A multi-source perception and hierarchical early warning method for road risk in adverse weather

Abstract

The application relates to the technical field of road monitoring, and discloses a road risk multi-source perception and hierarchical early warning method under severe weather, which comprises the following steps: collecting meteorological data and road state data of each road area in real time and carrying out pretreatment; establishing a multi-source data fusion model based on the pretreated meteorological data and road state data, carrying out data interpolation and supplement by using the multi-source data fusion model, and extracting key risk characteristic factors; determining the comprehensive risk score of each road area according to the key risk characteristic factors, and determining the early warning level of each road area according to the comprehensive risk score; generating a protection strategy according to the early warning level and the key risk characteristic factors, and controlling the working state of the corresponding physical protection facilities based on the protection strategy. The application reduces the influence of severe weather on road traffic safety and improves the safety and reliability of road operation.

Description

Technical Field

[0001] This invention relates to the field of road monitoring technology, and in particular to a method for multi-source perception and graded early warning of road risks under severe weather conditions. Background Technology

[0002] With the intensification of global climate change, extreme weather events (such as heavy rain, blizzards, dense fog, and freezing temperatures) are becoming more frequent, posing a serious threat to road traffic safety. Traditional road risk warning systems often rely on single data sources (such as weather station data or traffic flow monitoring), which suffer from limited perception dimensions, incomplete data coverage, and delayed response, making it difficult to accurately reflect the real-time risk status of complex road networks under severe weather conditions. Furthermore, existing systems generally lack effective mechanisms for integrating multi-source heterogeneous data (including meteorological data, road physical conditions, historical disaster records, and vehicle dynamic behavior), resulting in crude risk assessments, highly subjective warning level classifications, and an inability to form a closed-loop linkage control with physical protection facilities, thus hindering the improvement of proactive traffic safety prevention and control capabilities.

[0003] In recent years, although some studies have attempted to introduce data fusion or machine learning methods for road risk prediction, they generally suffer from the following shortcomings: First, they have not fully considered robustness handling in scenarios with missing data, affecting model stability; second, the risk feature factor extraction dimensions are too singular, failing to integrate multi-dimensional information such as meteorology, road itself, historical disasters, and traffic behavior; third, there is a lack of refined and executable mapping logic between warning levels and emergency strategies, especially the lack of a "one-vote veto" rapid response mechanism for extremely high risks based on key threshold triggers; and fourth, the automatic linkage control between warning results and physical protective facilities (such as defogging devices, shields, electronic warning screens, etc.) has not yet been realized, making it difficult to form an integrated intelligent closed loop of "perception-assessment-early warning-response". Summary of the Invention

[0004] The purpose of this invention is to provide a method for multi-source perception and graded early warning of road risks under severe weather conditions, in order to solve the above-mentioned problems.

[0005] This invention provides a method for multi-source perception and graded early warning of road risks under severe weather conditions, the method comprising: Real-time collection of meteorological and road condition data for each road area, and preprocessing of the meteorological and road condition data; A multi-source data fusion model is established based on preprocessed meteorological data and road condition data. The multi-source data fusion model is used for data interpolation and supplementation, and key risk characteristic factors are extracted. A comprehensive risk score is determined for each road area based on the key risk characteristic factors, and a warning level is determined for each road area based on the comprehensive risk score. A protection strategy is generated based on the warning level and key risk characteristic factors, and the working status of the corresponding physical protection facilities is controlled based on the protection strategy.

[0006] Preferably, when preprocessing the meteorological data and road condition data, the preprocessing includes data cleaning, standardization, and outlier removal.

[0007] Preferably, a multi-source data fusion model is established based on preprocessed meteorological data and road condition data, including: Historical disaster data and vehicle traffic data are acquired, and the historical disaster data, vehicle traffic data, meteorological data, and road condition data are time-aligned and spatially standardized. Data features are extracted for each type of data, and the data features are fused through intermediate fusion and attention mechanisms to obtain a multi-source data fusion model.

[0008] Preferably, the multi-source data fusion model is used for data interpolation and supplementation, and key risk characteristic factors are extracted, including: The missing data type, missing spatial region, and missing time window are determined, and the missing rate is calculated. Based on the missing spatial region, missing time window, and missing rate, the priority of the imputation strategy is determined. Based on the missing data type and the priority of the imputation strategy, the imputation submodule of the multi-source data fusion model is called to interpolate and supplement the data. After data interpolation and supplementation, key risk characteristic factors are extracted from the fused data. These key risk characteristic factors include meteorological risk characteristic factors, road risk characteristic factors, historical disaster risk characteristic factors, and traffic risk characteristic factors. Meteorological risk characteristic factors include: temperature, precipitation / snowfall, wind speed, visibility, road icing probability, and humidity. Road risk characteristic factors include: road surface temperature, friction coefficient, water depth, snow thickness, and camera image recognition results. Historical disaster risk characteristic factors include: frequency of disasters occurring on a certain road section in the past N years, time since the most recent disaster, and disaster severity level. Traffic risk characteristic factors include: average vehicle speed, traffic volume, speed standard deviation, frequency of sudden braking events, and trajectory deviation rate.

