A road surface wetness intelligent monitoring method, an electronic device and a storage medium

Abstract

The application relates to the technical field of road engineering and traffic safety monitoring, and specifically provides a road surface wet and slippery intelligent monitoring method, an electronic device and a storage medium, the method comprising the following steps: acquiring a road surface environment feature of a to-be-measured road section; collecting tire-road interaction force features, tire state features and motion state information in real time; performing window processing on a multi-source data set according to a preset time window or a preset distance window to obtain a plurality of road section-level analysis units; integrating the road surface environment feature, preset water film thickness corresponding to a preset wet and slippery working condition, preset friction detection unit working condition parameters and the kinematic features, friction statistical features and tire state features corresponding to the unit into a window-level fusion feature vector corresponding to the unit; generating an estimated friction coefficient and a risk assessment result corresponding to each road section-level analysis unit according to the window-level fusion feature vector, and then generating a safe operation suggestion parameter according to the estimated friction coefficient; and the method can accurately reflect the actual wet and slippery state of a road surface under the coupling action of multiple factors.

Description

Technical Field

[0001] This application relates to the field of road engineering and traffic safety monitoring technology, and more specifically, to an intelligent monitoring method, electronic device and storage medium for slippery road surfaces. Background Technology

[0002] Slippery road surfaces are one of the core adverse factors affecting road operation safety. Natural conditions such as precipitation, icing, and condensation on the road surface can lead to water accumulation, snow accumulation, or ice formation on the road surface, changing the original friction characteristics of the road surface, significantly reducing the road surface's anti-skid ability, and causing safety risks such as increased braking distance and vehicle skidding and rollover. Accurately grasping the actual slippery state of the road surface and making reasonable assessments of its safety risks are prerequisites for carrying out road slippery safety early warning, operation management, and emergency response, and are of great significance for ensuring road traffic safety.

[0003] Existing technologies for monitoring and assessing road surface slipperiness have developed along two main paths: one focuses on detecting and predicting the anti-skid performance of road surfaces, while the other focuses on analyzing vehicle operational safety. The former typically evaluates the anti-skid status of road surfaces based on indicators such as texture parameters, friction coefficients, or water film thickness; the latter analyzes operational safety issues under slippery conditions from the perspectives of vehicle dynamic response, friction demand and supply relationships, or risk boundaries.

[0004] However, existing technologies still have the following shortcomings: First, existing methods generally lack a closed-loop update mechanism based on continuous on-site friction data. They often rely on empirical regression, indirect inference, or offline analysis. In actual monitoring, the road surface slipperiness is affected by a combination of factors, such as road surface texture, ambient temperature, water film thickness, and the load, driving speed, and tire condition of the detection equipment. Relying solely on empirical parameters or indirect inference is insufficient to reflect the actual slipperiness under the combined effect of multiple factors, and the monitoring results often deviate significantly from the actual situation.

[0005] Secondly, existing methods require the use of multi-source heterogeneous data from different acquisition devices. These data from different sources usually differ in sampling frequency, time reference, and spatial location, making it difficult to achieve accurate spatiotemporal alignment. This can easily lead to deviations in the fusion analysis results, affecting the stability and reliability of the final monitoring results.

[0006] Third, the output of existing methods mostly stays at the level of single parameters such as friction coefficient, texture index or risk probability. These parameters cannot be directly used by road operation management departments or drivers and passengers, and are difficult to be directly converted into risk levels that can serve road operation management and safety early warning or operation control suggestions that can be referenced. Therefore, their engineering application value is limited.

[0007] There is currently no effective technical solution to the above problems. Summary of the Invention

[0008] The purpose of this application is to provide a method, electronic device and storage medium for intelligent monitoring of road surface slipperiness, which can accurately reflect the actual slippery state of the road surface under the coupled effects of multiple factors.

[0009] In a first aspect, this application provides a method for intelligent monitoring of slippery road surfaces, which includes the following steps: S1. Obtain the road surface environmental characteristics of the road segment to be tested; the road surface environmental characteristics include road surface texture characteristics and road surface temperature characteristics; S2. When the road section to be tested is under preset wet and slippery conditions, the friction detection unit is moved by a mobile carrier to collect tire road force characteristics, tire condition characteristics and motion state information in real time. S3. Perform time alignment processing on tire road force characteristics, tire state characteristics and motion state information to obtain a multi-source dataset under a unified time axis. Then, perform windowing processing on the multi-source dataset according to a preset time window or a preset distance window to obtain multiple road segment-level analysis units. The multi-source dataset includes multiple sets of tire road force characteristics, tire state characteristics and motion state information. S4. For each road segment-level analysis unit, obtain the kinematic features corresponding to the unit based on all motion state information corresponding to the unit, and obtain the friction statistical features corresponding to the unit based on all tire-road force features corresponding to the unit. Then, integrate the road surface environment features, the preset water film thickness corresponding to the preset wet and slippery working conditions, the preset friction detection unit working condition parameters, and the kinematic features, friction statistical features, and tire state features corresponding to the unit into the window-level fusion feature vector corresponding to the unit. S5. Using the pre-built fusion model, generate the estimated friction coefficient and risk assessment results for each road segment-level analysis unit based on the window-level fusion feature vector, and then generate the recommended safe operation parameters for each road segment-level analysis unit based on the estimated friction coefficient.

[0010] Optionally, step S2 includes: S21. Use surface water film spraying equipment to wet the road section to be tested in order to reduce the absorption of water film by the road surface. S22. Determine the equipment forward speed based on the preset equivalent water film thickness and preset spray flow rate corresponding to the preset wet and slippery working conditions. S23. Install a water-receiving component at the end of the road surface water film spraying equipment, control the road surface water film spraying equipment to move forward at the equipment forward speed in road sections other than the road section to be tested, and control the road surface water film spraying equipment to spray water film in road sections other than the road section to be tested at a preset spraying flow rate. S24. Obtain the water quality of the water-receiving component and calculate the actual water film thickness based on the water quality. S25. Analyze whether the absolute value of the deviation between the actual water film thickness and the preset equivalent water film thickness is less than the preset threshold. If yes, proceed to step S26. If no, generate an alarm message. S26. Control the road surface water film spraying equipment to move forward on the road section to be tested according to the equipment's forward speed, and control the road surface water film spraying equipment to spray water film on the road section to be tested at a preset spraying flow rate, so that the road section to be tested is in a preset wet and slippery condition. S27. The friction detection unit is moved by a mobile carrier, and the friction detection unit is used to collect the tire road force characteristics and tire state characteristics in real time. The spatiotemporal positioning unit is used to collect motion state information synchronously.

[0011] Optionally, the formula for calculating the actual water film thickness is: ; Where h represents the actual water film thickness, m represents the mass of water received, ρ represents the density of water, and A represents the preset bottom area corresponding to the water-receiving component.

[0012] Optionally, the motion state information is collected by a spatiotemporal positioning unit, which includes a BeiDou inertial navigation component. Each first timestamp corresponds to a piece of motion state information, and each second timestamp corresponds to a tire road force feature and a tire state feature. Step S3 includes: S31. Perform data cleaning and filtering on all tire road force characteristics and all tire condition characteristics; S32. Remove motion state information collected when the BeiDou inertial navigation component is in a disconnection segment, saturation segment, unit abnormal segment, positioning status degradation, number of data acquisition satellites less than the preset number, integrated navigation loss of lock, speed jump or attitude change. S33. Establish the correspondence between the first timestamp and the second timestamp to coarsely align the tire road force characteristics, tire state characteristics, and motion state information. S34. Perform interpolation and / or resampling alignment on the coarsely aligned tire road force features, tire state features and motion state information to obtain a multi-source dataset under a unified time axis. S35. Calculate the real-time mileage of the friction detection unit relative to the starting point of the road segment under test based on the motion state information; S36. Divide the real-time mileage into multiple windows according to the preset distance window to perform windowing processing on the multi-source dataset; each window corresponds to a real-time mileage interval and a road segment-level analysis unit, and each road segment-level analysis unit includes all tire-road force characteristics, all tire state characteristics and all motion state information within the corresponding real-time mileage interval.

[0013] Optionally, the motion state information includes motion velocity, lateral acceleration, and attitude information; the tire road force characteristics include traction direction horizontal force and normal load; step S4 includes: S41. For each segment-level analysis unit, calculate the window average velocity based on all motion velocities corresponding to the unit, calculate the window average lateral acceleration based on all lateral accelerations corresponding to the unit, and calculate the attitude change based on all attitude information corresponding to the unit. Then, integrate the window average velocity, window average lateral acceleration, and attitude change into the kinematic characteristics corresponding to the unit. S42. For each road segment-level analysis unit, calculate the average friction coefficient and friction coefficient fluctuation index corresponding to the unit based on the horizontal force in all traction directions and all normal loads corresponding to the unit, and then integrate the average friction coefficient and friction coefficient fluctuation index into the friction statistical characteristics corresponding to the unit. S43. For each road segment-level analysis unit, integrate the road surface environmental features, the preset water film thickness corresponding to the preset wet and slippery working conditions, the preset friction detection unit working condition parameters, and the kinematic features, friction statistical features, and tire state features corresponding to the unit into a window-level fusion feature vector corresponding to the unit.

[0014] Optionally, the formula for calculating the average friction coefficient is: ; Where, μ meas (k) represents the average friction coefficient corresponding to the kth road segment-level analysis unit, N k W represents the number of valid sampling points within the k-th road segment-level analysis unit. k F represents the k-th segment-level analysis unit. x (i) represents the horizontal traction force at the i-th valid sampling point within the k-th road segment-level analysis unit, F. z (i) represents the normal load of the i-th effective sampling point within the k-th road segment-level analysis unit; The formula for calculating the friction coefficient fluctuation index is: ; Where, σ μ (k) represents the friction coefficient fluctuation index corresponding to the kth road segment-level analysis unit.

