Road risk grading early warning method and system based on Beidou satellite system

Abstract

The invention discloses a road risk grading early warning method and system based on a Beidou satellite system, and the method comprises the steps: firstly carrying out the real-time positioning of a vehicle through Beidou dual-frequency signals, inertial navigation data, road side unit differential data and vehicle-mounted sensor data, and obtaining the precise position information; fusing the real-time position of the vehicle with meteorological data, vehicle-mounted OBD parameters and social media public opinion data to generate a dynamic risk factor matrix; a fuzzy rule base is constructed based on expert experience, the fuzzy weight of each risk factor in a matrix is obtained, meanwhile, the time sequence weight of each factor is predicted by means of an LSTM model, and a final road risk score is obtained through dynamic weighting. And performing graded early warning on the road risk in combination with the driver portrait, and feeding back and updating the fusion parameter, the fuzzy rule base or the LSTM model parameter according to the early warning effect to form a closed-loop optimization mechanism. Dynamic coupling analysis of multi-dimensional risk factors is realized, and the real-time performance and accuracy of road risk early warning are improved.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent transportation technology and relates to a road risk classification early warning method and system based on the Beidou satellite system. Background Art

[0002] With the comprehensive deployment of the BeiDou satellite navigation system and the widespread adoption of high-precision positioning services, BeiDou-based road risk monitoring technology has become a core research area in the field of intelligent transportation. By acquiring key information such as road conditions and vehicle operating status in real time, this technology provides strong support for traffic safety management, accident prevention, and emergency response, significantly improving traffic efficiency and ensuring travel safety.

[0003] However, existing Beidou-based road risk monitoring technology still faces many challenges and significant defects in practical applications. In terms of risk calibration mechanisms, traditional road risk calibration systems often rely on preset static thresholds or historical accident data to build risk assessment models. Although this static calibration mechanism has certain applicability in structured road scenarios, it exposes obvious shortcomings when faced with dynamic environmental changes and diverse road scenarios. First, it has poor adaptability to dynamic environments and cannot respond in real time to dynamic factors such as sudden changes in traffic flow, extreme weather or sudden accidents, resulting in the inability to issue effective warnings in a timely manner; second, the scenario generalization capability is insufficient. Due to the significant differences in risk characteristics of different road types, unified calibration rules are difficult to meet the risk assessment needs of multiple scenarios.

[0004] At the data fusion level, while current technologies attempt to integrate multiple data sources, they face serious bottlenecks. Different types of data lack unified spatiotemporal benchmarks, and traditional fusion methods rely on manual calibration and alignment, resulting in high processing latency and an inability to meet real-time warning requirements. Existing frameworks that rely on centralized cloud computing waste edge computing resources and make low-latency early warning difficult. Furthermore, unstructured data such as social media sentiment and roadside camera video streams have yet to be effectively incorporated into risk assessment systems, hindering the accuracy and sophistication of risk identification.

[0005] In terms of early warning strategies, existing warning systems employ a one-size-fits-all approach, which presents significant flaws. The system employs a single human-machine interaction model and lacks multimodal interaction, hindering drivers from quickly and accurately acquiring risk information. Driver characteristics are not considered, and warning thresholds cannot be adjusted based on driver behavior and physiological status, leading to excessive alerts or insufficient responses. Furthermore, the system lacks adaptability to different vehicle types and lacks a correlation model between vehicle dynamics parameters and risk calibration, making it incapable of accurately assessing the risks of various types of vehicles.

[0006] Furthermore, in complex scenarios, Beidou satellite signals are susceptible to interference, leading to increased positioning errors. Existing positioning accuracy compensation solutions often rely on single-use inertial navigation or visual SLAM, which suffer from issues such as sensor drift accumulation and environmental perception blind spots. These solutions cannot effectively complement Beidou positioning, severely impacting the reliability and accuracy of road risk monitoring. Summary of the Invention

[0007] The purpose of the present invention is to solve the technical problems of rough data fusion and generalized warning strategies in the prior art, and to provide a road risk classification warning method and system based on the Beidou satellite system.

