Highway road safety risk quantification method and system based on mutual information

Abstract

The present invention discloses a method and system for quantifying highway road safety risks based on mutual information, which relates to the technical field related to intelligent transportation. The method includes: collecting multi-dimensional features of highway road areas, and obtaining multi-source heterogeneous information based on a spatiotemporal alignment mechanism; forming a road area risk factor set and performing mutual information analysis to obtain a risk mutual information matrix; performing dynamic continuous monitoring of highway road areas to obtain a real-time risk factor value set; coordinating the risk mutual information matrix and the real-time risk factor value set to determine dynamic weight distribution, and calculating the real-time risk value of the highway road area. The method solves the technical problems existing in the prior art, such as insufficient spatiotemporal correlation of multi-source heterogeneous data, insufficient characterization of the superposition relationship of risk factors, and difficulty in adapting static weight distribution to dynamic traffic environments, which lead to poor real-time and accuracy of safety risk assessment. The method achieves the technical effect of dynamically and accurately quantifying road area safety risks and improving the real-time performance of risk assessment.

Description

Technical Field

[0001] The present application relates to the technical field related to intelligent transportation, and specifically to a method and system for quantifying highway road safety risks based on mutual information. Background Art

[0002] With the rapid expansion of the expressway network and the continued growth of traffic volume, the management and early warning of road safety risks are becoming increasingly important. Expressway safety risk assessments often rely on historical accident statistics or static risk factor analysis, making them difficult to adapt to complex and changing traffic environments. Dynamically and accurately quantifying road safety risks is crucial for improving the intelligent management of expressway safety, especially given the interplay of multi-source heterogeneous data, such as meteorological conditions, traffic flow patterns, road infrastructure, and vehicle operating parameters. Existing methods lack the ability to analyze multi-factor coupling, struggle to effectively handle the spatiotemporal correlations of multi-source heterogeneous data, and inadequately characterize the nonlinear interactions between risk factors. They are unable to adapt to dynamic changes in the traffic environment, such as sudden severe weather, temporary construction, or unusual traffic congestion. Furthermore, they lack support for real-time emergency lane control, limiting the real-time and accuracy of risk assessments.

[0003] Therefore, in the current relevant technologies, there are technical problems such as insufficient spatiotemporal correlation of multi-source heterogeneous data, insufficient characterization of the superposition relationship of risk factors, and static weight distribution that is difficult to adapt to the dynamic traffic environment, resulting in poor real-time and accuracy of safety risk assessment. Summary of the Invention

[0004] This application provides a method and system for quantifying highway road safety risks based on mutual information, which solves the technical problems existing in the existing technology, such as insufficient spatiotemporal correlation of multi-source heterogeneous data, insufficient characterization of the superposition relationship of risk factors, and difficulty in adapting static weight allocation to dynamic traffic environments, resulting in poor real-time and accuracy of safety risk assessment. It achieves the technical effect of dynamically and accurately quantifying road safety risks and improving the real-time performance of risk assessment.

[0005] The present application provides a method for quantifying highway road area safety risks based on mutual information, the method comprising: collecting multi-dimensional features of the highway road area, and obtaining multi-source heterogeneous information based on a spatiotemporal alignment mechanism; establishing a road area risk factor set based on the multi-source heterogeneous information, and performing mutual information analysis on the road area risk factor set to obtain a risk mutual information matrix; using the road area risk factor set as a monitoring constraint, dynamically and continuously monitoring the highway road area to obtain a real-time risk factor value set; coordinating the risk mutual information matrix and the real-time risk factor value set to determine a dynamic weight distribution, and calculating a real-time risk value for the highway road area.

[0006] In a possible implementation, the mutual information-based highway domain safety risk quantification method also performs the following processing: obtaining historical accident information of the highway domain; obtaining historical traffic information of the highway domain; obtaining historical meteorological information of the highway domain; obtaining historical road geometry information of the highway domain; obtaining historical pavement status information of the highway domain; performing spatiotemporal alignment processing on the historical accident information, the historical traffic information, the historical meteorological information, the historical road geometry information and the historical pavement status information according to the spatiotemporal alignment mechanism to obtain the multi-source heterogeneous information.

[0007] In a possible implementation, the mutual information-based highway road safety risk quantification method further performs the following processing: extracting the first accident information from the historical accident information, wherein the first accident information includes the first time and the first location of the first accident; using the first time as the time alignment reference, aligning the first accident, the historical traffic information, the historical meteorological information, the historical road geometry information and the historical road surface state information according to the time alignment strategy in the space-time alignment mechanism to obtain first alignment information; using the first location as the space alignment reference, aligning the first accident, the historical traffic information, the historical meteorological information, the historical road geometry information and the historical road surface state information according to the space alignment strategy in the space-time alignment mechanism to obtain second alignment information; the first alignment information and the second alignment information constitute the multi-source heterogeneous information; wherein the time alignment strategy refers to the alignment strategy based on 5G network timing, and the space alignment strategy refers to the alignment strategy based on Lagrange interpolation.

[0008] In a possible implementation, the mutual information-based highway road safety risk quantification method also performs the following processing: extracting the first severity of the first accident in the first accident information; forming an initial risk factor set based on a predetermined risk dimension, wherein the initial risk factor set includes any initial risk; based on the multi-source heterogeneous information, performing a correlation analysis on the arbitrary initial risk and the first severity to obtain an arbitrary correlation coefficient; if the arbitrary correlation coefficient reaches a predetermined correlation constraint, adding the arbitrary initial risk to the road risk factor set.