[0015] Optionally, step S2 includes: S21' When the road section to be tested is under preset wet and slippery conditions, the friction detection unit is moved by a mobile carrier to collect tire road force characteristics, tire state characteristics, spatiotemporal positioning information and motion state information in real time. Step S3 includes: S31': Time alignment processing is performed on tire road force characteristics, tire state characteristics, spatiotemporal positioning information, and motion state information to obtain a multi-source dataset under a unified time axis. Then, the multi-source dataset is windowed according to a preset time window or a preset distance window to obtain multiple road segment-level analysis units. The multi-source dataset includes multiple sets of tire road force characteristics, tire state characteristics, spatiotemporal positioning information, and motion state information. Step S5 includes: S51. Using a pre-built fusion model, generate the estimated friction coefficient and risk assessment results for each road segment-level analysis unit based on the window-level fusion feature vector; the risk assessment results include risk level and confidence level. S52. Remove road segment-level analysis units with confidence levels lower than the preset confidence level, estimate friction coefficients and risk assessment results; S53. Generate safe operation recommendation parameters for each road segment-level analysis unit based on the estimated friction coefficient, and bind the safe operation recommendation parameters with the spatiotemporal positioning information corresponding to the parameters.

[0016] Optionally, the recommended parameters for safe operation include recommended braking deceleration and recommended operating speed. The formula for calculating the recommended braking deceleration is as follows: ; Among them, a b This indicates a suggestion to brake and decelerate. This represents the estimated coefficient of friction, and g represents the acceleration due to gravity. The recommended formula for calculating running speed is: ; in, Indicates the suggested running speed, t r D represents the preset driving reaction time. max This indicates the preset parking distance.

[0017] Secondly, this application provides an electronic device including a processor and a memory, the memory storing computer-readable instructions, which, when executed by the processor, perform the steps of the method provided in the first aspect above.

[0018] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the method provided in the first aspect above.

[0019] As can be seen from the above, the intelligent monitoring method, electronic device and storage medium for road surface slipperiness provided in this application integrates road surface environmental information, operating parameters and multi-source measured data, extracts fusion features through time-aligned windowing processing, and finally generates directly usable monitoring results and safe operation suggestion parameters. It solves the problems of large deviation in monitoring results, insufficient spatiotemporal alignment accuracy of multi-source data and difficulty in directly applying the output results in the prior art. It can accurately reflect the actual slippery state of the road surface under the coupling effect of multiple factors, improve the accuracy and reliability of monitoring results, and the output results can directly serve road operation management and safety early warning, and have high engineering application value. Attached Figure Description

[0020] Figure 1 A flowchart of an intelligent monitoring method for slippery road surfaces provided in this application embodiment.

[0021] Figure 2 The figure shows the test results for water film thickness, vehicle load, driving speed, and single wheel deflection angle and their corresponding friction coefficient fluctuations under summer conditions, provided for embodiments of this application.

[0022] Figure 3 The figure shows the test results for water film thickness, vehicle load, driving speed, and single wheel deflection angle under winter conditions and their corresponding friction coefficient fluctuations, provided for embodiments of this application.

[0023] Figure 4 This is a schematic diagram of an electronic device structure provided in an embodiment of this application.

[0024] Reference numerals: 101, processor; 102, memory; 103, communication bus. Detailed Implementation

[0025] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0026] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0027] In a first aspect, this application provides a method for intelligent monitoring of slippery road surfaces, which includes the following steps: S1. Obtain the road surface environmental characteristics of the road segment to be tested; the road surface environmental characteristics include road surface texture characteristics and road surface temperature characteristics; S2. When the road section to be tested is under the preset wet and slippery conditions, the friction detection unit is moved by a mobile carrier to collect tire road force characteristics, tire condition characteristics and motion state information in real time. S3. Perform time alignment processing on tire road force characteristics, tire state characteristics and motion state information to obtain a multi-source dataset under a unified time axis. Then, perform windowing processing on the multi-source dataset according to a preset time window or a preset distance window to obtain multiple road segment-level analysis units. The multi-source dataset includes multiple sets of tire road force characteristics, tire state characteristics and motion state information. S4. For each road segment-level analysis unit, obtain the kinematic features corresponding to the unit based on all motion state information corresponding to the unit, and obtain the friction statistical features corresponding to the unit based on all tire-road force features corresponding to the unit. Then, integrate the road surface environment features, the preset water film thickness corresponding to the preset wet and slippery working conditions, the preset friction detection unit working condition parameters, and the kinematic features, friction statistical features, and tire state features corresponding to the unit into the window-level fusion feature vector corresponding to the unit. S5. Using the pre-built fusion model, generate the estimated friction coefficient and risk assessment results for each road segment-level analysis unit based on the window-level fusion feature vector, and then generate the recommended safe operation parameters for each road segment-level analysis unit based on the estimated friction coefficient.

[0028] To facilitate understanding, some key terms in this embodiment are explained below. The road surface environment characteristics in this embodiment refer to the physical properties of the road segment under test, including road surface texture characteristics and road surface temperature characteristics. Road surface texture characteristics describe the macroscopic and microscopic roughness and drainage performance of the road surface, such as the size, shape, and distribution of road aggregates, and the surface texture depth. Road surface temperature characteristics refer to the real-time temperature of the road surface of the road segment under test, which affects the evaporation rate of the water film and the formation of the freezing point, thus affecting the road surface's slipperiness. The preset slippery condition in this embodiment refers to a condition artificially set for the road surface to be in a slippery state during intelligent road slippery monitoring. This is usually achieved by spraying a water film of a certain thickness onto the road surface to simulate road conditions in rainy or humid environments. The purpose of the preset slippery condition is to obtain friction data and vehicle operation data of the road segment under test in a slippery state under controllable conditions. The mobile carrier in this embodiment refers to a vehicle or device capable of towing the friction detection unit along the test road segment. For example, it could be a specialized test vehicle or a regular vehicle. The function of the mobile carrier is to continuously move the friction detection unit along the test road segment, thereby monitoring the wet and slippery condition of the entire test road segment. The friction detection unit in this embodiment refers to a device used to collect the interaction force between the road surface and the test wheel in real time. This unit typically includes multiple test wheels and corresponding force sensors, capable of measuring the tire-road interaction force characteristics such as the traction direction horizontal force and normal load generated when the tire rolls or slips on a wet and slippery road surface. In addition, the friction detection unit also collects the tire's own operating status, such as tire pressure and tire temperature. The tire-road interaction force characteristics in this embodiment refer to the mechanical response generated when the tire contacts the road surface. These mainly include the traction direction horizontal force and the normal load. The traction direction horizontal force is the force generated by the interaction between the test wheel and the road surface in the direction of travel; this horizontal force reflects the longitudinal grip force provided by the road surface. The normal load is the force acting perpendicularly to the road surface on the test wheel; this load represents the pressure of the test wheel on the road surface. These mechanical characteristics are key indicators for evaluating road surface friction performance. The tire condition characteristics in this embodiment refer to the physical state parameters of the test wheel during operation, such as tire pressure and tire temperature. Tire pressure affects the tire's contact area and stiffness, thus affecting the friction performance between the tire and the road surface. Tire temperature affects the physical properties of the tire rubber (such as hardness and viscoelasticity), which also has a significant impact on the coefficient of friction. The motion state information in this embodiment refers to the dynamic parameters of the friction detection unit during movement, including motion speed, lateral acceleration, and attitude information. Motion speed reflects the speed of the vehicle's movement, lateral acceleration reflects the force on the vehicle in the lateral direction, which is related to the lateral force when the vehicle turns, and attitude information describes the spatial attitude of the vehicle, such as pitch angle, roll angle, and yaw angle. These parameters affect the contact geometry and force distribution between the tire and the road surface.The time alignment process in this embodiment refers to the process of synchronizing data from different data sources (such as friction detection units (for collecting tire road force characteristics and tire state characteristics) and spatiotemporal positioning units (for collecting motion state information)). Since different sensors may have different sampling frequencies and time bases, direct fusion can lead to data misalignment. Time alignment processes use interpolation, resampling, or other time synchronization algorithms to map all data onto a unified time axis, ensuring the consistency of different features over time. The multi-source dataset in this embodiment refers to a collection of data from multiple sources, including tire road force characteristics, tire state characteristics, and motion state information, after time alignment processing. This dataset, under a unified time axis, provides a complete and synchronized data foundation for subsequent road segment-level analysis. The road segment-level analysis unit in this embodiment refers to an independent data segment obtained by dividing the continuously collected multi-source dataset according to a preset time window or a preset distance window. Each road segment-level analysis unit represents a specific area on the road segment under test, containing all synchronously collected tire road force characteristics, tire state characteristics, and motion state information within that area, facilitating refined analysis of local road segments. The kinematic features in this embodiment refer to parameters reflecting the motion characteristics of the vehicle extracted from the motion state information corresponding to the road segment-level analysis unit. For example, window-average velocity, window-average lateral acceleration, and attitude change can be calculated. These features characterize the vehicle's driving state on a specific road segment and are crucial for understanding the vehicle's dynamic response under wet and slippery road conditions. The frictional statistical features in this embodiment refer to statistical quantities reflecting the road surface friction performance extracted from the tire-road force characteristics corresponding to the road segment-level analysis unit. For example, the average friction coefficient and friction coefficient fluctuation index can be calculated. These features directly quantify the frictional force level and stability provided by the road surface under wet and slippery conditions. The window-level fusion feature vector in this embodiment refers to a comprehensive feature vector formed by integrating road surface environmental features, preset water film thickness corresponding to preset wet and slippery conditions, preset friction detection unit operating parameters, and kinematic features, frictional statistical features, and tire state features corresponding to the road segment-level analysis unit. This vector contains multi-dimensional information affecting the road surface's wet and slippery state and vehicle operating safety, and is the basis for inputting the fusion model for wet and slippery risk assessment. The fusion model in this embodiment refers to a pre-built mathematical or machine learning model used to process window-level fused feature vectors and output estimated friction coefficients and risk assessment results. This model learns the relationship between features from a large amount of historical data and actual friction coefficients and risks, enabling it to predict and evaluate new input feature vectors. The estimated friction coefficient in this embodiment refers to the road surface friction coefficient predicted by the fusion model based on window-level fused feature vectors. This coefficient is a core indicator for quantifying the slipperiness of the road surface, reflecting the maximum friction force that the road surface can provide under current operating conditions.The risk assessment result in this embodiment refers to the assessment result of road surface slipperiness risk output by the fusion model based on the window-level fusion feature vector. It typically includes a risk level and a confidence level. The risk level classifies the degree of slipperiness into different levels for easy intuitive understanding and decision-making, while the confidence level indicates the reliability of the assessment result. The safe operation recommendation parameters in this embodiment refer to parameters generated based on the estimated friction coefficient that can be directly used to guide vehicle safe operation, such as recommended braking deceleration and recommended operating speed. These parameters transform the abstract friction coefficient into specific driving suggestions, providing practical safety guidance for road managers and drivers.