[0008] In order to achieve the above object, the present invention adopts the following technical solutions:

[0009] A first aspect of the present invention provides a road risk classification warning method based on the Beidou satellite system, comprising the following steps:

[0010] Based on real-time Beidou dual-frequency signals, inertial navigation data, roadside unit differential data, and on-board sensor data, the vehicle is positioned to obtain its real-time position.

[0011] Integrate real-time vehicle location with weather data, onboard OBD parameters, and social media sentiment to generate a dynamic risk factor matrix;

[0012] Based on expert experience, the fuzzy weights of each risk factor in the dynamic risk factor matrix are obtained through a pre-built fuzzy rule library;

[0013] Use the LSTM model to predict the time series weights of each risk factor in the dynamic risk factor matrix;

[0014] Dynamically weight the fuzzy weight and the temporal weight to obtain the final score of road risk;

[0015] Based on the final road risk score and driver profile, a graded warning of road risks is issued;

[0016] According to the warning effect after the graded warning, the fusion parameters, fuzzy rule base or LSTM model parameters are updated.

[0017] Furthermore, the vehicle is positioned to obtain the real-time position of the vehicle, specifically:

[0018] Based on the real-time Beidou dual-frequency signals, inertial navigation data, roadside unit differential data, and on-board sensor data, the Kalman filter is used to dynamically adjust the covariance matrix to obtain the real-time vehicle position;

[0019] Observe the real-time position of the vehicle to obtain the observation vector; and dynamically adjust the weight of each observation vector according to the Beidou signal carrier-to-noise ratio;

[0020] The roadside unit differential data is used to compensate the Beidou positioning results.

[0021] Furthermore, the generation of the dynamic risk factor matrix is ​​specifically as follows:

[0022] The meteorological grid data is converted from the ENU coordinate system to the vehicle coordinate system, and the vehicle position is matched by bilinear interpolation;

[0023] Temporal alignment was performed using cubic spline interpolation to generate a dynamic risk factor matrix.

[0024] Furthermore, based on expert experience, the fuzzy weights of each risk factor in the dynamic risk factor matrix are obtained through a pre-built fuzzy rule base, specifically:

[0025] Perform fuzzy processing on each risk factor in the dynamic risk factor matrix and assign a corresponding membership function to each risk factor;

[0026] According to the membership function corresponding to each risk factor and expert experience, the fuzzy weights of each risk factor in the dynamic risk factor matrix in the pre-built fuzzy rule base are adjusted.

[0027] Furthermore, the LSTM model is used to predict the time series weights of each risk factor in the dynamic risk factor matrix, specifically:

[0028] The real-time vehicle location, weather data, and on-board OBD parameters are input into the pre-trained LSTM model to obtain the time series weights of each risk factor.

[0029] Furthermore, the fuzzy weight and the temporal weight are dynamically weighted to obtain the final score of the road risk, which is specifically:

[0030] Dynamically weight the fuzzy weight and the time series weight to obtain the final weight of each risk factor;

[0031] The final score is obtained based on the final weight of each risk factor.

[0032] Furthermore, the final weights of the risk factors are:

[0033] w i (t) = α·w Fuzzy,i (t)+(1-α)·w LSTM,i (t)

[0034] Among them, w i (t) The final weight of the i-th risk factor at time t; w Fuzzy,i (t) is the fuzzy weight; w LSTM,i (t) is the temporal weight; α is the learnable fusion coefficient.

[0035] Furthermore, the final score is:

[0036]

[0037] Among them, f i is the normalized function of the ith risk factor; M is the total number of risk factors; x i (t) represents the i-th risk factor.

[0038] A second aspect of the present invention provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the road risk classification warning method based on the Beidou satellite system.

