultaneously.

[0181] S5-4, Risk Level Classification and Early Warning. Based on the coupled risk index. The risk level is divided into four levels:

[0182]

[0183] When the risk level reaches Level III or above, the system will simultaneously push early warning information to the transportation company's monitoring platform, driver's terminal, and traffic management department.

[0184] The present invention can be implemented according to the above embodiments, but its application scope is not limited thereto. The above embodiments are only to explain the implementation process of the present invention for dynamic identification of ice and snow low-temperature coupling risks under certain specific conditions, but should not be construed as limiting the application scope of the present invention.

Claims

1. A method for dynamic identification of ice and snow low-temperature coupled risks of passenger and hazardous goods vehicles in seasonally frozen areas, characterized in that, The method includes the following steps: S1. Based on the BeiDou high-precision positioning system, the driving status parameters of "two passengers and one dangerous goods" vehicles are obtained in real time, including latitude and longitude coordinates, instantaneous speed, heading angle, acceleration and braking distance, and the raw data is preprocessed by Kalman filtering. S2. By integrating data from roadside ice and snow sensors and road surface temperature sensors, and combining real-time temperature and snowfall information from meteorological stations, a spatiotemporal distribution model of road surface ice and snow conditions is constructed to quantify the attenuation characteristics of road surface friction coefficient. S3. Collect driving behavior characteristic parameters through the vehicle CAN bus and extract key indicators such as emergency braking frequency, steering wheel angle change rate, lane departure, yaw rate and lateral acceleration. S4. Establish a spatiotemporal coupling model of multi-source heterogeneous data, including ice and snow low-temperature environment parameters, road surface condition parameters, BeiDou driving parameters, and driving behavior parameters, and achieve multi-dimensional data fusion through timestamp alignment and spatial matching. S5. Based on dynamic Bayesian networks and fuzzy comprehensive evaluation methods, a dynamic identification model for the coupled risk of ice and snow low temperature in seasonally frozen areas is constructed, which outputs the risk level and triggers corresponding early warning strategies.

2. The method according to claim 1, characterized in that, The specific steps of step S1 are as follows: S1-1. Acquisition of BeiDou high-precision positioning data: Utilizing the BeiDou-3 system's RTK positioning mode, the vehicle's geocentric and geofixed coordinates are obtained in real time. The positioning accuracy is better than 0.1m, and the update frequency is not less than 10Hz; S1-2, Coordinate Transformation: Convert Earth-centered, Earth-fixed coordinates to geodetic coordinates. : in, , , For the Earth's semi-major axis, For the Earth's minor axis, For the first eccentricity, The second eccentricity; S1-3. Calculation of travel speed and acceleration: Based on the position information of two consecutive epochs, calculate the instantaneous velocity. : in, , , The average radius of the Earth Sampling interval, longitudinal acceleration Obtained from velocity difference: S1-4, Braking distance calculation, when vehicle longitudinal acceleration is detected. ( When the braking acceleration threshold is reached, it is determined to be a braking behavior, and the braking distance is... for: in, The braking start time, The moment the vehicle stops; S1-5. Kalman filter preprocessing: Establish vehicle motion state equations and observation equations, and perform filtering processing on the raw BeiDou positioning data. in, For state vectors, Here is the state transition matrix. For the observation matrix, and The process noise and observation noise are respectively assumed to be zero-mean Gaussian white noise.

3. The method according to claim 1, characterized in that, The specific steps of step S2 are as follows: S2-1, Road surface temperature acquisition and freezing determination: Road surface temperature is acquired through a road surface temperature sensor. relative humidity is obtained by combining road surface humidity sensors. The road surface is considered to be in an icy state when the following conditions are met: in, The road surface freezing temperature threshold (taken as -2°C) is used. The humidity threshold is set to 85%. S2-2, Snow thickness monitoring: Snow thickness is obtained through roadside ice and snow sensors. And establish a snow cover level classification model: S2-3, Road surface friction coefficient attenuation model: Establishing a road surface friction coefficient model under the dual influence of temperature and snow accumulation: in, The standard friction coefficient and temperature decay coefficient for dry road surfaces. for: Snow attenuation coefficient for: when hour ;when mm ;when mm This conforms to the physical law that the thicker the snow, the lower the coefficient of friction; S2-4. Spatiotemporal distribution model of road surface condition: Based on the sensor network deployed along the route, the spatiotemporal distribution field of road surface ice and snow condition is constructed using ordinary kriging interpolation. in, The pavement condition value of the point to be estimated is... For the first Observations from each sensor The Kriging weights satisfy the unbiased constraint. The weighting coefficients are obtained by solving the Kriging equations.