[0029] This application proposes an intelligent monitoring method for slippery road surfaces, aiming to address the shortcomings of existing slippery road surface monitoring methods in terms of data acquisition, processing, and result application.

[0030] First, the method involves acquiring the road surface environmental features of the road segment to be tested, including road surface texture features and road surface temperature features. For example, the road surface texture features can be acquired using the existing LS-40 portable 3D surface analyzer, and the road surface temperature features can be acquired using the existing DWIR infrared thermometer. These features can be acquired in a single data collection before monitoring begins.

[0031] Secondly, when the test section is under preset wet and slippery conditions, a mobile carrier moves the friction detection unit to collect tire-road force characteristics, tire condition characteristics, and motion status information in real time. For example, the mobile carrier can be a towing vehicle, and the friction detection unit is connected to the towing vehicle. When the towing vehicle pulls the friction detection unit through the test section at a preset speed, the force sensor can measure the tire-road force characteristics such as the horizontal force in the traction direction and the normal load between the test wheel and the road surface in real time. At the same time, the sensors installed on the test wheel can monitor tire condition characteristics such as tire pressure and tire temperature in real time. In addition, the vehicle's GPS or inertial navigation system can acquire motion status information such as the friction detection unit's speed, lateral acceleration, and attitude information in real time.

[0032] Next, the tire force characteristics, tire state characteristics, and motion state information are time-aligned to obtain a multi-source dataset under a unified time axis. Then, the multi-source dataset is windowed according to a preset time window or a preset distance window to obtain multiple road segment-level analysis units. For example, since different sensors may have different sampling frequencies and timestamps, it is necessary to time-align the collected raw data. This can be achieved through methods such as linear interpolation, spline interpolation, or resampling to map all data onto a unified time axis. After time alignment, the continuous data stream can be divided according to a fixed time length (e.g., a time window every 5 seconds) or a fixed distance length (e.g., a distance window every 10 meters) to obtain multiple independent road segment-level analysis units. Each analysis unit contains all time-aligned data within the corresponding time or distance interval.

[0033] Then, for each road segment-level analysis unit, the kinematic features corresponding to that unit are obtained based on all motion state information, and the friction statistical features are obtained based on all tire-road force characteristics. Finally, the road surface environmental features, the preset water film thickness corresponding to the preset wet-slip condition, the preset friction detection unit operating parameters, and the kinematic features, friction statistical features, and tire state features of that unit are integrated into a window-level fusion feature vector corresponding to that unit. For example, within a road segment-level analysis unit, the average value of all motion velocities can be calculated as the window average velocity, the average value of all lateral accelerations can be calculated as the window average lateral acceleration, and the standard deviation of attitude information can be calculated as the attitude change; these together constitute the kinematic features. Simultaneously, the average value and standard deviation of the friction coefficient can be calculated based on the traction direction horizontal force and normal load within the unit, serving as friction statistical features. Furthermore, the preset water film thickness and preset friction detection unit operating parameters are values ​​pre-set before friction detection begins; these parameters may include tire deflection angle, counterweight, or axle load. Finally, these features from different sources are integrated into a single window-level fusion feature vector.

[0034] Finally, using a pre-built fusion model, estimated friction coefficients and risk assessment results are generated for each road segment-level analysis unit based on window-level fusion feature vectors. Then, safe operation recommendations are generated for each road segment-level analysis unit based on the estimated friction coefficients. For example, the fusion model can be a model built on machine learning algorithms (such as support vector machines, neural networks, or random forests). During the training phase, this model learns the mapping relationship between feature vectors and actual friction coefficients and risks using a large amount of historical data. During actual monitoring, the window-level fusion feature vectors of each road segment-level analysis unit are input into this model, and the model outputs the estimated friction coefficients and risk assessment results (such as risk level and confidence level) for that road segment. Furthermore, safe operation recommendations, such as recommended braking deceleration and recommended operating speed, can be calculated based on the estimated friction coefficients for that road segment.

[0035] The following example provides a more detailed explanation of the above technical solution: Suppose that on a section of highway (location A) in the suburbs of a city, a 5-kilometer stretch needs to be monitored for slippery conditions using intelligent monitoring. Traditional monitoring methods may rely solely on meteorological data or single information provided by road surface sensors, making it difficult to accurately reflect the actual slippery conditions under the combined effects of multiple factors, and the output results are not readily available to guide traffic management.

[0036] To address this issue, this application proposes an intelligent monitoring method for slippery road surfaces. First, before monitoring begins, staff acquire the surface environmental characteristics of the road section to be monitored at location A. For example, an LS-40 portable 3D surface analyzer is used to scan the road surface of the section to be monitored, acquiring surface texture features such as texture depth and roughness. Simultaneously, an infrared thermometer is used to measure the road surface temperature, obtaining its characteristics.

[0037] Next, to simulate slippery conditions, a mobile vehicle equipped with a road surface water film spraying device (such as a water truck) travels at a preset speed along the test section, wetting the road surface with a preset spray flow rate to achieve a preset water film thickness (e.g., 0.8 mm), thus placing the test section under preset slippery conditions. Subsequently, another mobile vehicle towing a friction detection unit follows closely behind, passing through the test section at a set test speed (e.g., 60 km / h). During the test vehicle's movement, the friction detection unit collects real-time tire-road force characteristics (such as traction direction horizontal force and normal load) and tire condition characteristics (such as tire pressure and tire temperature), and the spatiotemporal positioning unit collects real-time motion state information (such as speed, lateral acceleration, and attitude information). This data is recorded at high frequency and includes its own timestamp.

[0038] Because the sampling frequencies and time bases of the friction detection unit and the spatiotemporal positioning unit may differ, the acquired raw data needs to be time-aligned. For example, an interpolation algorithm is used to map all data points onto a unified 100Hz time axis, resulting in a multi-source dataset with a unified time axis. Then, for ease of analysis, this multi-source dataset is windowed according to a preset distance window (e.g., one window every 20 meters) to generate multiple road segment-level analysis units. Each analysis unit corresponds to a 20-meter-long area on the road segment under test and contains all synchronously acquired tire-road force, tire condition, and motion state data within that area.

[0039] For each road segment-level analysis unit, the system further extracts features. For example, for a 20-meter-long road segment-level analysis unit, the system calculates the average of all motion velocities within the unit as the window average velocity, the average of all lateral accelerations as the window average lateral acceleration, and the standard deviation of attitude information as the attitude change. These together constitute the kinematic characteristics of the unit. Simultaneously, based on all traction direction horizontal forces and normal loads within the unit, the system calculates the average friction coefficient and friction coefficient fluctuation index, serving as the friction statistical characteristics of the unit. Furthermore, the previously acquired road surface environmental features, preset water film thickness (0.8 mm), and preset friction detection unit operating parameters (such as a tire deflection angle of 5 degrees and a counterweight of 200 kg) are integrated with these kinematic characteristics, friction statistical characteristics, and tire condition characteristics (such as an average tire pressure of 240 kPa and an average tire temperature of 35°C) to form the corresponding window-level fused feature vector for that unit.

[0040] Finally, the window-level fusion feature vector of each road segment-level analysis unit is input into a pre-built fusion model. The fusion model (e.g., a deep learning model) generates an estimated friction coefficient (e.g., 0.35) and a risk assessment result (e.g., a risk level of "moderate" with a confidence level of 92%) for that road segment-level analysis unit based on these multi-dimensional features. Based on this estimated friction coefficient, the system further generates recommended safe operating parameters. For example, based on the estimated friction coefficient of 0.35, a recommended braking deceleration of 3.43 m / s² is calculated, and combined with a preset driving reaction time (e.g., 1.5 seconds) and a preset stopping distance (e.g., 50 meters), a recommended operating speed of 50 km / h is calculated. These results, including the estimated friction coefficient, risk level, and recommended operating speed, are all linked to the corresponding road segment spatial location information and displayed in real-time on the monitoring screen of the traffic management center, or pushed to the driver via an in-vehicle terminal.

[0041] As can be seen from the above examples, the intelligent monitoring method for road surface slippage in this application integrates road surface environmental characteristics, continuously measured tire-road forces, tire condition, and motion state information, and performs rigorous time alignment and windowing processing to construct a multi-dimensional fused feature vector. This enables the system to comprehensively and accurately reflect the actual slippage state under the coupled effects of multiple factors, overcoming the limitations of traditional methods that rely on empirical regression, indirect inference, or offline analysis. Compared to the difficulty of spatiotemporal alignment of multi-source heterogeneous data in existing technologies, this method ensures data consistency and the reliability of fusion results through meticulous time alignment processing. Furthermore, this method not only outputs estimated friction coefficients and risk assessment results, but also generates safe operation recommendation parameters that can be directly used for road operation management and safety early warning, such as recommended braking deceleration and recommended operating speed. This directly transforms the monitoring and analysis results into practical guidance, significantly improving the engineering application value of the method and solving the problem that the output results of existing technologies are difficult to directly translate into practical applications.