[0009] In a possible implementation, the mutual information-based highway road safety risk quantification method further performs the following processing: collecting a vehicle behavior feature set in the vehicle dimension; collecting a road state feature set in the road dimension; collecting an environmental risk feature set in the environmental dimension; and forming the initial risk factor set based on the vehicle behavior feature set, the road state feature set, and the environmental risk feature set; wherein the predetermined risk dimension includes at least the vehicle dimension, the road dimension, and the environmental dimension.

[0010] In a possible implementation, the mutual information-based highway road safety risk quantification method also performs the following processing: randomly extracting the first risk and the second risk from the road risk factor set, and calculating the mutual information value of the first risk and the second risk; forming the risk mutual information matrix based on the correspondence between the first risk, the second risk and the mutual information value; which also includes: reading a predetermined mutual information value limit; if the mutual information value reaches the predetermined mutual information value limit, issuing a risk assessment correction instruction; based on the risk assessment correction instruction, introducing a cross term between the first risk and the second risk, wherein the cross term is used to correct the real-time risk value.

[0011] In a possible implementation, the mutual information-based highway road safety risk quantification method also performs the following processing: analyzing the real-time risk factor value set based on the attention mechanism to obtain real-time weight distribution; calibrating the real-time weight distribution in combination with the risk mutual information matrix to obtain a first calibrated weight distribution; and using the first calibrated weight distribution as the dynamic weight distribution.

[0012] In a possible implementation, the mutual information-based highway road safety risk quantification method also performs the following processing: extracting the arbitrary factor value time series of any risk factor in the real-time risk factor value set, and analyzing it to obtain any real-time importance coefficient; calibrating the first calibration weight distribution based on the arbitrary real-time importance coefficient to obtain a second calibration weight distribution; and replacing the first calibration weight distribution with the second calibration weight distribution as the dynamic weight distribution.

[0013] In a possible implementation, the mutual information-based highway road safety risk quantification method also performs the following processing: drawing an arbitrary factor scatter plot based on the time series of the arbitrary factor value; performing polynomial fitting analysis on the arbitrary factor scatter plot to obtain an arbitrary fitting curve; obtaining the target slope at the target moment based on the arbitrary fitting curve, and using the target slope as the arbitrary real-time important coefficient at the target moment.

[0014] The present application also provides a highway road area safety risk quantification system based on mutual information, which includes: a multi-source heterogeneous information acquisition unit, which is used to collect multi-dimensional features of the highway road area and obtain multi-source heterogeneous information according to a spatiotemporal alignment mechanism; a mutual information analysis unit, which is used to form a road area risk factor set based on the multi-source heterogeneous information, and perform mutual information analysis on the road area risk factor set to obtain a risk mutual information matrix; a dynamic continuous monitoring unit, which is used to perform dynamic continuous monitoring of the highway road area with the road area risk factor set as a monitoring constraint to obtain a real-time risk factor value set; a real-time risk value calculation unit, which is used to coordinate the risk mutual information matrix and the real-time risk factor value set to determine the dynamic weight distribution and calculate the real-time risk value of the highway road area.

[0015] The mutual information-based highway road safety risk quantification method and system proposed in this application is intended to collect multi-dimensional features of highway roads and obtain multi-source heterogeneous information based on a spatiotemporal alignment mechanism; to establish a road risk factor set and conduct mutual information analysis to obtain a risk mutual information matrix; to use the road risk factor set as a monitoring constraint, to conduct dynamic continuous monitoring of the highway road to obtain a real-time risk factor value set; to coordinate the risk mutual information matrix and the real-time risk factor value set to determine a dynamic weight distribution and calculate the real-time risk value of the highway road. This solves the technical problems existing in the prior art of insufficient spatiotemporal correlation of multi-source heterogeneous data, insufficient characterization of the superposition relationship of risk factors, and difficulty in adapting static weight distribution to dynamic traffic environments, resulting in poor real-time and accuracy of safety risk assessment. It achieves the technical effect of dynamically and accurately quantifying road safety risks and improving the real-time performance of risk assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] To more clearly illustrate the technical solutions of the embodiments of the present disclosure, the accompanying drawings of the embodiments of the present disclosure are briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in precise order. Instead, various steps may be processed in reverse order or simultaneously as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0017] Figure 1 A flowchart of a method for quantifying highway road safety risks based on mutual information is provided in an embodiment of the present application.

[0018] Figure 2 Schematic diagram of the structure of the highway road safety risk quantification system based on mutual information provided in an embodiment of the present application.

[0019] Explanation of the reference numerals: multi-source heterogeneous information obtaining unit 10 , mutual information analyzing unit 20 , dynamic continuous monitoring unit 30 , real-time risk value calculating unit 40 . DETAILED DESCRIPTION

[0020] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0021] In order to make the purpose, technical solutions and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limiting this application. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0022] In the following description, reference is made to “some embodiments”, which describes a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict, and the terms “first\second” involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. The terms “including” and “having” and any variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or that are inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs. The terms used herein are for the purpose of describing the embodiments of this application only.

[0023] The embodiment of the present application provides a method for quantifying highway safety risks based on mutual information, such as Figure 1 As shown, the method includes:

[0024] In step S100 , multi-dimensional features of the highway area are collected, and multi-source heterogeneous information is obtained according to a spatiotemporal alignment mechanism.