[0039] A third aspect of the present invention provides a road risk classification warning system based on the Beidou satellite system, comprising:

[0040] The vehicle location acquisition module locates the vehicle and obtains its real-time location based on real-time Beidou dual-frequency signals, inertial navigation data, roadside unit differential data, and on-board sensor data;

[0041] A dynamic risk factor matrix construction module that integrates real-time vehicle location with meteorological data, onboard OBD parameters, and social media sentiment to generate a dynamic risk factor matrix;

[0042] The risk factor fuzzy weight calculation module obtains the fuzzy weights of each risk factor in the dynamic risk factor matrix through a pre-built fuzzy rule library based on expert experience;

[0043] The risk factor time series weight calculation module uses the LSTM model to predict the time series weights of each risk factor in the dynamic risk factor matrix;

[0044] The weight fusion module dynamically weights the fuzzy weights and the temporal weights to obtain the final road risk score;

[0045] The graded warning module provides graded warnings for road risks based on the final road risk score and driver profile;

[0046] The feedback optimization module updates the fusion parameters, fuzzy rule base or LSTM model parameters according to the warning effect after the graded warning.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] This invention discloses a road risk classification and early warning method based on the Beidou satellite system. By integrating Beidou dual-frequency signals, inertial navigation data, roadside unit differential data, and on-board sensor data, this method overcomes the limitations of a single positioning method, effectively eliminating interference such as signal obstruction and multipath effects. This method achieves high-precision positioning of vehicles at the centimeter or even millimeter level, accurately capturing the vehicle's position in real time and providing a reliable foundation for subsequent risk assessment. The method deeply integrates the vehicle's real-time location with meteorological data, on-board optical boresight (OBD) parameters, and social media public opinion to generate a dynamic risk factor matrix. This process incorporates multi-dimensional information such as weather changes, vehicle conditions, and social media feedback, comprehensively capturing potential risks in the road environment, timely detecting dynamic changes in risk factors, and avoiding risk misjudgments due to information loss. A fuzzy rule base is constructed based on expert experience to obtain the fuzzy weights of each risk factor in the dynamic risk factor matrix, ensuring a professional and rational risk factor assessment. Furthermore, an LSTM model is used to predict the temporal weights of each risk factor and explore the patterns of risk factor changes over time. The dynamic weighting of fuzzy and temporal weights makes road risk scoring more realistic and improves the accuracy and scientific nature of risk assessment. Based on the final road risk score and combined with driver profiles, a graded warning system is implemented, taking into account individual factors such as driving habits, experience, and emergency response capabilities. This differentiated warning mechanism provides more targeted risk alerts for different drivers, enhancing the effectiveness and practicality of warnings, helping drivers take proactive countermeasures, and reducing the probability of accidents. Based on the actual results of the graded warnings, the fused parameters, fuzzy rule base, or LSTM model parameters are updated, forming a closed-loop "warning-feedback-optimization" system. Through continuous iterative optimization, the system can adapt to the road risk assessment needs of different regions and scenarios, continuously improving the accuracy and reliability of warnings and ensuring road traffic safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0050] Figure 1 This is a flowchart of the road risk classification warning method based on the Beidou satellite system of the present invention;

[0051] Figure 2 This is a data preprocessing flow chart of the present invention;

[0052] Figure 3 This is a flow chart of the risk score calculation of the present invention;

[0053] Figure 4 This is the hierarchical warning decision-making process of the present invention;

[0054] Figure 5 This is a closed-loop flow chart for feedback optimization of the present invention. DETAILED DESCRIPTION

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0056] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0057] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0058] The present invention is described in further detail below with reference to the accompanying drawings:

[0059] See also Figure 1-Figure 5 The present invention provides a road risk classification warning method based on the Beidou satellite system, comprising the following steps:

[0060] S1, first, collects multi-source heterogeneous data such as Beidou dual-frequency signals, roadside unit (RSU) differential data, inertial navigation (INS) and on-board sensors, and uses anti-interference spatiotemporal compensation algorithms to solve the positioning drift problem in complex scenarios. That is, based on the adaptive Kalman filter to dynamically correct the Beidou observation weights, combined with the RSU geographical attenuation compensation model, the positioning error is controlled within 0.2 meters in signal-blocked areas such as tunnels.

[0061] By integrating BeiDou-3 dual-frequency signals, roadside RSU differential data, and inertial navigation (INS), an anti-interference spatiotemporal compensation algorithm is designed, achieving a positioning error of ≤0.2m in scenarios such as tunnels and urban canyons. When BeiDou signals are interfered with by multipath effects, ionospheric delay, or obstruction, the anti-interference spatiotemporal compensation algorithm achieves sub-meter positioning accuracy compensation by integrating BeiDou dual-frequency observations, inertial navigation system (INS), and roadside unit (RSU) differential data. Specifically:

[0062] S101 uses an adaptive Kalman filter (AKF) to dynamically adjust the noise covariance matrix to suppress interference. The state equation of the adaptive Kalman filter is:

[0063]

[0064] Among them, the state vector x=[longitude, latitude, elevation, F is the state transfer matrix, and Q is the process noise covariance.