[0009] Preferably, the comprehensive risk score for each road area is determined based on the key risk characteristic factors, including: A risk assessment index system was constructed using the analytic hierarchy process (AHP). Meteorological risk characteristic factors, road risk characteristic factors, historical disaster risk characteristic factors, and traffic risk characteristic factors were used as primary indicators. Each characteristic factor under a primary indicator was assigned a corresponding weight, which was determined by a combination of expert scoring and entropy weighting. Each key risk characteristic factor is standardized and converted into a unified scoring range. The indicator type for each key risk characteristic factor is determined, including positive and negative indicators. For positive indicators, a standardization method of "the larger the better" is adopted, and for negative indicators, a standardization method of "the smaller the better" is adopted. The comprehensive risk score for each road area is calculated by weighted summation.

[0010] Preferably, the warning level for each road area is determined based on the comprehensive risk score, including: A first score, a second score, and a third score are preset, with the first score, the second score, and the third score increasing sequentially. The warning level for the corresponding road area is set according to the relationship between the comprehensive risk score and the first score, the second score, and the third score. If the comprehensive risk score is less than the first score, the warning level for the corresponding road area is set to low risk. If the comprehensive risk score is greater than or equal to the first score and the comprehensive risk score is less than the second score, then the warning level for the corresponding road area is set to medium risk. If the comprehensive risk score is greater than or equal to the second score, and the comprehensive risk score is less than the third score, then the warning level for the corresponding road area is set to a high-risk level. If the comprehensive risk score is greater than or equal to the third score, then the warning level for the corresponding road area is set to an extremely high risk level.

[0011] Preferably, before determining the warning level for each road area based on the comprehensive risk score, the process includes: A threshold is set for each key risk characteristic factor. The corresponding value of the key risk characteristic factor is compared with the threshold. If the corresponding value of the key risk characteristic factor is greater than the threshold, the warning level of the corresponding road area is directly determined to be an extremely high risk level.

[0012] Preferably, a protection strategy is generated based on the warning level and key risk characteristic factors, including: If the warning level is low risk, no protection strategy will be generated, and monitoring will continue. If the warning level is medium risk or high risk, a protection strategy is generated. If the warning level is high-risk, the emergency response mechanism will be activated immediately, the road area will be closed, and vehicles will be clearly informed of the prohibition of passage through road closure warning signs and electronic displays.

[0013] Preferably, a protection strategy is generated based on the warning level and key risk characteristic factors, including: Preset limits for precipitation / snowfall, visibility, and humidity; The precipitation / snowfall is compared with the precipitation / snowfall limit, the visibility is compared with the visibility limit, and the humidity is compared with the humidity limit. If the precipitation / snowfall exceeds the precipitation / snowfall limit, a shading strategy is generated. If the visibility is less than the visibility limit, or the humidity is greater than the humidity limit, a warm air defogging strategy is generated.

[0014] Preferably, controlling the operating status of the corresponding physical protection facilities based on the protection strategy includes: The physical protection facilities include: a protective cover, a lighting unit, and a warm air defogging unit; If the protection strategy includes a shielding strategy, then the protective shield and lighting unit are activated; If the protection strategy includes a warm air defogging strategy, then the warm air defogging unit and the lighting unit are activated.