[0025] Preferably, multi-dimensional feature collection of highway domains refers to obtaining various dynamic and static information related to traffic safety through different data sources such as traffic monitoring cameras, weather stations, road sensors, GPS trajectories, and traffic management platforms, mainly including historical accident information, traffic flow information, weather information, road geometry information, and road surface status information of highway domains. These information are multi-source heterogeneous, and direct fusion will lead to analysis deviation. The data is standardized through the spatiotemporal alignment mechanism, including time alignment processing, space alignment processing, and data association matching processing. Among them, time alignment processing is to standardize the data of different frequencies. Data is unified to the same time granularity, for example, hourly updated weather data is interpolated to a timestamp that matches second-updated traffic flow data; spatial alignment processing is to map data to a unified spatial unit, for example, sparse weather station data is allocated to specific road sections through spatial interpolation; data association matching processing is to establish association rules between different data. For example, when analyzing the accident risk of a certain curve, the geometric line shape, real-time vehicle speed and rainfall data of the location need to be synchronously associated; ultimately, structured, spatiotemporally consistent multi-source heterogeneous information is formed to ensure that features of different dimensions are accurately associated in time and space, avoiding misjudgment due to data dislocation.

[0026] Furthermore, step S100 also includes step S110, obtaining historical accident information of the highway area; step S120, obtaining historical traffic information of the highway area; step S130, obtaining historical meteorological information of the highway area; step S140, obtaining historical road geometry information of the highway area; step S150, obtaining historical pavement status information of the highway area; step S160, performing spatiotemporal alignment processing on the historical accident information, the historical traffic information, the historical meteorological information, the historical road geometry information and the historical pavement status information according to the spatiotemporal alignment mechanism to obtain the multi-source heterogeneous information.

[0027] Preferably, historical accident information on the expressway is obtained using the traffic management department's accident database, traffic police accident reports, and highway operating company records. This includes the time and location of the accident, such as the road section stake number; the type of accident, such as rear-end collision, rollover, and collision with a guardrail; and the severity of the accident, such as the number of casualties and the level of property damage. Historical traffic information on the expressway is obtained through ETC gantries, traffic monitoring cameras, geomagnetic coils, and GPS data from floating vehicles, including the time and section of data collection, such as the specific stake number interval; historical traffic volume, such as the number of vehicles per unit time; historical average speed, i.e., the average speed of vehicles on the road section; historical lane occupancy, i.e., the proportion of time a lane is occupied by vehicles; and historical traffic congestion indices, such as speed-based congestion levels.

[0028] Preferably, historical meteorological information for the expressway is obtained through meteorological monitoring stations, meteorological sensors along the expressway, and satellite remote sensing data. This includes the recording time, location, visibility, rainfall, wind speed, temperature, and road dryness / slipperiness / iciness. Highway design drawings, GIS, and drone aerial survey data are used to obtain historical road geometry information for the expressway, including section IDs, such as stake number intervals; linear metrics, such as horizontal curve radius, longitudinal slope, and superelevation; number of lanes (i.e., the number of one-way or two-way lanes); shoulder width (i.e., the width of the emergency lane); and the location of interchanges, tunnels, bridges, and special structures. Historical road surface condition information for the expressway is obtained through road inspection vehicles, IoT sensors, and manual inspection reports, including inspection time, inspection location, road surface roughness, friction coefficient, cracks and potholes, and construction and maintenance status.

[0029] Preferably, the temporal granularity, spatial coverage, and data format of historical accident information, historical traffic information, historical meteorological information, historical road geometry information, and historical road surface condition information are all inconsistent. A spatiotemporal alignment mechanism is used to standardize spatiotemporal alignment, including both temporal and spatial alignment. Specifically, temporal alignment uses linear interpolation to match low-frequency data to high-frequency time points. For example, hourly meteorological data is interpolated to minute-by-minute data to match traffic flow data. Accident data is aggregated according to a predetermined time window, and the number of accidents within that time period is calculated. Spatial alignment uniformly maps all data to the highway's pile number system. Kriging interpolation is used for meteorological data to estimate weather conditions along the highway, and gridding is used for data association in complex areas such as interchanges. Ultimately, multi-source data aligned in time and space—that is, multi-source heterogeneous information—is generated, ensuring improved accuracy and reliability of risk assessments.

[0030] Furthermore, step S160 also includes step S161, extracting the first accident information from the historical accident information, wherein the first accident information includes the first time and the first location of the first accident; step S162, taking the first time as the time alignment reference, aligning the first accident, the historical traffic information, the historical meteorological information, the historical road geometry information and the historical road surface state information according to the time alignment strategy in the space-time alignment mechanism to obtain first alignment information; step S163, taking the first location as the space alignment reference, aligning the first accident, the historical traffic information, the historical meteorological information, the historical road geometry information and the historical road surface state information according to the space alignment strategy in the space-time alignment mechanism to obtain second alignment information; step S164, the first alignment information and the second alignment information constitute the multi-source heterogeneous information; wherein, the time alignment strategy refers to the alignment strategy based on 5G network timing, and the space alignment strategy refers to the alignment strategy based on Lagrange interpolation.