[0042] In some preferred embodiments, the friction detection unit includes a testing vehicle and a friction sensing module, a mating module, a shock absorption module, a height adjustment module, a mating-friction coordination module, a tire travel module, and a data integration module mounted on the testing vehicle. The testing vehicle is connected to a tractor. The friction sensing module provides conditions for acquiring data such as the internal pressure of the test wheel, the internal temperature of the test wheel, the vertical vibration of the test wheel, the total traction force of the testing vehicle, and the partial traction force of the testing vehicle. Specifically, the friction sensing module includes detection elements such as tire pressure / temperature sensors, acceleration sensors, and spoke-type force sensors. The module also includes isolation and buffer components to protect the sensors from dust, shock, and external damage, thereby ensuring that the detection data is stably and continuously transmitted to the data integration module. The counterweight module provides an adjustable vertical load for the test wheels of the friction detection unit to meet the friction detection requirements under different working conditions. Preferably, the counterweight module is implemented by loading counterweight blocks, with a counterweight range of 0–100 kg and a single counterweight block mass of 10 kg. The counterweight module also includes a counterweight fixing structure to ensure that the counterweight blocks remain stable and do not shift significantly during the movement of the detection vehicle. The vibration damping module reduces the horizontal and vertical vibration amplitudes generated during the movement of the detection vehicle, providing a stable working environment for all modules on the vehicle. Preferably, the vibration damping module includes a vertical damping spring and a hydraulic horizontal damper, which respectively mitigate vertical impact and horizontal vibration during the operation of the detection vehicle. Specifically, the vertical damping spring and the hydraulic horizontal damper are stably connected to the counterweight module, the counterweight-friction coordination module, and the friction sensing module, respectively, thereby achieving buffering and attenuation of the overall vibration response of the detection vehicle. The height adjustment module is used to transfer the traction force of the towing vehicle to the testing vehicle and provide vertical stability support when the testing vehicle is stationary. The height adjustment module includes a balance wheel, a height adjustment component, and a traction connection component. The height adjustment component can be adjusted via a guide wheel jack or equivalent structure to ensure height matching between the testing vehicle and the towing vehicle, and to maintain a constant and effective height adjustment during the testing vehicle's movement. The traction connection component preferably includes a traction connector and a length adjustment structure to achieve a reliable connection between the testing vehicle and the towing vehicle, and to meet the requirements of flexible installation, adjustable length, and easy disassembly. The counterweight-friction coordination module is used to stably transfer the vertical load applied by the counterweight module to the friction sensing module, while preventing the traction force on the counterweight module from being transferred to the friction sensing module, thereby reducing interference with the friction measurement results. Preferably, the counterweight-friction coordination module includes a counterweight support rail or an equivalent support force transmission structure. The counterweight support rail is used to complete the vertical load transfer from the counterweight module to the friction sensing module, and reduces the possibility of traction force transmission by reducing frictional wear between the support rail components.The tire travel module enables the test wheels of the testing vehicle to have an adjustable deflection angle and to switch between a locked state and a pure rolling state. Preferably, the tire travel module connects the test wheels to the axle pins, giving the test wheels the ability to adjust their deflection. The deflection angle of the test wheels is preferably adjustable within the range of 0 to 10°. The module preferably uses high-strength pin connections to meet the strength and wear resistance requirements of the connection between the tire and the axle. In addition, the tire travel module may also include a tire braking device to achieve the test wheel locking function.

[0043] In some preferred embodiments, step S2 includes: S21. Use surface water film spraying equipment to wet the road section to be tested in order to reduce the absorption of water film by the road surface. S22. Determine the equipment forward speed based on the preset equivalent water film thickness and preset spray flow rate corresponding to the preset wet and slippery working conditions. S23. Install a water-receiving component at the end of the road surface water film spraying equipment, control the road surface water film spraying equipment to move forward at the equipment forward speed in road sections other than the road section to be tested, and control the road surface water film spraying equipment to spray water film in road sections other than the road section to be tested at a preset spraying flow rate. S24. Obtain the water quality of the water-receiving component and calculate the actual water film thickness based on the water quality. S25. Analyze whether the absolute value of the deviation between the actual water film thickness and the preset equivalent water film thickness is less than the preset threshold. If yes, proceed to step S26. If no, generate an alarm message. S26. Control the road surface water film spraying equipment to move forward on the road section to be tested according to the equipment's forward speed, and control the road surface water film spraying equipment to spray water film on the road section to be tested at a preset spraying flow rate, so that the road section to be tested is in a preset wet and slippery condition. S27. The friction detection unit is moved by a mobile carrier, and the friction detection unit is used to collect the tire road force characteristics and tire state characteristics in real time. The spatiotemporal positioning unit is used to collect motion state information synchronously.

[0044] A road surface water film spraying device is used to apply a water film to a road surface. Its function is to transition the road section under test to a preset wet-slip condition. The purpose of the wetting treatment is to pre-wet the road surface to reduce its absorption of water during the actual water film spraying, thereby ensuring the stability of the water film thickness. This wetting treatment can be performed on the road section under test using a low-flow spraying system carried by a water truck. The preset wet-slip condition refers to a road surface condition simulating a specific degree of slipperiness. Its core parameter is the preset equivalent water film thickness. This preset equivalent water film thickness refers to the thickness of the uniform water layer formed by rainwater on the road surface under ideal conditions, assuming the road surface is absolutely smooth and even. The thickness of this uniform water layer remains constant along the assumed ideal smooth surface, without considering the influence of actual road surface unevenness or surface texture on local water accumulation. The preset spray flow rate is the pre-set amount of water sprayed by the spraying equipment per unit time, and the equipment forward speed is the speed at which the spraying equipment moves on the road surface. Determining the equipment's forward speed ensures that, under a given spray flow rate, a water film with a preset equivalent water film thickness can be formed on the tested road section. This speed can be calculated using the formula d = q / V × w, where d represents the preset equivalent water film thickness, q represents the preset spray flow rate, V represents the equipment's forward speed, and w represents the preset spray width. Both the preset spray flow rate and preset spray width are related to the setting of the road surface water film spraying equipment. Specifically, the mapping relationship between the preset spray flow rate and preset spray width and the setting of the road surface water film spraying equipment is as follows: when the setting of the road surface water film spraying equipment is the fifth setting, the preset spray width is 6.8 meters, and the preset spray flow rate is 8.75 m³ / h. 3 / h; When the road surface water film spraying equipment is in the fourth gear, the preset spraying width is 6.2 meters and the preset spraying flow rate is 8m³ / h. 3 / h; When the road surface water film spraying equipment is in the third gear, the preset spraying width is 5.7 meters and the preset spraying flow rate is 7.25 m³ / h. 3 / h; When the road surface water film spraying equipment is in the second gear, the preset spraying width is 5.3 meters and the preset spraying flow rate is 6.5m³ / h. 3 / h; When the road surface water film spraying equipment is in the first gear, the preset spraying width is 4.9 meters and the preset spraying flow rate is 5.75 m³ / h. 3 / h. A water-collecting assembly is a device used to collect water sprayed onto the road surface. The preferred water-collecting assembly is a stainless steel water-collecting tray with dimensions of 100mm × 100mm × 70mm. This tray is horizontally fixed in the center of the sprinkler truck's driving path and meets the following conditions: 1. The centerline of the water-collecting tray is aligned with the centerline of the nozzle of the sprinkler truck (road surface water film spraying equipment); 2. The water-collecting tray maintains a distance of 1.5–2.0m from the rear wheels of the sprinkler truck to reduce the impact of edge splashing; 3. The bottom of the water-collecting tray is basically flush with the ground to avoid the influence of slope and height differences on the amount of water collected. For the pre-verification of spraying parameters, it should be understood that the water-collecting tray is only used for local sampling and quantitative verification during the water film thickness verification process and is not placed in the test area during the friction test. Therefore, the local water collection by the water-collecting tray during the verification stage will not affect the overall water film distribution of the test section during the formal test. Pre-spraying on sections of road other than the test section is to verify whether the spraying parameters (equipment travel speed and preset spraying flow rate) can achieve the expected water film thickness before the formal operation, avoiding unnecessary interference to the actual test section. The collected water mass refers to the amount of water collected by the water-collecting component during pre-spraying. By measuring the collected water mass, the actual water film thickness formed under the current spraying parameters can be indirectly calculated. This mass can be obtained by weighing the water in the water-collecting component using an electronic scale. Analyzing whether the absolute value of the deviation between the actual water film thickness and the preset equivalent water film thickness is less than a preset threshold is the core of parameter verification. By comparing the difference between the actual water film thickness and the preset equivalent water film thickness, it can be determined whether the current spraying parameters meet the requirements. The preset threshold represents the allowable deviation range. Excessive deviation indicates inaccurate spraying parameters, while deviation within the allowable range indicates accurate parameters. The road surface water film spraying equipment will spray water film onto the actual test section according to the verified equipment travel speed and preset spray flow rate, thereby precisely constructing a road surface that meets the preset wet and slippery conditions. This equipment can be equipped with a GPS or RTK positioning system to ensure it travels precisely along the preset path and speed on the test section. Simultaneously, the spray flow rate can be monitored and controlled in real time via a flow meter to ensure stable spray volume. The mobile carrier can be a modified tractor unit, the friction detection unit can be a friction detection trolley connected to the tractor unit, and the spatiotemporal positioning unit can be a Beidou inertial navigation system.

[0045] This solution revolves around the accurate construction of preset wet and slippery conditions. Pre-verification ensures that the actual wet and slippery conditions of the road section under test meet the preset requirements, providing a reliable foundation for accurate wet and slippery monitoring. First, the road section under test is pre-wetted to saturate the surface beforehand, reducing additional water absorption during the actual water film spraying process. This prevents the water film thickness from continuously decreasing due to road surface water absorption after spraying, thus minimizing deviations in water film thickness caused by road surface water absorption. The equipment's forward speed is determined based on the preset equivalent water film thickness and preset spray flow rate required by the preset wet and slippery conditions. The operating speed is not arbitrarily set but matched to the required spray volume per unit area, ensuring that the spray volume per unit area meets the preset water film thickness requirements and that the parameter logic conforms to the laws governing water film thickness formation. A water-receiving component is installed at the end of the spraying equipment. A spraying test is first conducted outside the test section according to determined parameters, without altering the original state of the test section. Simultaneously, the water-receiving component collects the water falling onto it during the spraying, providing a basis for calculating the actual water film thickness. This allows for verification of spraying parameters without operating on the test section. The actual water film thickness is calculated based on the collected water mass. Utilizing the conversion relationship between mass and water film thickness, the actual water film thickness corresponding to the current parameters can be obtained without complex on-site water film thickness measurement equipment, making the process simple and reliable. The deviation between the actual water film thickness and the preset equivalent water film thickness is compared. Only if the deviation meets the requirements will subsequent formal operations proceed; otherwise, an alarm is triggered, prematurely terminating the non-compliant operation. This avoids wasted work caused by discovering thickness discrepancies after the test section has been sprayed, ensuring that the operating conditions for subsequent testing stages meet the requirements. After confirming that the deviation meets the requirements, the water film is officially sprayed onto the test section according to the verified parameters, ensuring that the final water film thickness meets the requirements of the preset wet-slip condition, thus completing the accurate construction of the wet-slip condition. Finally, a mobile carrier drives the friction detection unit and spatiotemporal positioning unit to simultaneously collect different types of detection data, achieving simultaneous acquisition of multi-source data and providing accurate and complete basic data for subsequent fusion analysis. This ability to accurately construct slippery working conditions, combined with the aforementioned intelligent monitoring method for slippery road surfaces, greatly improves the reliability of subsequent multi-source data fusion analysis and the accuracy of friction coefficient estimation, making risk assessment results and safe operation recommendations more instructive.