[0065] Observation equation:

[0066]

[0067] Observation vector z k Including Beidou pseudorange / carrier phase observation values, INS estimated position (angular velocity + acceleration integral) and RSU differential correction values.

[0068] S102, dynamically adjust the observation noise according to the BeiDou signal carrier-to-noise ratio (CNR) to adjust the weight of each observation vector:

[0069]

[0070] CNR th The CNR is a threshold (e.g., 45dB-Hz). When the CNR is lower than the threshold, the BeiDou observation weight is reduced and INS / RSU data is used first.

[0071] S103, RSU differential compensation term, using the differential correction value Δp broadcast by the roadside unit RSU , correct the Beidou positioning results:

[0072] p corrected =p GNSS +Δp RSU W geo

[0073] Among them, W geo is the geographical attenuation weight matrix, which decays exponentially as the distance between the vehicle and the RSU increases.

[0074] S2 then performs data preprocessing. A spatiotemporal alignment matrix is ​​used to overcome the spatiotemporal benchmarking barriers of multimodal data. In the spatial dimension, ENU coordinate system transformation and bilinear interpolation are used to achieve millimeter-level matching of meteorological grid data and vehicle positions. In the temporal dimension, cubic spline interpolation is used to align asynchronous data streams, generating a highly consistent dynamic risk factor matrix.

[0075] By building a "Beidou + 5G + Edge Computing" hybrid architecture, we synchronize Beidou positioning data, weather radar (rainfall / visibility), vehicle OBD parameters (tire pressure, brake status), and social media public opinion (accident reporting) at the millisecond level through a spatiotemporal alignment matrix to generate a dynamic risk factor matrix. The following is an introduction to the spatiotemporal alignment matrix model:

[0076] Resolve the differences in time and space benchmarks among Beidou positioning data (time and space coordinate system), meteorological data (gridized area), and on-board sensors (vehicle body coordinate system), and achieve millimeter-level synchronization of multi-source data.

[0077] S201, constructing a spatial alignment matrix, specifically: aligning the meteorological grid data (latitude and longitude grid) with the vehicle body coordinates (front / right / sky direction):

[0078]

[0079] in, is the rotation matrix from the ENU coordinate system to the vehicle body coordinate system, p origin is the origin of the local coordinate system.

[0080] Perform bilinear interpolation on meteorological data (such as rainfall intensity) to match the vehicle's current position:

[0081]

[0082] Where (x, y) is the current position of the vehicle, I ij The neighboring grid point data.

[0083] S202, build a time alignment mechanism, specifically: Timestamp synchronization: align the clocks of each device based on the PTP precise time protocol, and use cubic spline interpolation for asynchronous data:

[0084]

[0085] Among them, B i (t) is the basis function, D i is discrete time point data.

[0086] S203, constructing a spatiotemporal unified matrix. Specifically, defining the spatiotemporal alignment matrix M align Integrated spatial transformation and temporal interpolation:

[0087] D aligned =M align ·D raw =T space ·T time ·D raw

[0088] Among them, T timeis the spatial transformation matrix, T time is the time interpolation matrix.

[0089] S3 dynamically calibrates the generated dynamic risk factors, with a hybrid model of fuzzy logic and LSTM playing a key role. Fuzzy logic is embedded in expert rules such as landslide probability and visibility deterioration. LSTM captures the cumulative effect of continuous rainfall on geological risk. The learnable coefficient α dynamically balances the weighted outputs of rule reasoning and data prediction, improving the accuracy of risk scoring. A hybrid model of fuzzy logic and LSTM is used to dynamically adjust risk factor weights based on real-time data streams (e.g., increasing visibility weight during heavy rain and accident probability weight on congested roads). Specifically:

[0090] S301: Perform fuzzy processing on each risk factor (rainfall intensity, traffic volume, Beidou positioning error, etc.) and define a membership function. For example, the rainfall intensity membership function (triangular function) is:

[0091]

[0092] Define dynamic weight adjustment rules (IF-THEN form) based on expert experience, for example:

[0093] Rule 1: IF the rainfall intensity is "heavy" AND the visibility is "low" THEN the visibility weight is increased to 0.4;

[0094] Rule 2: IF the traffic volume is “high” AND the frequency of sudden braking is “high” THEN the collision risk weight is increased to 0.6.