[0031] Preferably, an accident information is randomly extracted from the historical accident information as the first accident information, and its corresponding first time and first location are obtained; the first accident, historical traffic information, historical meteorological information, historical road geometry information and historical road surface state information are aligned according to the time alignment strategy in the spatiotemporal alignment mechanism, wherein the time alignment strategy refers to an alignment strategy based on 5G network timing, that is, utilizing the ultra-low latency and nanosecond time synchronization capabilities of the 5G network to unify the time base of all data collection to eliminate clock errors, and then using the first time as the time alignment base, that is, the time of occurrence of the first accident as the alignment anchor point, setting the time window, that is, selecting data within a certain time range before and after the accident, and then supplementing the low-frequency data through Lagrange interpolation, that is, using linear interpolation to estimate the value of low-frequency data such as meteorology, such as 1-minute meteorological data and 100ms radar data, and then outputting the value of the traffic information, historical meteorological information, historical road geometry information and historical road surface state information at the time alignment base, which is the first alignment information.

[0032] Preferably, the first accident, historical traffic information, historical meteorological information, historical road geometry information, and historical road surface condition information are aligned according to a spatial alignment strategy within the spatiotemporal alignment mechanism. The spatial alignment strategy is an alignment strategy based on Lagrangian interpolation, which estimates the value of a target location by fitting a polynomial to data at known discrete points. The first location is then used as the spatial alignment reference, i.e., the first accident location is used as the spatial anchor point. Neighboring data from each data point is selected, such as detector data at adjacent locations and observation data from the two nearest meteorological stations. The target value is then interpolated. For example, for traffic flow data, the vehicle speed at the spatial alignment reference is calculated using the Lagrangian interpolation formula; for meteorological data, the rainfall at the spatial alignment reference is interpolated. The values ​​of the traffic information, historical meteorological information, historical road geometry information, and historical road surface condition information at the spatial alignment reference are then output as the second alignment information. Finally, the first and second alignment information are merged to form the final multi-source heterogeneous information, thereby improving the high-precision data input for dynamic risk assessment and ensuring improved accuracy and reliability of risk assessment.

[0033] In step S200 , a road risk factor set is constructed based on the multi-source heterogeneous information, and mutual information analysis is performed on the road risk factor set to obtain a risk mutual information matrix.

[0034] Step S200 further includes step S210, extracting the first severity of the first accident from the first accident information; step S220, forming an initial risk factor set based on a predetermined risk dimension, wherein the initial risk factor set includes any initial risk; step S230, based on the multi-source heterogeneous information, performing a correlation analysis on the arbitrary initial risk and the first severity to obtain an arbitrary correlation coefficient; step S240, if the arbitrary correlation coefficient reaches a predetermined correlation constraint, adding the arbitrary initial risk to the road risk factor set.

[0035] Preferably, a road risk factor set is formed based on multi-source heterogeneous information, and factors that are significantly correlated with the severity of the accident are screened out from the initial risk factor set to construct a road risk factor set for risk assessment. Specifically, quantitative indicators such as the number of casualties, property loss level, traffic congestion duration, and number of closed lanes are extracted from the first accident information, and the corresponding first severity is obtained according to the injury severity assessment, and then an initial risk factor set is formed based on a predetermined risk dimension, wherein the predetermined risk dimension is a category of factors that may affect the severity of the accident, which is pre-defined based on domain knowledge, and includes at least vehicle dimension, road dimension, and environmental dimension. For example, the initial risk may include traffic flow, vehicle speed, congestion index, visibility, rainfall, road wetness, pothole area ratio, and friction coefficient attenuation rate, and then an initial risk factor set is formed and includes any initial risk.

[0036] Preferably, based on multi-source heterogeneous information, a correlation analysis is performed on any initial risk and the first severity. Specifically, the statistical correlation between each initial risk factor and the severity of the accident is calculated through the Pearson correlation coefficient or the Spearman rank correlation to obtain an arbitrary correlation coefficient. For example, a correlation coefficient of 0.65 indicates a strong positive correlation, which may be due to speeding aggravating the consequences of the accident; a correlation coefficient of -0.72 indicates a strong negative correlation, and low visibility leads to more serious accidents; then a predetermined correlation constraint is used to determine whether the risk factor is significant. If any correlation coefficient reaches the predetermined correlation constraint, it means that the initial risk factor is significantly correlated with the severity of the accident, and it is added to the road risk factor set; if any correlation coefficient does not reach the predetermined correlation constraint, it means that the initial risk factor is not correlated with the severity of the accident, and it is eliminated, and finally a road risk factor set is obtained, which may include average vehicle speed, rainfall, visibility, pothole area ratio, etc.

[0037] Furthermore, step S220 also includes step S221, collecting a vehicle behavior feature set in the vehicle dimension; step S222, collecting a road state feature set in the road dimension; step S223, collecting an environmental risk feature set in the environmental dimension; step S224, forming the initial risk factor set based on the vehicle behavior feature set, the road state feature set and the environmental risk feature set; wherein the predetermined risk dimension includes at least the vehicle dimension, the road dimension and the environmental dimension.

[0038] Optimally, vehicle-level vehicle behavior features are collected through on-board OBD terminals, ETC gantries, traffic cameras, radar speed cameras, and GPS tracks of floating vehicles. These features represent the dynamic operating state and driving behavior of vehicles on highways, potentially including real-time speed, acceleration, lane change frequency, and the proportion of large vehicles. Road state features are collected through road inspection vehicles, IoT sensors, drone inspections, and manual maintenance records. These features represent the physical state and geometric characteristics of highway infrastructure, potentially including pothole area, crack length density, road surface smoothness, and road marking clarity. Furthermore, environmental risk features are collected through weather stations, visibility meters, road surface temperature sensors, and vehicle-to-everything (V2X) communications. These features represent meteorological conditions and special environmental events, potentially including visibility, friction coefficient decay rate, rainfall, and icing probability. Finally, these three-dimensional state feature sets are integrated into a structured dataset to form an initial risk factor set.