[0015] Compared with existing technologies, the advantages of this invention lie in its ability to establish a multi-source data fusion model by real-time acquisition and preprocessing of meteorological and road condition data, combined with historical disaster and traffic flow data. This effectively overcomes the limitations of a single data source and comprehensively depicts the complex risk status of the road system under severe weather conditions. Utilizing the interpolation and supplementation mechanisms in the fusion model, the optimal interpolation strategy is dynamically selected for different missing types, spatial regions, and time windows, ensuring the complete extraction of key risk characteristic factors (including meteorological, road, historical disaster, and traffic categories), providing high-quality input for subsequent assessments. The weights of each risk factor are determined by combining the analytic hierarchy process (AHP) and the entropy weight method, and the scoring scale is standardized. Finally, a weighted summation method is used to calculate the comprehensive risk score, making the warning level classification both scientific and interpretable. On the one hand, by preset thresholds for key risk factors, immediate judgment of extremely high risks (such as sudden drops in visibility, excessive snow accumulation, etc.) is achieved, triggering emergency road closures. On the other hand, based on the comprehensive score range, four levels of warnings—low, medium, high, and extremely high—are divided to achieve differentiated response strategies, balancing safety and traffic efficiency. Based on the warning level and specific risk factors (such as precipitation / snowfall, visibility, and humidity), the system automatically generates shielding strategies or warm air defogging strategies, and links and controls physical facilities such as protective covers, lighting units, and warm air defogging units to achieve automated and intelligent closed-loop control from risk perception to active protection, significantly improving the resilience and safety of the road system under severe weather conditions. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating a method for multi-source perception and graded early warning of road risks under severe weather conditions according to the present invention. Detailed Implementation

[0018] 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 some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0019] like Figure 1 As shown, this invention provides a method for multi-source perception and graded early warning of road risks under severe weather conditions, the method comprising: Meteorological data and road condition data for each road area are collected in real time, and the meteorological data and road condition data are preprocessed.

[0020] A multi-source data fusion model is established based on preprocessed meteorological and road condition data. The multi-source data fusion model is then used for data interpolation and supplementation, and key risk characteristic factors are extracted.

[0021] A comprehensive risk score is determined for each road area based on the key risk characteristic factors, and a warning level is determined for each road area based on the comprehensive risk score.

[0022] A protection strategy is generated based on the warning level and key risk characteristic factors, and the working status of the corresponding physical protection facilities is controlled based on the protection strategy.

[0023] This invention, through real-time acquisition and fusion processing of multi-source data, can comprehensively and accurately perceive the actual risk status of road areas under severe weather conditions, overcoming the limitations and errors that may exist in single data sources. The established multi-source data fusion model realizes effective data interpolation and supplementation, ensuring the reliability of key risk characteristic factor extraction and providing a high-quality data foundation for subsequent risk assessment. The warning level is determined based on the comprehensive risk score, making the warning results more scientific and quantitative, accurately reflecting the risk differences in different road areas. Furthermore, the generation of protection strategies and control of physical protection facilities based on the warning level and key risk characteristic factors achieves closed-loop management from risk perception to proactive protection, effectively reducing the impact of severe weather on road traffic safety, improving the safety and reliability of road operations, and providing timely and effective decision support for traffic management departments and travelers.

[0024] In some embodiments of this application, when preprocessing the meteorological data and road condition data, the preprocessing includes data cleaning, standardization, and outlier removal.

[0025] Understandably, by eliminating noise, inconsistencies, and invalid information in the original data, data quality can be improved, providing an accurate, reliable, and uniformly formatted data foundation for subsequent multi-source data fusion, feature extraction, and risk assessment.

[0026] Specifically, data cleaning involves checking the integrity and correcting the format of the collected raw meteorological data (such as temperature, precipitation, wind speed, visibility, etc.) and road condition data (such as road surface temperature, friction coefficient, water depth, snow thickness, camera image recognition results, etc.). Data containing null values, garbled characters, duplicate records, or misplaced timestamps due to sensor malfunctions, communication interruptions, or transmission errors is removed, and missing fields are marked to provide a basis for subsequent interpolation. Standardization is also crucial because data from different sources have different dimensions and numerical ranges (e.g., wind speed is measured in m / s, humidity in percentage, and friction coefficient is dimensionless). Direct fusion can lead to a model bias towards features with larger numerical values. Therefore, data standardization is necessary to convert various types of data to a uniform numerical range (e.g., 0–1 or a distribution with a mean of 0 and a variance of 1) to eliminate the influence of dimensions and ensure that each feature has comparability and equal weight during fusion and scoring. Outlier removal: Statistical methods (such as the 3σ principle and box plots) or rules based on physical plausibility (such as road surface temperature not being lower than -50℃ and visibility not exceeding 50 kilometers) are used to identify and remove outlier observations that significantly deviate from the normal range. These outliers may originate from equipment false alarms or extreme, accidental events; retaining them would severely interfere with the accuracy of risk feature extraction and comprehensive scoring. After removal, appropriate interpolation can be performed based on context or the fusion model to ensure data continuity.

[0027] In some embodiments of this application, a multi-source data fusion model is established based on preprocessed meteorological data and road condition data, including: acquiring historical disaster data and vehicle traffic data; performing time alignment and spatial standardization on the historical disaster data, vehicle traffic data, meteorological data, and road condition data; extracting data features for each type of data; and fusing the data features through an intermediate fusion method and an attention mechanism to obtain a multi-source data fusion model.