[0095] S302, input time series data: Beidou positioning trajectory, vehicle OBD parameters (vehicle speed, brake pressure), meteorological data and other time series, and output the time series weight of each risk factor.

[0096] LSTM unit state update, specifically:

[0097] f t =σ(W f ·[h t-1 ,x t ]+b f )(Forget Gate)

[0098] i t =σ(W i ·[h t-1 ,x t ]+b i )(Input gate)

[0099] (Candidate Memory)

[0100] (Memory Update)

[0101] 0 t =σ(W o ·[h t-1 ,x t ]+b o )(Output gate)

[0102] h t =o t ⊙tanh(C t )(Hidden state output)

[0103] Among them, W * ,b * is a trainable parameter, σ is the sigmoid function, and ⊙ is the Hadamard product.

[0104] S303, dynamically weighting the fuzzy weight and the time series weight to obtain the final weight of each risk factor. Specifically, the dynamic weight allocation formula is:

[0105] w i (t) = α·w Fuzzy,i (t)+(1-α)·w LSTM,i (t)

[0106] Among them, w i (t) The final weight of the i-th risk factor (such as visibility, traffic flow) at time t; w Fuzzy,i (t) is the weight of the fuzzy logic output (based on the rule base); w LSTM,i (t) is the weight of LSTM prediction (based on time series features); α is the learnable fusion coefficient (the initial value is set by expert experience and dynamically optimized during training).

[0107] Based on the final weight of each risk factor, the final score is obtained:

[0108]

[0109] where f i is the normalized function of the i-th risk factor (e.g., mapping rainfall intensity to a 0-1 risk value); N is the total number of risk factors.

[0110] S4, based on the final score, makes graded warning decisions for road risks and implements precise intervention based on a multimodal graded warning strategy: constructs a driver risk profile based on reinforcement learning, and dynamically adjusts the threshold to trigger AR-HUD three-dimensional risk projection or vehicle active speed limit.

[0111] S401: Driver profile matching uses a dynamic risk threshold adaptive mechanism, specifically:

[0112] Generate risk tolerance labels based on historical driving data (such as nighttime driving preferences and frequency of sudden acceleration) and dynamically adjust warning thresholds (such as increasing the collision probability threshold for aggressive drivers to trigger warnings by 20%).

[0113] Extract multidimensional feature vectors from historical driving data:

[0114] Indicator Category Driving behavior aggressiveness Visual sensitivity Risk appetite Physiological state Behavioral indicators Rapid acceleration frequency Proportion of night driving time Average following distance - Physiological indicators - - - Heart rate variability (HRV)

[0115] Reinforcement learning uses a dual deep Q network (DDQN) to dynamically optimize the warning threshold:

[0116]

[0117] Among them, state s is the current driving environment (such as visibility and Beidou positioning risk level); action a is to adjust the warning threshold (±5% to ±25%); reward r is a weighted score of the driver's response time and comfort feedback.

[0118] S402, multi-modal graded warning output uses AR-HUD projection technology, specifically:

[0119] Mapping Beidou coordinates (vehicle position) to AR space, establishing a 3D geospatial registration model, and mapping Beidou latitude and longitude (BDS-84 coordinate system) to the windshield field of view in real time:

[0120]

[0121] Among them, K is the camera intrinsic parameter matrix; R enu2cam is the rotation matrix from the ENU coordinate system to the AR camera; p BDS Beidou positioning coordinates.