[0039] Furthermore, step S200 also includes step S250, randomly extracting the first risk and the second risk in the road risk factor set, and calculating the mutual information value of the first risk and the second risk; step S260, forming the risk mutual information matrix based on the correspondence between the first risk, the second risk and the mutual information value; wherein, it also includes: step a, reading the predetermined mutual information value limit; step b, if the mutual information value reaches the predetermined mutual information value limit, issuing a risk assessment correction instruction; step c, based on the risk assessment correction instruction, introducing the cross term of the first risk and the second risk, wherein the cross term is used to correct the real-time risk value.

[0040] Preferably, two factors are randomly extracted from the road risk factor set as the first risk and the second risk, and the mutual information value between the first risk and the second risk is calculated. The mutual information is used to quantify the nonlinear dependency between the two risk factors. A larger value indicates a stronger synergy. For example, the mutual information value of visibility and vehicle speed is greater than 0.7 (strong coupling), while the mutual information value of traffic flow and slope is only 0.3 (weak coupling). Specifically, based on the joint probability distribution of historical data, the reduction in uncertainty of one factor on another is measured. For example, the mutual information value of vehicle speed and rainfall is 0.4, indicating a strong correlation, indicating that speed reduction is required on rainy days. Then, based on the corresponding relationship between the first risk, the second risk, and the mutual information value, the mutual information values ​​of all risk factors are filled into a matrix to form a symmetric matrix as the risk mutual information matrix. For example, the quantitative indicator of vehicle behavior is vehicle speed, which is obtained by the ETC gantry; the quantitative indicator of road condition is the pothole area ratio, which is obtained by the road surface inspection vehicle; and the quantitative indicator of environmental risk is visibility or friction coefficient attenuation rate, which is obtained by the weather station and road surface sensors. An exemplary risk mutual information matrix is ​​shown in Table 1:

[0041] Table 1 Risk coupling matrix data table

[0042]

[0043] Preferably, a predetermined mutual information value limit is configured to determine whether the risk assessment model needs to be revised. If the mutual information value reaches the predetermined mutual information value limit of 0.6, it indicates that there is a significant interaction between the risk factors, and a risk assessment correction instruction is issued, indicating that the current risk assessment model does not consider the synergistic effect of the two and needs to be revised, including introducing the cross-term of the first risk and the second risk based on the risk assessment correction instruction, that is, adding a new item in the risk assessment model to represent the joint impact of the two risks, such as the cross-term of vehicle speed and rainfall is 0.3×vehicle speed×rainfall, which means that the risk increases exponentially when speeding on rainy days; finally, a weighted sum is performed to correct the real-time risk value, which significantly improves the risk warning capability for complex risk scenarios such as extreme weather and sudden changes in traffic flow.

[0044] Step S300 , using the road area risk factor set as a monitoring constraint, dynamically and continuously monitoring the highway road area to obtain a real-time risk factor value set.

[0045] Preferably, based on a pre-established set of road risk factors, such as vehicle speed, rainfall, visibility, road friction coefficient, etc., real-time data collection and analysis is carried out on all sections and all time periods of the expressway, wherein the road risk factor set is used as a monitoring constraint to ensure that only data related to the risk factors are collected and the monitoring frequency is defined at the same time; dynamic and continuous monitoring of the expressway road area is carried out, vehicle speed data is obtained through ETC gantry / radar speed measurement, rainfall information is obtained through meteorological sensors / V2X vehicle network, road friction coefficient is obtained through embedded road sensors, and visibility is obtained using visibility meter / camera AI analysis, and finally a real-time risk factor value set reflecting the current risk status is obtained.

[0046] Step S400 : Coordinate the risk mutual information matrix and the real-time risk factor value set to determine dynamic weight distribution, and calculate the real-time risk value of the highway area.

[0047] Step S400 also includes step S410, analyzing the real-time risk factor value set based on the attention mechanism to obtain real-time weight distribution; step S420, calibrating the real-time weight distribution in combination with the risk mutual information matrix to obtain a first calibrated weight distribution; step S430, using the first calibrated weight distribution as the dynamic weight distribution.

[0048] Preferably, the collaborative risk mutual information matrix and the real-time risk factor value set, that is, the dynamic importance of the real-time risk factors is analyzed through the attention mechanism, and the risk mutual information matrix is ​​used to calibrate the weights, the synergistic effect of multiple factors is captured, and finally a dynamic weight distribution reflecting the current road conditions is generated. Specifically, the real-time risk factor value set is analyzed based on the attention mechanism, the contribution of each factor to the current risk is automatically calculated and the normalized weight is output through the Softmax function to obtain the initial real-time weight distribution; then, the real-time weight distribution is calibrated in combination with the risk mutual information matrix to reflect the synergistic effect between the risk factors. If the mutual information value of the two factors is high, It is necessary to increase their joint weight, that is, to increase the weight of high mutual information factor pairs proportionally to enhance the cross-term, and then perform normalization to keep the total weight at 1, and then generate a first calibration weight distribution, and use the first calibration weight distribution as the dynamic weight distribution, and calculate the real-time risk value of the highway domain, including mapping the real-time risk factor value to the [0, 1] interval through range normalization, and then using the dynamic allocation weight for weighted summation to obtain the real-time risk value, for example, 0~0.3 low risk, 0.3~0.7 medium risk, 0.7~1.0 high risk, thereby significantly improving the accuracy of highway safety risk assessment in complex environments.