[0046] In some preferred embodiments, the formula for calculating the actual water film thickness is as follows: ; Where h represents the actual water film thickness, m represents the mass of water received, ρ represents the density of water, and A represents the preset bottom area corresponding to the water-receiving component.

[0047] In some preferred embodiments, the motion state information is collected by a spatiotemporal positioning unit, which includes a BeiDou inertial navigation component. Each first timestamp corresponds to a piece of motion state information, and each second timestamp corresponds to a tire track force feature and a tire state feature. Step S3 includes: S31. Perform data cleaning and filtering on all tire road force characteristics and all tire condition characteristics; S32. Remove motion state information collected when the BeiDou inertial navigation component is in a disconnection segment, saturation segment, unit abnormal segment, positioning status degradation, number of data acquisition satellites less than the preset number, integrated navigation loss of lock, speed jump or attitude change. S33. Establish the correspondence between the first timestamp and the second timestamp to coarsely align the tire road force characteristics, tire state characteristics, and motion state information. S34. Perform interpolation and / or resampling alignment on the coarsely aligned tire road force features, tire state features and motion state information to obtain a multi-source dataset under a unified time axis. S35. Calculate the real-time mileage of the friction detection unit relative to the starting point of the road segment under test based on the motion state information; S36. Divide the real-time mileage into multiple windows according to the preset distance window to perform windowing processing on the multi-source dataset; each window corresponds to a real-time mileage interval and a road segment-level analysis unit, and each road segment-level analysis unit includes all tire-road force characteristics, all tire state characteristics and all motion state information within the corresponding real-time mileage interval.

[0048] Motion state information refers to data describing the dynamic characteristics of the friction detection unit in space, such as position, velocity, and attitude. The spatiotemporal positioning unit includes a BeiDou inertial navigation component, which is a device integrating BeiDou satellite positioning and inertial navigation functions. This component can be an independent hardware module that outputs data through a serial interface or Ethernet interface. Each first timestamp corresponds to a piece of motion state information. The first timestamp is a time stamp synchronously generated by the spatiotemporal positioning unit when collecting motion state information, used to accurately record the data generation time of each piece of motion state information, serving as the basis for subsequent time alignment. This timestamp can be a high-precision clock synchronization signal within the system, or a unified time reference provided by an external time synchronization device (such as a GPS time synchronization module). Each second timestamp corresponds to a tire road force feature and a tire state feature. The second timestamp is a time stamp synchronously generated by the friction detection unit when collecting tire road force and tire state features, used to accurately record the data generation time of each tire road force and tire state feature, serving as the basis for subsequent time alignment. This timestamp can be the clock of the data acquisition system within the friction detection unit, or a unified clock synchronized with the mobile carrier's main control system.

[0049] Step S31 aims to improve the data quality of tire road force characteristics and tire condition characteristics. Specifically, the specific process of step S31 may include: 1. Performing unit and range checks on the tire road force characteristics and tire condition characteristics to identify abnormal and saturated segments; 2. Performing zero-point drift correction on the tire road force characteristics and tire condition characteristics. In this process, the average value of the pre-test static stage, the no-load baseline stage, or a preset reference segment can be used as the zero-point reference to perform baseline subtraction on subsequent sampling sequences; 3. Performing noise reduction filtering on the tire road force characteristics and tire condition characteristics. Preferably, FIR can be invoked. 4. Suppress high-frequency noise using filters or other digital filtering methods; 5. Identify and suppress abnormal spikes in the sampling sequence. Preferably, abnormal spikes can be identified based on the amplitude of abrupt changes in adjacent sampling points, the statistical characteristics of a sliding window, or exceeding a threshold, and processed using methods such as removal, neighborhood replacement, or interpolation correction; 6. Mark short-term missing segments and complete them using linear interpolation, nearest neighbor filling, or other conventional interpolation methods when continuity requirements are met. It should be understood that for missing segments whose duration exceeds a preset threshold, the missing mark is retained without forced completion. Step S32 aims to remove unreliable motion state information collected by the BeiDou inertial navigation component under abnormal working conditions. This can be done by judging and filtering based on the status flag bit output by the BeiDou inertial navigation component, or by analyzing continuous data to detect abrupt changes or discontinuities in parameters such as speed and attitude, thereby identifying abnormal data segments. Step S33 aims to initially resolve the inconsistency of time bases between different data sources, laying the foundation for subsequent accurate alignment. The nearest neighbor matching method can be used to match the first timestamp with the closest second timestamp. Alternatively, linear interpolation or spline interpolation can be used to map the data of one time series onto another time series. Step S34 aims to eliminate the differences in sampling frequency and time offset between different data sources and achieve accurate data synchronization. Methods such as linear interpolation and cubic spline interpolation can be used to interpolate the data so that it has corresponding values ​​at a unified time point. Alternatively, resampling technology can be used to resample all data to a common, preset sampling frequency. The specific process of step S34 may include: 1. Selecting a unified time axis as the target time reference. Preferably, the system time axis of the friction detection unit can be selected as the unified target time axis, or a unified target time axis can be constructed according to a preset sampling interval; 2. Interpolating and mapping the remaining BeiDou inertial navigation component data according to the unified target time axis so that the motion state information corresponds one-to-one with the tire road force characteristics and tire state characteristics, thereby obtaining a multi-source dataset under the unified time axis. It should be understood that for high sampling rate channels, this embodiment can use downsampling or resampling methods to map to the unified time axis.Step S35 aims to map the time-series data to a spatial dimension, providing a basis for subsequent distance-based windowing. This can be achieved by integrating the velocity data from the motion state information or by continuously accumulating the distances from the position coordinates output by the spatiotemporal positioning unit. Step S36 aims to divide the continuous, multi-source data along a unified time axis into discrete data units with clear spatial correspondences, facilitating subsequent road segment-level analysis. Fixed-length distance windows can be set (e.g., dividing a window every 5 or 10 meters).

[0050] This application's solution clarifies that motion state information is collected by a spatiotemporal positioning unit containing a BeiDou inertial navigation component, and distinguishes the timestamps corresponding to the two types of data, thus establishing a benchmark for subsequent time alignment operations and avoiding confusion regarding the time sources of different data during alignment. By cleaning and filtering all tire force characteristics and tire state characteristics, invalid and erroneous data in the original collected data are removed in advance, preventing invalid data from entering the subsequent alignment process and reducing interference with the alignment results. By removing motion state information collected under various abnormal operating conditions of the BeiDou inertial navigation component, motion state data with insufficient acquisition reliability is eliminated at the data source level, ensuring that the motion state information entering subsequent processing has sufficient reliability. A coarse alignment is achieved by establishing a correspondence between the first and second timestamps, initially mapping the two types of data with different time bases to a unified framework, completing preliminary alignment, reducing the difficulty of subsequent fine alignment, and improving alignment efficiency. By interpolating and / or resampling the coarsely aligned data, a multi-source dataset with a unified time axis is obtained. This further eliminates time discrepancies between data with different sampling frequencies, ensuring accurate correspondence between multi-source data at the same time point and resolving the issue of asynchronous multi-source data. By calculating the real-time mileage of the friction detection unit relative to the starting point of the road segment under test based on motion state information, the time-dimensional collected data is mapped to the actual spatial location, providing a positional basis for subsequent analysis unit division according to actual road segments, allowing windowing to match the spatial location of the actual road. Windowing is completed by dividing the real-time mileage according to preset distance windows, ensuring that each analysis unit corresponds to a clear actual road segment interval. This guarantees that the data of each analysis unit accurately corresponds to the specific road segment under test, facilitating subsequent wet-slip monitoring and analysis for specific road segments and improving the spatial specificity and accuracy of the monitoring results. This scheme addresses the issues of abnormal data interference and inconsistent time benchmarks during the alignment of multi-source collected data. Through a process of preprocessing to remove invalid data, step-by-step alignment, and finally windowing according to spatial location, accurate and reliable analysis units are obtained, providing accurate basic data for subsequent wet-slip state analysis.

[0051] In some preferred embodiments, the motion state information includes motion velocity, lateral acceleration, and attitude information; the tire track force characteristics include traction direction horizontal force and normal load; and step S4 includes: S41. For each segment-level analysis unit, calculate the window average velocity based on all motion velocities corresponding to the unit, calculate the window average lateral acceleration based on all lateral accelerations corresponding to the unit, and calculate the attitude change based on all attitude information corresponding to the unit. Then, integrate the window average velocity, window average lateral acceleration, and attitude change into the kinematic characteristics corresponding to the unit. S42. For each road segment-level analysis unit, calculate the average friction coefficient and friction coefficient fluctuation index corresponding to the unit based on the horizontal force in all traction directions and all normal loads corresponding to the unit, and then integrate the average friction coefficient and friction coefficient fluctuation index into the friction statistical characteristics corresponding to the unit. S43. For each road segment-level analysis unit, integrate the road surface environmental features, the preset water film thickness corresponding to the preset wet and slippery working conditions, the preset friction detection unit working condition parameters, and the kinematic features, friction statistical features, and tire state features corresponding to the unit into a window-level fusion feature vector corresponding to the unit.