4. The method according to claim 1, characterized in that, The specific steps of step S3 are as follows: S3-1, CAN bus data acquisition: Accesses the vehicle's CAN bus via the OBD-II interface to read the following parameters in real time: brake pedal opening. Steering wheel angle yaw rate Wheel speed of each wheel ( and engine speed The sampling frequency is not less than 50Hz; S3-2, Emergency Braking Behavior Recognition: When the braking deceleration exceeds a set threshold and the brake pedal opening exceeds the threshold, it is determined as an emergency braking event. in, m / s² is the deceleration threshold for emergency braking on icy and snowy roads. The brake pedal opening threshold is defined within the slip time window. The frequency of emergency braking within 300 seconds is as follows: S3-3 Lane Departure Calculation: Based on the BeiDou positioning trajectory and the lane centerline of a high-precision map, calculate the vertical distance from the vehicle's current position to the nearest lane centerline. : in, The vehicle's current location. The coordinates of the nearest projected point on the lane centerline. The heading angle of the lane centerline at that point. For the average radius of the Earth, when ( A distance of 0.6m is considered a lane departure event. S3-4. Sideslip Risk Indicator: Based on the deviation between the actual yaw rate and the ideal yaw rate, as well as the lateral acceleration, a normalized sideslip risk index is constructed. : The ideal yaw rate is calculated based on a linear two-degree-of-freedom vehicle model: This refers to the vehicle's wheelbase. For insufficient steering gradient, It is lateral acceleration. and These are the upper safety limits for yaw rate and lateral acceleration, respectively. S3-5, Driving behavior comprehensive feature vector: The above parameters are used to construct a driving behavior feature vector. in, The rate of change of steering wheel angle. This represents the standard deviation of velocity within the time window.

5. The method according to claim 1, characterized in that, The specific steps of step S4 are as follows: S4-1. Multi-source data time alignment: Using the BeiDou Time (BDT) timestamp of the BeiDou positioning data as the reference clock, time alignment is performed on the road sensor data and CAN bus data respectively. in, and These are the communication transmission delay compensation amounts for the road surface sensor and the CAN bus, respectively. S4-2, Spatial Matching: Spatial matching is performed between the vehicle's real-time BeiDou positioning and the positions of various roadside sensors. When the vehicle is in contact with the... When the ground distance of a sensor is less than the matching radius, the data from that sensor is associated with the vehicle. in, The distance is spherical. m is the spatial matching radius; S4-3, Construction of Multidimensional Coupled Feature Vectors: Fusing spatiotemporally aligned multi-source data into a 12-dimensional coupled feature vector: S4-4. Feature normalization processing: The Min-Max normalization method is used to map the features of each dimension to... Interval: in, For the first Original values ​​of dimensional features and These are the minimum and maximum values ​​of the feature in the historical samples, respectively.

6. The method for dynamic identification of ice and snow low-temperature coupling risks for "passenger and dangerous goods" vehicles in seasonally frozen areas according to claim 1, characterized in that, The specific steps of step S5 are as follows: S5-1, Fuzzy Quantization of Risk Factors: A triangular fuzzy membership function is established for each dimension of the normalized coupled feature vector, using the road surface friction coefficient as an example. For example, three fuzzy sets are established: "low" (high risk), "medium", and "high" (low risk): The above three membership functions satisfy the following conditions on the entire domain: This ensures the completeness of the fuzzy classification. S5-2, Dynamic Bayesian Network Construction, Establishing a Risk Level-Based System (Values) A dynamic Bayesian network with target nodes (corresponding to level four risk) is constructed, and the network contains three layers of evidence nodes: the environment layer ( ), train level ( ) and driving behavior layer ( Time slice The posterior probability of the risk is calculated using Bayes' theorem: in, Let be the likelihood probability of the current evidence. The risk level represents the state transition probability, and the transition probability matrix is ​​obtained through statistical analysis of historical accident data. S5-3. Comprehensive assessment of coupling risks: Constructing an ice and snow low-temperature coupling risk index by integrating fuzzy assessment and Bayesian inference results. : in, Scoring environmental risk factors To score the driving risk factors, Scoring of driving behavior risk factors For multi-factor coupling enhancement terms, the weighting coefficients are... Determined using the Analytic Hierarchy Process (AHP), satisfying... , The coupling enhancement term is calculated as follows: in, For coupling enhancement coefficient, and These are the upper limit normalized reference values ​​for the frequency of emergency braking and the sideslip index, respectively. This item is only applicable when the road surface is icy ( It is activated when there is sudden braking or skidding, reflecting the synergistic risk amplification effect when icy and snowy roads and abnormal driving behavior occur simultaneously. S5-4. Risk Level Classification and Early Warning, based on the coupled risk index The risk level is divided into four levels: When the risk level reaches Level III or above, the system will simultaneously push early warning information to the transportation company's monitoring platform, driver's terminal, and traffic management department.

Images