[0049] Furthermore, step S430 also includes step S431, extracting the time series of any factor value of any risk factor in the real-time risk factor value set, and analyzing to obtain any real-time importance coefficient; step S432, calibrating the first calibration weight distribution based on the arbitrary real-time importance coefficient to obtain a second calibration weight distribution; step S433, replacing the first calibration weight distribution with the second calibration weight distribution as the dynamic weight distribution.

[0050] Preferably, the time series of any factor value of any risk factor is extracted from the real-time risk factor value set, that is, the recent time series data of any risk factor such as the friction coefficient, and then the time series feature calculation is performed, including calculating the rate of change, that is, the deviation between the current value and the historical mean; the trend strength, that is, the linear regression slope; the volatility, that is, the standard deviation; and then the weighted calculation of the comprehensive value of several time series features is used as the arbitrary real-time important coefficient. The first calibration weight distribution is then calibrated based on the arbitrary real-time important coefficient. If the arbitrary real-time important coefficient is greater than the preset threshold, the corresponding risk factor weight is increased, and other factors are compressed proportionally to keep the sum of 1, thereby obtaining the second calibration weight distribution. Finally, the second calibration weight distribution replaces the first calibration weight distribution as the dynamic weight distribution to ensure that the weight adjustment takes into account both the real-time data and the coupling relationship between factors, thereby ensuring the accuracy and rationality of the weight distribution.

[0051] Furthermore, step S431 also includes step S4311, drawing a scatter plot of any factor based on the time series of the arbitrary factor value; step S4312, performing a polynomial fitting analysis on the scatter plot of any factor to obtain an arbitrary fitting curve; step S4313, obtaining a target slope at the target moment based on the arbitrary fitting curve, and using the target slope as the arbitrary real-time important coefficient at the target moment.

[0052] Preferably, accurate quantification of the changing trend of risk factors is achieved through scatter plot fitting analysis. Specifically, a scatter plot of any factor is drawn according to the time series of the arbitrary factor value, that is, the continuous monitoring values ​​of the target risk factor in the recent time window are selected to form a time series data set, and then the original data points are plotted with time as the horizontal axis and the factor value as the vertical axis to obtain a scatter plot of any factor; the least squares method is then used to perform a quadratic polynomial fitting analysis on the scatter plot of any factor to obtain an arbitrary fitting curve, and the goodness of fit is judged by the determination coefficient; then the slope function is obtained by differentiating the arbitrary fitting curve with respect to time, and the target time is substituted into the slope function to calculate and determine the target slope. If the target slope is positive, it indicates that visibility has improved; if the target slope is negative, it indicates that the friction coefficient has decreased; and the larger the absolute value of the target slope, the more drastic the change; finally, the target slope is mapped to the [0, 1] interval through the Sigmoid function to obtain the final arbitrary real-time important coefficient, thereby ensuring dynamic and accurate quantification of road safety risks and improving the real-time and reliability of risk assessment.

[0053] In the above, refer to Figure 1 The highway road safety risk quantification method based on mutual information according to an embodiment of the present invention is described in detail. Figure 2 A highway road safety risk quantification system based on mutual information according to an embodiment of the present invention is described.

[0054] The highway road safety risk quantification system based on mutual information according to the embodiment of the present invention is used to solve the technical problems existing in the prior art, such as insufficient spatiotemporal correlation of multi-source heterogeneous data, insufficient characterization of the superposition relationship of risk factors, and difficulty in adapting static weight allocation to dynamic traffic environments, which lead to poor real-time and accuracy of safety risk assessment. It achieves the technical effect of dynamically and accurately quantifying road safety risks and improving the real-time performance of risk assessment. Figure 2 As shown, the highway road safety risk quantification system based on mutual information includes: a multi-source heterogeneous information acquisition unit 10, a mutual information analysis unit 20, a dynamic continuous monitoring unit 30, and a real-time risk value calculation unit 40.

[0055] The multi-source heterogeneous information acquisition unit 10 is used to collect multi-dimensional features of the highway road area and obtain multi-source heterogeneous information based on the spatiotemporal alignment mechanism; the mutual information analysis unit 20 is used to form a road area risk factor set based on the multi-source heterogeneous information, and perform mutual information analysis on the road area risk factor set to obtain a risk mutual information matrix; the dynamic continuous monitoring unit 30 is used to perform dynamic continuous monitoring of the highway road area with the road area risk factor set as the monitoring constraint to obtain a real-time risk factor value set; the real-time risk value calculation unit 40 is used to coordinate the risk mutual information matrix and the real-time risk factor value set to determine the dynamic weight distribution and calculate the real-time risk value of the highway road area.

[0056] The specific configuration of the multi-source heterogeneous information acquisition unit 10 will be described in detail below. The multi-source heterogeneous information acquisition unit 10 further includes: obtaining historical accident information for the highway area; obtaining historical traffic information for the highway area; obtaining historical meteorological information for the highway area; obtaining historical road geometry information for the highway area; obtaining historical road surface condition information for the highway area; and performing spatiotemporal alignment processing on the historical accident information, the historical traffic information, the historical meteorological information, the historical road geometry information, and the historical road surface condition information according to the spatiotemporal alignment mechanism to obtain the multi-source heterogeneous information.