[0052] Motion velocity can refer to the instantaneous linear velocity of the friction detection unit, lateral acceleration can refer to the instantaneous acceleration of the friction detection unit in the lateral direction, and attitude information can refer to the pitch angle, roll angle, and yaw angle of the friction detection unit. Besides the BeiDou inertial navigation components, motion state information can also be obtained through the Global Positioning System (GPS) or Global Navigation Satellite System (GNSS) combined with an inertial measurement unit (IMU). Horizontal force in the traction direction refers to the horizontal force exerted by the tire on the road surface in the traction or braking direction, and normal load refers to the force acting on the tire perpendicular to the road surface. A road segment-level analysis unit refers to a data set representing a specific road segment, obtained by windowing the data according to a preset time window or preset distance window after time alignment of multi-source datasets. Each road segment-level analysis unit contains all tire-road force characteristics, tire state characteristics, and motion state information within the corresponding time or distance interval. Windowing can use a fixed time length window, such as dividing the data into units every 1 second; or a fixed distance length window, such as dividing the data into units every 10 meters. Kinematic characteristics refer to statistical quantities extracted from the motion state information of a road segment-level analysis unit that characterize the overall motion characteristics of that road segment. These can include window-averaged speed, window-averaged lateral acceleration, and attitude change. Window-averaged speed is calculated by averaging all motion speeds within the unit, reflecting the overall driving speed level of the road segment. Window-averaged lateral acceleration is calculated by averaging all lateral accelerations within the unit, reflecting the overall lateral force state of the road segment. Attitude change can be calculated by the difference between the maximum and minimum values, standard deviation, or variance of all attitude information (such as pitch angle, roll angle, and heading angle) within the unit, reflecting the attitude fluctuation of the friction detection unit in that road segment. Friction statistical features refer to the statistical quantities extracted from the tire-road force characteristics of a road segment analysis unit, which characterize the overall frictional properties of that segment. These can include the average friction coefficient and the friction coefficient fluctuation index. The average friction coefficient is calculated by averaging all instantaneous friction coefficients within the unit (calculated as the ratio of the horizontal force in the traction direction to the normal load), reflecting the overall frictional level of the road segment. The friction coefficient fluctuation index can be calculated by calculating the standard deviation or variance of all instantaneous friction coefficients within the unit, reflecting the dispersion or uniformity of the friction coefficients in the road segment. In addition to the average and fluctuation index, friction statistical features can further include statistics such as the peak value, minimum value, skewness, and kurtosis of the friction coefficient. A window-level fusion feature vector is a vector formed by integrating features from various sources, including road surface environmental features, preset water film thickness corresponding to preset wet and slippery conditions, preset friction detection unit operating parameters, kinematic features, friction statistical features, and tire state features. This integration can unify scattered multi-source information into a standardized form suitable for the input of the fusion model. The integration method can be simply concatenating all features into a high-dimensional vector.

[0053] This application's solution addresses the problems of feature dimension chaos, inability to reflect the overall characteristics of road segments, and difficulty in reflecting multi-factor coupled wet and slippery conditions caused by directly inputting raw data into the fusion model in traditional methods. Specifically, after acquiring the road surface environment characteristics of the road segment under test, and when the segment is under a preset wet and slippery condition, a moving carrier drives the friction detection unit to move, collecting the tire-road force characteristics, tire state characteristics, and motion state information of the friction detection unit in real time. First, these multi-source data are time-aligned to obtain a multi-source dataset under a unified time axis. Then, the multi-source dataset is windowed according to a preset time window or preset distance window to obtain multiple road segment-level analysis units. Based on this, for each road segment-level analysis unit, this application's solution further clarifies the specific process for extracting effective features from the raw discrete sampling data. By calculating the window-averaged velocity, window-averaged lateral acceleration, and attitude change, the discrete motion state information is transformed into kinematic features that reflect the overall motion characteristics of the road segment. Simultaneously, by calculating the average friction coefficient and friction coefficient fluctuation index, the discrete tire-road force characteristics are transformed into friction statistical characteristics that reflect the overall friction level and uniformity of the road segment. These statistically aggregated features effectively avoid errors caused by random fluctuations at individual sampling points, more accurately reflecting the overall characteristics of the road segment. Finally, these extracted kinematic and friction statistical features are integrated with road surface environmental characteristics, preset water film thickness corresponding to preset wet and slippery conditions, preset friction detection unit operating parameters, and tire state characteristics to form a window-level fusion feature vector. This integration method enables the fusion model to fully utilize multiple types of information, such as road surface attributes, wet and slippery condition parameters, detection equipment parameters, and measured motion, friction, and tire state, thereby more comprehensively and accurately understanding and assessing the wet and slippery state of the road surface. The entire process forms an effective conversion mechanism from raw data to high-level features, providing a reliable and high-quality input foundation for subsequently generating estimated friction coefficients and risk assessment results for each road segment-level analysis unit using the pre-built fusion model.

[0054] In some preferred embodiments, the formula for calculating the average friction coefficient is: ; Where, μ meas (k) represents the average friction coefficient corresponding to the kth road segment-level analysis unit, N k W represents the number of valid sampling points within the k-th road segment-level analysis unit. k F represents the k-th segment-level analysis unit. x(i) represents the horizontal traction force at the i-th valid sampling point within the k-th road segment-level analysis unit, F. z (i) represents the normal load of the i-th effective sampling point within the k-th road segment-level analysis unit; The formula for calculating the friction coefficient fluctuation index is: ; Where, σ μ (k) represents the friction coefficient fluctuation index corresponding to the kth road segment-level analysis unit.

[0055] In some preferred embodiments, step S2 includes: S21' When the road section to be tested is under preset wet and slippery conditions, the friction detection unit is moved by a mobile carrier to collect tire road force characteristics, tire state characteristics, spatiotemporal positioning information and motion state information in real time. Step S3 includes: S31': Time alignment processing is performed on tire road force characteristics, tire state characteristics, spatiotemporal positioning information, and motion state information to obtain a multi-source dataset under a unified time axis. Then, the multi-source dataset is windowed according to a preset time window or a preset distance window to obtain multiple road segment-level analysis units. The multi-source dataset includes multiple sets of tire road force characteristics, tire state characteristics, spatiotemporal positioning information, and motion state information. Step S5 includes: S51. Using a pre-built fusion model, generate the estimated friction coefficient and risk assessment results for each road segment-level analysis unit based on the window-level fusion feature vector; the risk assessment results include risk level and confidence level. S52. Remove road segment-level analysis units with confidence levels lower than the preset confidence level, estimate friction coefficients and risk assessment results; S53. Generate safe operation recommendation parameters for each road segment-level analysis unit based on the estimated friction coefficient, and bind the safe operation recommendation parameters with the spatiotemporal positioning information corresponding to the parameters.

[0056] Spatiotemporal positioning information refers to data that accurately describes the geographical location and timestamp of the friction detection unit at a specific moment. This data can include longitude, latitude, altitude, and the corresponding precise time, and can be collected by the spatiotemporal positioning unit. The risk assessment results include risk level and confidence level. The risk level is a qualitative or semi-quantitative assessment of the road surface slipperiness or potential hazards, such as being categorized as low risk, medium risk, and high risk. In this embodiment, the preset level corresponding to the preset mapping interval to which the estimated friction coefficient belongs can be used as the risk level. Specifically, Among them, R k Indicates the risk level. The estimated friction coefficient is represented by μ1, μ2, and μ3, which represent the first, second, and third preset risk thresholds, respectively. Preferably, μ1 = 0.42, μ2 = 0.32, and μ3 = 0.25. Confidence level is a quantitative indicator of the reliability or accuracy of the risk assessment result, usually expressed as a probability value or a score. For example, the fusion model can output a value between 0 and 1, representing the model's confidence level in the current assessment result. The introduction of confidence level makes the assessment result no longer a single judgment, but rather includes its reliability basis, providing more comprehensive information for subsequent decisions. Removing road segment-level analysis units, estimated friction coefficients, and risk assessment results with confidence levels lower than the preset confidence level refers to quality control and screening of the risk assessment results output by the fusion model. By setting a preset confidence threshold, all road segment-level analysis units with confidence levels below this threshold, along with their corresponding estimated friction coefficients and risk assessment results, will be considered unreliable data and discarded. This effectively avoids interference from low-quality or erroneous assessment results caused by factors such as sensor noise, model uncertainty, or abnormal operating conditions on the final safety recommendations. Binding safety operation recommendation parameters with their corresponding spatiotemporal positioning information means that after generating safety operation recommendation parameters (such as recommended braking deceleration and recommended operating speed) for a specific road segment, these parameters are associated with the precise geographical location information of that road segment (such as starting latitude and longitude, ending latitude and longitude, or center point latitude and longitude) for storage or transmission. This binding gives the safety recommendations clear spatial directionality, directly indicating which specific road segment is at risk and what safety measures should be taken, greatly improving the practicality and operability of the information.

[0057] This application's solution effectively addresses the problems of inaccurate spatial positioning, difficulty in measuring result reliability, and difficulty in implementing safety recommendations in wet and slippery road surface monitoring by introducing spatiotemporal positioning information acquisition, result confidence assessment and screening, and spatial output binding into the basic monitoring method. Specifically, in the data acquisition stage, step S21' collects not only the tire-road force characteristics, tire state characteristics, and motion state information of the friction detection unit in real time, but also simultaneously collects spatiotemporal positioning information. This improvement ensures that all subsequent monitoring data naturally carry accurate geographical locations, laying the foundation for subsequent spatial analysis. Subsequently, in the data preprocessing stage, step S31' performs strict time alignment processing on all multi-source data, including spatiotemporal positioning information, ensuring a high degree of synchronization of data from different sensors in the time dimension. Based on this, windowing processing is performed according to preset time windows or distance windows, dividing the continuous monitoring data into multiple road segment-level analysis units with clear spatial and temporal ranges. This processing method ensures that the data within each analysis unit has a unified time reference and spatial context, avoiding the problem of mismatch between location and monitoring results due to data misalignment. In the fusion model evaluation phase, step S51 generates not only the estimated friction coefficient and risk level for each road segment-level analysis unit, but more importantly, it also outputs the confidence level of the evaluation results. The confidence level, as a quantitative indicator of the reliability of the evaluation results, provides a basis for subsequent quality control. Next, step S52 introduces a confidence-based result filtering mechanism. All road segment-level analysis units with confidence levels below a preset threshold, along with their corresponding estimated friction coefficients and risk assessment results, are automatically removed. This step effectively filters out low-reliability results caused by poor data quality, environmental interference, or model uncertainty, thereby significantly improving the accuracy and reliability of the final output safety recommendations and avoiding misleading decisions due to inaccurate information. Finally, in the safety recommendation generation and output phase, step S53 generates safe operation recommendation parameters based on the filtered estimated friction coefficients. This solution further binds these recommendation parameters to the spatiotemporal location information corresponding to the road segment. This binding ensures that the final safety recommendations are no longer abstract numerical values, but rather guiding information with clear geographical coordinates that can be directly applied to specific road segments. For example, road management departments can accurately locate risky road sections on a map based on suggested operating speeds with latitude and longitude information, and issue targeted warnings or take control measures accordingly. Through the synergistic effect of the above series of steps, the solution in this application constructs a full-chain optimization mechanism from data collection to result output, enabling the intelligent monitoring method for slippery road surfaces to provide more accurate, reliable, and spatially oriented risk assessments and safe operation recommendations, greatly enhancing its application value in actual road traffic management and safety early warning.