[0122] S5. Finally, feedback optimization is performed on the fused parameters, fuzzy rule base or LSTM model parameters; the α coefficient and rule base are optimized in real time through the online learning mechanism to form a self-evolution closed loop of "risk identification-warning execution-effect evaluation". Through the synergistic effect of the four major innovations of positioning enhancement, hybrid modeling, spatiotemporal alignment, and multimodal interaction, the three major industry pain points of poor dynamic adaptability, data silos, and high false alarm rate in traditional methods are systematically solved. Significantly improve the long-term stability and scenario adaptability of dynamic risk calibration. Feedback optimization first collects multi-dimensional verification data, including warning false alarm rate, driver response delay, vehicle active control effect (such as braking distance deviation) and the matching degree between the actual accident position and the predicted coordinates, and constructs a feedback feature vector with false alarm rate, human-computer interaction efficiency, and positioning accuracy attenuation coefficient as the core. Feedback feature vector example:

[0123]

[0124] Based on the feedback feature vector, a hybrid learning architecture is used to achieve multi-level optimization. At the parameter level, Bayesian optimization is used to dynamically adjust the fusion coefficient α of fuzzy logic and LSTM, weighing the contribution of expert rules and data-driven methods.

[0125] The online parameter search model based on Bayesian optimization is as follows:

[0126]

[0127] Among them, λ is the penalty factor, which is automatically calibrated through historical feedback data.

[0128] This embodiment provides an example scenario: a sudden landslide on a mountain road after continuous rainfall to illustrate the implementation results of the present invention.

[0129] Example input data table:

[0130]

[0131] Example of model output table:

[0132]

[0133] One embodiment of the present invention provides a road risk classification warning system based on the Beidou satellite system, comprising:

[0134] The vehicle location acquisition module locates the vehicle and obtains its real-time location based on real-time Beidou dual-frequency signals, inertial navigation data, roadside unit differential data, and on-board sensor data;

[0135] A dynamic risk factor matrix construction module that integrates real-time vehicle location with meteorological data, onboard OBD parameters, and social media sentiment to generate a dynamic risk factor matrix;

[0136] The risk factor fuzzy weight calculation module obtains the fuzzy weights of each risk factor in the dynamic risk factor matrix through a pre-built fuzzy rule library based on expert experience;

[0137] The risk factor time series weight calculation module uses the LSTM model to predict the time series weights of each risk factor in the dynamic risk factor matrix;

[0138] The weight fusion module dynamically weights the fuzzy weights and the temporal weights to obtain the final road risk score;

[0139] The graded warning module provides graded warnings for road risks based on the final road risk score and driver profile;

[0140] The feedback optimization module updates the fusion parameters, fuzzy rule base or LSTM model parameters according to the warning effect after the graded warning.

[0141] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory). The computer-readable storage medium is a memory device in a terminal device for storing programs and data. It is understood that the computer-readable storage medium herein may include both built-in storage media in the terminal device and, of course, extended storage media supported by the terminal device. It may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device, or component. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for being loaded and executed by a processor. These instructions may be one or more computer programs (including program code). It should be noted that more specific examples (a non-exhaustive list) of computer-readable storage media herein include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0142] Computer-readable storage media also include data signals propagated in baseband or as part of a carrier wave, which carry readable program code. Such propagated data signals can take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The readable storage medium can also be any readable medium other than a readable storage medium, which can send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, device, or device. The program code contained on the readable storage medium can be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above.

[0143] The program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, and the like, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a stand-alone software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0144] The processor may load and execute one or more instructions stored in a computer-readable storage medium to implement the corresponding steps of the road risk classification warning method based on the Beidou satellite system in the above embodiment.

[0145] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A road risk classification warning method based on the Beidou satellite system, characterized in that: The following steps are involved: Based on real-time Beidou dual-frequency signals, inertial navigation data, roadside unit differential data, and on-board sensor data, the vehicle is positioned to obtain its real-time position. Integrate real-time vehicle location with weather data, onboard OBD parameters, and social media sentiment to generate a dynamic risk factor matrix; Based on expert experience, the fuzzy weights of each risk factor in the dynamic risk factor matrix are obtained through a pre-built fuzzy rule library; Use the LSTM model to predict the time series weights of each risk factor in the dynamic risk factor matrix; Dynamically weight the fuzzy weight and the temporal weight to obtain the final score of road risk; Based on the final road risk score and driver profile, a graded warning of road risks is issued; According to the warning effect after the graded warning, the fusion parameters, fuzzy rule base or LSTM model parameters are updated.