[0057] The specific configuration of the multi-source heterogeneous information acquisition unit 10 will be described in detail below. The multi-source heterogeneous information acquisition unit 10 further includes: extracting the first accident information from the historical accident information, wherein the first accident information includes the first time and the first location of the first accident; using the first time as the time alignment reference, aligning the first accident, the historical traffic information, the historical meteorological information, the historical road geometry information, and the historical road surface state information according to the time alignment strategy in the spatiotemporal alignment mechanism to obtain first alignment information; using the first location as the spatial alignment reference, aligning the first accident, the historical traffic information, the historical meteorological information, the historical road geometry information, and the historical road surface state information according to the spatial alignment strategy in the spatiotemporal alignment mechanism to obtain second alignment information; the first alignment information and the second alignment information constitute the multi-source heterogeneous information; wherein the time alignment strategy refers to an alignment strategy based on 5G network timing, and the spatial alignment strategy refers to an alignment strategy based on Lagrange interpolation.

[0058] The specific configuration of the mutual information analysis unit 20 will be described in detail below. The mutual information analysis unit 20 further includes: extracting the first severity of the first accident from the first accident information; forming an initial risk factor set based on a predetermined risk dimension, wherein the initial risk factor set includes any initial risk; performing a correlation analysis between the arbitrary initial risk and the first severity based on the multi-source heterogeneous information to obtain an arbitrary correlation coefficient; and adding the arbitrary initial risk to the road domain risk factor set if the arbitrary correlation coefficient meets a predetermined correlation constraint.

[0059] The specific configuration of mutual information analysis unit 20 will be described in detail below. Mutual information analysis unit 20 further includes: collecting a vehicle behavior feature set in the vehicle dimension; collecting a road state feature set in the road dimension; and collecting an environmental risk feature set in the environment dimension; and assembling the initial risk factor set based on the vehicle behavior feature set, the road state feature set, and the environmental risk feature set. The predetermined risk dimensions include at least the vehicle dimension, the road dimension, and the environmental dimension.

[0060] The specific configuration of the mutual information analysis unit 20 will be described in detail below. The mutual information analysis unit 20 further includes: randomly extracting the first risk and the second risk from the road risk factor set, and calculating the mutual information value between the first risk and the second risk; forming the risk mutual information matrix based on the correspondence between the first risk, the second risk, and the mutual information value; wherein, it also includes: reading a predetermined mutual information value limit; if the mutual information value reaches the predetermined mutual information value limit, issuing a risk assessment correction instruction; based on the risk assessment correction instruction, introducing a cross term between the first risk and the second risk, wherein the cross term is used to correct the real-time risk value.

[0061] The specific configuration of the real-time risk value calculation unit 40 will be described in detail below. The real-time risk value calculation unit 40 further includes: analyzing the real-time risk factor value set based on the attention mechanism to obtain a real-time weight distribution; calibrating the real-time weight distribution based on the risk mutual information matrix to obtain a first calibrated weight distribution; and using the first calibrated weight distribution as the dynamic weight distribution.

[0062] The specific configuration of the real-time risk value calculation unit 40 will be described in detail below. The real-time risk value calculation unit 40 further includes: extracting the time series of any factor value of any risk factor in the real-time risk factor value set and analyzing it to obtain any real-time importance coefficient; calibrating the first calibration weight distribution based on the arbitrary real-time importance coefficient to obtain a second calibration weight distribution; and replacing the first calibration weight distribution with the second calibration weight distribution as the dynamic weight distribution.

[0063] The specific configuration of the real-time risk value calculation unit 40 will be described in detail below. The real-time risk value calculation unit 40 further includes: plotting an arbitrary factor scatter plot based on the arbitrary factor value time series; performing polynomial fitting analysis on the arbitrary factor scatter plot to obtain an arbitrary fitting curve; and obtaining a target slope at a target moment based on the arbitrary fitting curve, and using the target slope as the arbitrary real-time importance coefficient at the target moment.

[0064] The highway road safety risk quantification system based on mutual information provided by an embodiment of the present invention can execute the highway road safety risk quantification method based on mutual information provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0065] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other and are not used to limit the scope of protection of the present invention.

[0066] The above specific embodiments do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application shall be included within the scope of protection of this application.

Claims

1. A highway safety risk quantification method based on mutual information, characterized by: include: Collect multi-dimensional features of highway areas and obtain multi-source heterogeneous information based on the spatiotemporal alignment mechanism; Building a road risk factor set based on the multi-source heterogeneous information, and performing mutual information analysis on the road risk factor set to obtain a risk mutual information matrix; Taking the road area risk factor set as a monitoring constraint, dynamically and continuously monitoring the highway road area to obtain a real-time risk factor value set; Collaborating the risk mutual information matrix with the real-time risk factor value set to determine dynamic weight distribution and calculate the real-time risk value of the highway area; Collect multi-dimensional features of highway areas and obtain multi-source heterogeneous information based on the spatiotemporal alignment mechanism, including: Obtaining historical accident information of the highway area; Obtaining historical traffic information of the expressway area; Obtaining historical meteorological information of the highway area; Obtaining historical road geometry information of the highway area; Obtaining historical road surface status information of the highway area; Performing spatiotemporal alignment processing on the historical accident information, the historical traffic information, the historical meteorological information, the historical road geometry information, and the historical road surface state information according to the spatiotemporal alignment mechanism to obtain the multi-source heterogeneous information; Perform mutual information analysis on the road risk factor set to obtain a risk mutual information matrix, including: Randomly extracting a first risk and a second risk from the road risk factor set, and calculating a mutual information value between the first risk and the second risk; forming the risk mutual information matrix based on the corresponding relationship between the first risk, the second risk and the mutual information value; Among them, also include: Reading a predetermined mutual information value limit; If the mutual information value reaches the predetermined mutual information value limit, issuing a risk assessment correction instruction; Based on the risk assessment correction instruction, an intersection term of the first risk and the second risk is introduced, wherein the intersection term is used to correct the real-time risk value.