[0058] The following is a concrete example. Suppose we are conducting slippery road monitoring on a section of highway to be tested. As a specific implementation method, in step S21', the mobile carrier can be a tractor-trailer connected to a friction detection unit, which can be a friction detection trolley. The tractor-trailer is also equipped with a high-precision spatiotemporal positioning unit, such as a combined navigation system integrating a BeiDou satellite receiver and an inertial measurement unit (IMU). When the tractor-trailer passes through the wetted section of road at a preset speed, the system collects real-time data at a high frequency (e.g., 100Hz) on the tire-road force characteristics and tire condition characteristics, including the horizontal force in the traction direction, normal load, tire pressure, and tire temperature. Simultaneously, the combined navigation system collects the vehicle's speed, lateral acceleration, attitude information, and precise latitude, longitude, altitude, and timestamps, as well as spatiotemporal positioning and motion state information. All of this data carries its own original timestamp. In step S31', the collected multi-source data is first time-aligned. For example, the data from the friction detection unit is sampled at 100Hz, while the data from the spatiotemporal positioning unit is sampled at 50Hz. The system can use linear interpolation to interpolate the spatiotemporal positioning data to 100Hz, aligning it with the friction detection data on the time axis. After time alignment, the system can perform windowing processing on the data according to a preset distance window (e.g., one window every 20 meters). This means that every 20 meters traveled forms a segment-level analysis unit, which contains all synchronous tire-road force characteristics, tire state characteristics, spatiotemporal positioning information, and motion state information for that 20-meter segment. Each segment-level analysis unit will be associated with its corresponding start and end geographic coordinates. In step S51, a pre-trained deep learning fusion model, such as a model based on a convolutional neural network (CNN) or a recurrent neural network (RNN), receives the window-level fusion feature vector of each segment-level analysis unit. This model not only outputs the estimated friction coefficient (e.g., 0.35) and risk level (e.g., "high risk") for that segment, but also outputs a confidence score (e.g., 0.68), which reflects the model's certainty regarding the estimation result. In step S52, the system checks the confidence level of each road segment-level analysis unit. If the preset confidence threshold is 0.70, the road segment-level analysis units with a confidence level of 0.68, along with their corresponding estimated friction coefficients and risk assessment results, will be removed and not used for subsequent safety recommendation generation because their reliability does not meet the preset standard. For road segments with a confidence level higher than 0.70, their results are retained. In step S53, for those road segment-level analysis units that pass the confidence level screening, the system generates safe operation recommendation parameters based on their estimated friction coefficients.For example, if the estimated friction coefficient of a certain road segment is 0.45, the system can calculate a recommended operating speed of 60 km / h and a recommended braking deceleration of 4.41 m / s² based on a preset driving reaction time (e.g., 1.5 seconds) and a preset stopping distance (e.g., 50 meters). Subsequently, these recommended parameters (recommended operating speed and recommended braking deceleration) are bound to the precise spatiotemporal location information of the road segment (e.g., the starting latitude and longitude of the 20-meter segment: 10.12345°N, 120.67890°E; ending latitude and longitude: 10.12360°N, 120.67910°E) to form a safety recommendation data package containing location information, facilitating accurate risk warnings and traffic management by road management departments.

[0059] In some preferred embodiments, the recommended parameters for safe operation include recommended braking deceleration and recommended operating speed, and the formula for calculating the recommended braking deceleration is as follows: ; Among them, a b This indicates a suggestion to brake and decelerate. This represents the estimated coefficient of friction, and g represents the acceleration due to gravity. The recommended formula for calculating running speed is: ; in, Indicates the suggested running speed, t r D represents the preset driving reaction time. max This indicates the preset parking distance.

[0060] Taking the single-factor working condition test conducted under actual road conditions as an example, the influence of four key factors—water film thickness, vehicle load, vehicle speed, and single-wheel deflection angle—on the friction coefficient was statistically analyzed. The results for summer and winter are as follows: Figure 2 and Figure 3 As shown. Therefore, it is evident that this application can achieve the following beneficial effects: (I) This application enables online continuous friction detection under actual road conditions and generates stable friction statistical output. Specifically, addressing the problem that existing technologies rely on complex 3D reconstruction, numerical simulation, or offline analysis, which are insufficient for real-time applications, this application uses a friction detection unit to collect traction force and normal load online under actual road conditions and calculate the friction coefficient. Simultaneously, it uses the mean (average friction coefficient) and standard deviation (friction coefficient fluctuation index) to describe the friction statistical characteristics, thus generating continuously updated friction capability characterization results. Experimental results show that continuous and stable friction statistical output can be obtained under both summer and winter conditions: the average friction coefficient is generally distributed in the range of 0.31–0.35 (summer) and 0.32–0.38 (winter), and the standard deviation ranges from approximately 0.0036 to 0.0355. These results demonstrate that this application can continuously output friction level and its dispersion information under different seasonal environments, providing a reliable data foundation for real-time updates of slippery risk.

[0061] (ii) This application can reflect the influence trend of typical dynamic working conditions on friction capability. Specifically, in view of the problem that indoor tests in the prior art are difficult to reflect the multi-factor coupling effect under real driving conditions, this application controls or records factors such as water film thickness, vehicle load, driving speed and single wheel deflection angle under actual road conditions, and performs statistical analysis on the relationship between them and the friction coefficient.

[0062] The experimental results show that different operating conditions all have identifiable effects on the friction coefficient, and exhibit consistent trends under summer and winter conditions: 1) Water film thickness factor: In both seasons, there is a distinguishable difference in the mean friction value of 0.5mm water film compared to 1.0 / 1.5mm (approximately 0.02 in summer; approximately 0.01–0.02 in winter), and the corresponding standard deviation range can be given to support risk calibration under thin water film conditions. 2) Vehicle load factors: In both seasons, the average friction value increased significantly when the load increased from 2.16kN to 3.16kN (approximately +0.05 in summer and approximately +0.06 in winter), indicating that this application can stably reproduce and quantify the differences in adhesion characteristics caused by load changes; 3) Driving speed factor: In both seasons, the average friction value decreased when the speed increased from 10km / h to 30km / h (approximately -0.02 in summer and -0.03 in winter), providing a direct basis for adopting "speed constraint / speed correction" in engineering implementation; 4) Single wheel deflection angle factor: In both seasons, the average friction value decreased slightly when the deflection angle increased from 4° to 8° (approximately -0.01 in both seasons). At the same time, the standard deviations for different deflection angles were different, which provides a basis for optimizing the standard testing conditions (such as selecting a more stable deflection angle level).

[0063] The above results demonstrate that this application can simultaneously perceive and quantify key operating factors affecting friction capability under actual road conditions, providing a basis for subsequent wet skid risk identification and parameter calibration.

[0064] (iii) This application can improve the intuitiveness and stability of friction characterization based on physical measurement results. Specifically, in view of the problem that friction parameters in the prior art mainly rely on empirical models for indirect estimation and lack direct physical measurement support, this application calculates the instantaneous friction coefficient by directly measuring the traction force and normal load, and uses the standard deviation to characterize the data dispersion, so that the friction results have both average level characterization and fluctuation degree characterization.

[0065] The experimental results show that the standard deviation obtained under different working conditions is approximately 0.0036 to 0.0355, indicating that this application can not only output the friction mean, but also reflect the stability of friction changes, thus providing a basis for abnormal data identification, risk threshold setting and result credibility evaluation.

[0066] Based on the above technical solution, this application can support continuous road segment identification and risk assessment of slippery road conditions, and generate road segment-level slippery risk levels, result confidence levels, and corresponding safe operation parameters, thereby providing technical basis for vehicle safety driving warnings and road operation management. This application can carry out continuous detection and analysis in actual road environments. Compared with methods that rely on complex offline simulations or single static evaluations, it is more conducive to improving the efficiency of slippery road segment identification, reducing the time consumption caused by complex analysis processes, and providing direct data support for road maintenance, operation management, and safety warnings.

[0067] Please refer to Figure 4 , Figure 4This application provides a schematic diagram of the structure of an electronic device according to an embodiment of the present application. The electronic device includes a processor 101 and a memory 102. The processor 101 and the memory 102 are interconnected and communicate with each other via a communication bus 103 and / or other forms of connection mechanisms (not shown). The memory 102 stores computer-readable instructions executable by the processor 101. When the electronic device is running, the processor 101 executes these computer-readable instructions to perform the method in any optional implementation of the above embodiments, thereby achieving the following functions: Step S1: Obtain the road surface environment characteristics of the road segment to be tested; the road surface environment characteristics include road surface texture characteristics and road surface temperature characteristics; Step S2: When the road segment to be tested is under a preset wet and slippery condition, a friction detection unit is moved by a moving carrier to collect tire road force characteristics, tire state characteristics, and motion state information in real time; Step S3: Perform time alignment processing on the tire road force characteristics, tire state characteristics, and motion state information to obtain a unified time... The multi-source dataset under the inter-axis is then windowed according to a preset time window or a preset distance window to obtain multiple road segment-level analysis units. The multi-source dataset includes multiple sets of tire-road force features, tire state features, and motion state information. Step S4: For each road segment-level analysis unit, the kinematic features corresponding to the unit are obtained based on all motion state information corresponding to the unit, and the friction statistical features corresponding to the unit are obtained based on all tire-road force features corresponding to the unit. Then, the road surface environment features, the preset water film thickness corresponding to the preset wet and slippery working conditions, the preset friction detection unit working condition parameters, and the kinematic features, friction statistical features, and tire state features corresponding to the unit are integrated into the window-level fusion feature vector corresponding to the unit. Step S5: Using the pre-built fusion model, the estimated friction coefficient and risk assessment results corresponding to each road segment-level analysis unit are generated based on the window-level fusion feature vector. Then, the safe operation recommendation parameters corresponding to each road segment-level analysis unit are generated based on the estimated friction coefficient.