2. The road risk classification warning method based on the BeiDou satellite system according to claim 1 is characterized in that: Positioning the vehicle to obtain the real-time position of the vehicle is specifically as follows: Based on the real-time Beidou dual-frequency signals, inertial navigation data, roadside unit differential data, and on-board sensor data, the Kalman filter is used to dynamically adjust the covariance matrix to obtain the real-time vehicle position; Observe the real-time position of the vehicle to obtain the observation vector; and dynamically adjust the weight of each observation vector according to the Beidou signal carrier-to-noise ratio; When the Beidou signal carrier-to-noise ratio decreases, the roadside unit differential data is used to compensate the Beidou positioning results.

3. The road risk classification warning method based on the BeiDou satellite system according to claim 1 is characterized in that: The generating of the dynamic risk factor matrix is ​​specifically as follows: The meteorological grid data is converted from the ENU coordinate system to the vehicle coordinate system, and the vehicle position is matched by bilinear interpolation; Temporal alignment was performed using cubic spline interpolation to generate a dynamic risk factor matrix.

4. The road risk classification warning method based on the BeiDou satellite system according to claim 1 is characterized in that: Based on expert experience, the fuzzy weights of each risk factor in the dynamic risk factor matrix are obtained through a pre-built fuzzy rule base, specifically: Perform fuzzy processing on each risk factor in the dynamic risk factor matrix and assign a corresponding membership function to each risk factor; According to the membership function corresponding to each risk factor and expert experience, the fuzzy weights of each risk factor in the dynamic risk factor matrix in the pre-built fuzzy rule base are adjusted.

5. The road risk classification warning method based on the BeiDou satellite system according to claim 1 is characterized in that: The LSTM model is used to predict the time series weights of each risk factor in the dynamic risk factor matrix, specifically: The real-time vehicle location, weather data, and on-board OBD parameters are input into the pre-trained LSTM model to obtain the time series weights of each risk factor.

6. The road risk classification warning method based on the BeiDou satellite system according to claim 1 is characterized in that: The fuzzy weight and the temporal weight are dynamically weighted to obtain the final road risk score, which is specifically: Dynamically weight the fuzzy weight and the time series weight to obtain the final weight of each risk factor; The final score is obtained based on the final weight of each risk factor.

7. The road risk classification warning method based on the BeiDou satellite system according to claim 6 is characterized in that: The final weights of the various risk factors are: w i (t)=α·w Fuzzy,i (t)+(1-α)·w LSTM,i (t) Among them, w i (t) The final weight of the i-th risk factor at time t; w Fuzzy,i (t) is the fuzzy weight; w LSTM,i (t) is the temporal weight; α is the learnable fusion coefficient.

8. The road risk classification warning method based on the BeiDou satellite system according to claim 7 is characterized in that: The final rating is: Among them, f i is the normalized function of the ith risk factor; N is the total number of risk factors; x i (t) represents the i-th risk factor.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the road risk classification warning method based on the Beidou satellite system according to any one of claims 1 to 8 is implemented.

10. A road risk classification warning system based on the Beidou satellite system, characterized in that: include: The vehicle location acquisition module locates the vehicle and obtains its real-time location based on real-time Beidou dual-frequency signals, inertial navigation data, roadside unit differential data, and on-board sensor data; A dynamic risk factor matrix construction module that integrates real-time vehicle location with meteorological data, onboard OBD parameters, and social media sentiment to generate a dynamic risk factor matrix; The risk factor fuzzy weight calculation module obtains the fuzzy weights of each risk factor in the dynamic risk factor matrix through a pre-built fuzzy rule library based on expert experience; The risk factor time series weight calculation module uses the LSTM model to predict the time series weights of each risk factor in the dynamic risk factor matrix; The weight fusion module dynamically weights the fuzzy weights and the temporal weights to obtain the final road risk score; The graded warning module provides graded warnings for road risks based on the final road risk score and driver profile; The feedback optimization module updates the fusion parameters, fuzzy rule base or LSTM model parameters according to the warning effect after the graded warning.

Images