2. The highway safety risk quantification method based on mutual information according to claim 1 is characterized in that: Performing spatiotemporal alignment processing on the historical accident information, the historical traffic information, the historical meteorological information, the historical road geometry information, and the historical road surface state information according to the spatiotemporal alignment mechanism to obtain the multi-source heterogeneous information includes: Extracting first accident information from the historical accident information, wherein the first accident information includes a first time and a first location of the first accident; Using the first time as a time alignment reference, aligning the first accident, the historical traffic information, the historical meteorological information, the historical road geometry information, and the historical road surface state information according to a time alignment strategy in the spatiotemporal alignment mechanism to obtain first alignment information; Using the first location as a spatial alignment reference, aligning the first accident, the historical traffic information, the historical meteorological information, the historical road geometry information, and the historical road surface condition information according to a spatial alignment strategy in the spatiotemporal alignment mechanism to obtain second alignment information; The first alignment information and the second alignment information constitute the multi-source heterogeneous information; Among them, the time alignment strategy refers to an alignment strategy based on 5G network timing, and the space alignment strategy refers to an alignment strategy based on Lagrange interpolation.

3. The highway safety risk quantification method based on mutual information according to claim 2 is characterized in that: A road risk factor set is established based on the multi-source heterogeneous information, including: extracting a first severity of the first accident from the first accident information; forming an initial risk factor set based on predetermined risk dimensions, wherein the initial risk factor set includes any initial risk; Based on the multi-source heterogeneous information, performing a correlation analysis between the arbitrary initial risk and the first severity to obtain an arbitrary correlation coefficient; If the arbitrary correlation coefficient reaches a predetermined correlation constraint, the arbitrary initial risk is added to the road area risk factor set.

4. The highway safety risk quantification method based on mutual information according to claim 3 is characterized in that: An initial set of risk factors is formed based on predetermined risk dimensions, including: Collecting vehicle behavior feature sets in vehicle dimensions; Collect road state feature sets in road dimension; Collect environmental risk feature sets from environmental dimensions; composing the initial risk factor set based on the vehicle behavior feature set, the road state feature set, and the environmental risk feature set; The predetermined risk dimensions include at least the vehicle dimension, the road dimension and the environment dimension.

5. The highway safety risk quantification method based on mutual information according to claim 1 is characterized in that: Determining dynamic weight distribution by coordinating the risk mutual information matrix with the real-time risk factor value set includes: Analyzing the real-time risk factor value set based on an attention mechanism to obtain real-time weight distribution; Calibrate the real-time weight distribution in combination with the risk mutual information matrix to obtain a first calibrated weight distribution; The first calibration weight distribution is used as the dynamic weight distribution.

6. The highway safety risk quantification method based on mutual information according to claim 5 is characterized in that: Before allocating the first calibration weight as the dynamic weight, the method further includes: Extracting a time series of any factor value of any risk factor in the real-time risk factor value set, and analyzing it to obtain any real-time important coefficient; calibrating the first calibration weight distribution based on the arbitrary real-time important coefficient to obtain a second calibration weight distribution; The first calibration weight distribution is replaced by the second calibration weight distribution as the dynamic weight distribution.

7. The highway safety risk quantification method based on mutual information according to claim 6 is characterized in that: Extracting the arbitrary factor value time series of any risk factor in the real-time risk factor value set and analyzing it to obtain any real-time important coefficient includes: Draw a scatter plot of any factor according to the time series of the arbitrary factor values; Performing polynomial fitting analysis on the scatter plot of the arbitrary factors to obtain an arbitrary fitting curve; A target slope at a target moment is obtained according to the arbitrary fitting curve, and the target slope is used as the arbitrary real-time important coefficient at the target moment.

8. The highway safety risk quantification system based on mutual information is characterized by: The system is used to implement the mutual information-based highway road safety risk quantification method according to any one of claims 1 to 7, and the system includes: A multi-source heterogeneous information acquisition unit is used to collect multi-dimensional features of the highway area and obtain multi-source heterogeneous information based on the spatiotemporal alignment mechanism; a mutual information analysis unit, configured to construct a road risk factor set based on the multi-source heterogeneous information, and perform mutual information analysis on the road risk factor set to obtain a risk mutual information matrix; A dynamic continuous monitoring unit is used to perform dynamic continuous monitoring of the highway road area with the road area risk factor set as the monitoring constraint to obtain a real-time risk factor value set; a real-time risk value calculation unit is used to coordinate the risk mutual information matrix and the real-time risk factor value set to determine the dynamic weight distribution and calculate the real-time risk value of the highway road area.

Images