[0068] This application embodiment also provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it executes the method in any optional implementation of the above embodiments to achieve the following functions: Step S1: Acquire the road surface environment features of the road segment to be tested; the road surface environment features include road surface texture features and road surface temperature features; Step S2: When the road segment to be tested is under a preset wet and slippery condition, a friction detection unit is moved by a moving carrier to collect tire-road force features, tire state features, and motion state information in real time; Step S3: Perform time alignment processing on the tire-road force features, tire state features, and motion state information to obtain a multi-source dataset under a unified time axis, and then perform windowing processing on the multi-source dataset according to a preset time window or a preset distance window to obtain multiple road segment-level analysis units; Multi-source dataset This includes multiple sets of tire-road force characteristics, tire state characteristics, and motion state information; Step S4: For each road segment-level analysis unit, obtain the kinematic characteristics corresponding to the unit based on all motion state information corresponding to the unit, and obtain the friction statistical characteristics corresponding to the unit based on all tire-road force characteristics corresponding to the unit. Then, integrate the road surface environment characteristics, the preset water film thickness corresponding to the preset wet and slippery working conditions, the preset friction detection unit working condition parameters, and the kinematic characteristics, friction statistical characteristics, and tire state characteristics corresponding to the unit into a window-level fusion feature vector corresponding to the unit; Step S5: Use the pre-built fusion model to generate the estimated friction coefficient and risk assessment results corresponding to each road segment-level analysis unit based on the window-level fusion feature vector, and then generate the safe operation recommendation parameters corresponding to each road segment-level analysis unit based on the estimated friction coefficient. The computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0069] As can be seen from the above, the intelligent monitoring method, electronic device and storage medium for road surface slipperiness provided in this application integrates road surface environmental information, operating parameters and multi-source measured data, extracts fusion features through time-aligned windowing processing, and finally generates directly usable monitoring results and safe operation suggestion parameters. It solves the problems of large deviation in monitoring results, insufficient spatiotemporal alignment accuracy of multi-source data and difficulty in directly applying the output results in the prior art. It can accurately reflect the actual slippery state of the road surface under the coupling effect of multiple factors, improve the accuracy and reliability of monitoring results, and the output results can directly serve road operation management and safety early warning, and have high engineering application value.

[0070] In the embodiments provided in this application, it should be understood that relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0071] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent monitoring of slippery road surfaces, characterized in that, The intelligent monitoring method for slippery road surfaces includes the following steps: S1. Obtain the road surface environmental features of the road segment to be tested; the road surface environmental features include road surface texture features and road surface temperature features; S2. When the road section to be tested is under a preset wet and slippery condition, the friction detection unit is moved by a mobile carrier to collect tire road force characteristics, tire condition characteristics and motion state information in real time. S3. Perform time alignment processing on the tire road force features, tire state features, and motion state information to obtain a multi-source dataset under a unified time axis. Then, perform windowing processing on the multi-source dataset according to a preset time window or a preset distance window to obtain multiple road segment-level analysis units. The multi-source dataset includes multiple sets of the tire road force features, tire state features, and motion state information. S4. For each road segment-level analysis unit, obtain the kinematic features corresponding to the unit based on all the motion state information corresponding to the unit, and obtain the friction statistical features corresponding to the unit based on all the tire-road force features corresponding to the unit. Then, integrate the road surface environment features, the preset water film thickness corresponding to the preset wet and slippery working condition, the preset friction detection unit working condition parameters, and the kinematic features, friction statistical features, and tire state features corresponding to the unit into a window-level fusion feature vector corresponding to the unit. S5. Using the pre-built fusion model, generate the estimated friction coefficient and risk assessment results corresponding to each road segment-level analysis unit based on the window-level fusion feature vector, and then generate the safe operation recommendation parameters corresponding to each road segment-level analysis unit based on the estimated friction coefficient.

2. The intelligent monitoring method for slippery road surfaces according to claim 1, characterized in that, Step S2 includes: S21. Use a surface water film spraying device to wet the road section to be tested in order to reduce the absorption of water film by the road surface. S22. Determine the equipment forward speed based on the preset equivalent water film thickness and preset spray flow rate corresponding to the preset wet and slippery working conditions. S23. Install a water-collecting component at the end of the road surface water film spraying device, control the road surface water film spraying device to move forward at the forward speed of the device on the road section other than the road section to be tested, and control the road surface water film spraying device to spray water film on the road section other than the road section to be tested at the preset spraying flow rate. S24. Obtain the water mass received by the water-receiving component, and calculate the actual water film thickness based on the water mass received; S25. Analyze whether the absolute value of the deviation between the actual water film thickness and the preset equivalent water film thickness is less than a preset threshold. If yes, proceed to step S26; otherwise, generate an alarm message. S26. Control the road surface water film spraying equipment to move forward on the road section to be tested at the speed of the equipment, and control the road surface water film spraying equipment to spray water film on the road section to be tested at the preset spraying flow rate, so that the road section to be tested is in the preset wet and slippery condition. S27. The friction detection unit is moved by a mobile carrier, and the friction detection unit is used to collect the tire road force characteristics and tire state characteristics in real time, and the motion state information is collected synchronously by the spatiotemporal positioning unit.

3. The intelligent monitoring method for slippery road surfaces according to claim 2, characterized in that, The formula for calculating the actual water film thickness is as follows: ； Where h represents the actual water film thickness, m represents the mass of water received, ρ represents the density of water, and A represents the preset bottom area corresponding to the water-receiving component.

4. The intelligent monitoring method for road surface slipperiness according to claim 1, characterized in that, The motion state information is collected by a spatiotemporal positioning unit, which includes a BeiDou inertial navigation component. Each first timestamp corresponds to one piece of motion state information, and each second timestamp corresponds to one tire road force feature and one tire state feature. Step S3 includes: S31. Perform data cleaning and data filtering on all the tire road force characteristics and all the tire state characteristics; S32. Remove motion state information collected when the BeiDou inertial navigation component is in a disconnection segment, saturation segment, unit abnormal segment, positioning status degradation, number of data acquisition satellites less than a preset number, integrated navigation loss of lock, speed jump or attitude change. S33. Establish the correspondence between the first timestamp and the second timestamp to coarsely align the tire road force characteristics, the tire state characteristics, and the motion state information. S34. Perform interpolation and / or resampling alignment on the coarsely aligned tire road force features, tire state features and motion state information to obtain a multi-source dataset under a unified time axis. S35. Calculate the real-time mileage of the friction detection unit relative to the starting point of the road segment to be tested based on the motion state information; S36. Divide the real-time mileage into multiple windows according to a preset distance window to perform windowing processing on the multi-source dataset; each window corresponds to a real-time mileage interval and a road segment-level analysis unit, and each road segment-level analysis unit includes all the tire-road force features, all the tire state features and all the motion state information within the corresponding real-time mileage interval.

5. The intelligent monitoring method for road surface slipperiness according to claim 1, characterized in that, The motion state information includes motion speed, lateral acceleration, and attitude information; the tire track force characteristics include traction direction horizontal force and normal load; step S4 includes: S41. For each of the road segment-level analysis units, calculate the window average velocity based on all the motion velocities corresponding to the unit, calculate the window average lateral acceleration based on all the lateral accelerations corresponding to the unit, and calculate the attitude change based on all the attitude information corresponding to the unit. Then, integrate the window average velocity, the window average lateral acceleration, and the attitude change into the kinematic features corresponding to the unit. S42. For each road segment-level analysis unit, calculate the average friction coefficient and friction coefficient fluctuation index corresponding to the unit based on all horizontal forces in the traction direction and all normal loads corresponding to the unit, and then integrate the average friction coefficient and friction coefficient fluctuation index into the friction statistical characteristics corresponding to the unit. S43. For each road segment-level analysis unit, the road surface environmental features, the preset water film thickness corresponding to the preset wet and slippery working conditions, the preset friction detection unit working condition parameters, and the kinematic features, friction statistical features, and tire state features corresponding to the unit are integrated into a window-level fusion feature vector corresponding to the unit.

6. The intelligent monitoring method for road surface slipperiness according to claim 5, characterized in that, The formula for calculating the average friction coefficient is: ； Where, μ meas (k) represents the average friction coefficient corresponding to the kth road segment-level analysis unit, N k W represents the number of valid sampling points within the k-th road segment-level analysis unit. k F represents the k-th segment-level analysis unit. x (i) represents the horizontal traction force at the i-th valid sampling point within the k-th road segment-level analysis unit, F. z (i) represents the normal load of the i-th effective sampling point within the k-th road segment-level analysis unit; The formula for calculating the friction coefficient fluctuation index is as follows: ； Where, σ μ (k) represents the friction coefficient fluctuation index corresponding to the kth road segment-level analysis unit.

7. The intelligent monitoring method for road surface slipperiness according to claim 1, characterized in that, Step S2 includes: S21' When the road section to be tested is under a preset wet and slippery condition, the friction detection unit is moved by a mobile carrier to collect tire road force characteristics, tire state characteristics, spatiotemporal positioning information and motion state information in real time. Step S3 includes: S31': Perform time alignment processing on the tire road force features, tire state features, spatiotemporal positioning information, and motion state information to obtain a multi-source dataset under a unified time axis. Then, perform windowing processing on the multi-source dataset according to a preset time window or a preset distance window to obtain multiple road segment-level analysis units; the multi-source dataset includes multiple sets of the tire road force features, tire state features, spatiotemporal positioning information, and motion state information; Step S5 includes: S51. Using a pre-constructed fusion model, generate the estimated friction coefficient and risk assessment results corresponding to each road segment-level analysis unit based on the window-level fusion feature vector; the risk assessment results include risk level and confidence level. S52. Remove road segment-level analysis units with confidence levels lower than the preset confidence level, estimate friction coefficients and risk assessment results; S53. Generate safe operation recommendation parameters for each of the road segment-level analysis units based on the estimated friction coefficient, and bind the safe operation recommendation parameters with the spatiotemporal positioning information corresponding to the parameters.

8. The intelligent monitoring method for road surface slipperiness according to claim 7, characterized in that, The recommended parameters for safe operation include recommended braking deceleration and recommended operating speed. The formula for calculating the recommended braking deceleration is as follows: ； Among them, a b This indicates a suggestion to brake and decelerate. This represents the estimated coefficient of friction, and g represents the acceleration due to gravity. The formula for calculating the recommended operating speed is as follows: ； in, Indicates the suggested running speed, t r D represents the preset driving reaction time. max This indicates the preset parking distance.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing computer-readable instructions that, when executed by the processor, perform the steps of the method as described in any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it performs the steps of the method as described in any one of claims 1-8.

Images