[0028] Understandably, by integrating road-related data from different sources and dimensions, a multi-source data fusion model can be constructed that comprehensively and accurately reflects the risk status of roads under severe weather conditions. Through time alignment, spatial standardization, feature extraction, and the design of fusion mechanisms, inconsistencies in the spatiotemporal scale and semantic level of multi-source data are eliminated, improving the model's ability to perceive complex risk scenarios and its generalization performance.

[0029] Specifically, in addition to preprocessed meteorological and road condition data, historical disaster data (such as records of landslides, flooding, icing, etc., that occurred on a certain road section over the past N years) and vehicle traffic data (such as traffic flow, average vehicle speed, frequency of emergency braking, trajectory deviation rate, etc.) are further introduced. Since these data come from different systems, have different sampling frequencies, and coordinate systems, time alignment is required first, that is, all data are unified to the same time window (such as aggregating in 5-minute or 10-minute units) to ensure that the data are comparable at the same time; at the same time, spatial standardization is performed, mapping different data to a unified road area division unit (such as dividing by road segment ID, kilometer markers, or grid), so that various types of data correspond one-to-one in space.

[0030] Representative features are extracted for each of the four types of data. For example, precipitation intensity variation rate and visibility decline trend are extracted from meteorological data; friction coefficient fluctuation and snow accumulation rate are extracted from road condition data; disaster frequency and recent disaster time decay factor are extracted from historical disaster data; and speed standard deviation and abnormal driving behavior density are extracted from vehicle traffic data. These features can more effectively characterize the dynamic evolution of risk, rather than relying solely on raw observations.

[0031] A mid-stage fusion approach combined with an attention mechanism is employed for feature fusion: Mid-stage fusion involves extracting features independently from various data types before fusing them at the feature level. Compared to early fusion (directly concatenating raw data), this method preserves more semantic information from each modality and overcomes the insufficient collaborative ability of late fusion (merging results after separate modeling). Building upon this, an attention mechanism is introduced to dynamically allocate fusion weights for each feature based on the current risk context of the road area. For example, in foggy weather, the attention weights for visibility and camera image recognition features are automatically increased; in icy and snowy weather, the weights for road surface temperature, icing probability, and friction coefficient are enhanced. Through this mechanism, the model can adaptively focus on the information source that contributes the most to the current risk, thereby generating a more discriminative fusion representation and ultimately forming a multi-source data fusion model.

[0032] In some embodiments of this application, the multi-source data fusion model is used for data interpolation and supplementation, and key risk characteristic factors are extracted. This includes: determining the data missing type, missing spatial region, and missing time window, and calculating the missing rate; determining the priority of the interpolation strategy based on the missing spatial region, missing time window, and missing rate; and calling the interpolation submodule of the multi-source data fusion model to interpolate and supplement the data according to the data missing type and interpolation strategy priority. After data interpolation and supplementation are completed, key risk characteristic factors are extracted from the fused data, wherein the key risk characteristic factors include atmospheric... Meteorological risk characteristics include: meteorological risk factors, road risk characteristics, historical disaster risk characteristics, and traffic risk characteristics. Meteorological risk characteristics include: temperature, precipitation / snowfall, wind speed, visibility, probability of road icing, and humidity. Road risk characteristics include: road surface temperature, friction coefficient, water depth, snow thickness, and camera image recognition results. Historical disaster risk characteristics include: frequency of disasters occurring on a certain road section in the past N years, time since the most recent disaster, and disaster severity level. Traffic risk characteristics include: average vehicle speed, traffic volume, speed standard deviation, frequency of sudden braking events, and trajectory deviation rate.

[0033] Understandably, when multi-source data is missing or incomplete, intelligent interpolation and supplementation mechanisms are used to restore data integrity. Based on this, key feature factors that can comprehensively characterize the road risk status are systematically extracted, providing reliable and structured input for subsequent risk scoring and early warning decisions.

[0034] Specifically, the preprocessed multi-source data is first subjected to missing data diagnosis to identify the missing type of each data category (e.g., random missing data, missing data over consecutive time periods, sensor failure across an entire area, etc.) and determine the specific spatial region (e.g., a road segment or grid cell) and time window (e.g., a specific hour, a specific day, or multiple consecutive time periods) where the missing data occurs. Simultaneously, the missing rate of each data type in different regions and time periods is calculated. Based on this information, the impact of missing data on risk assessment is comprehensively evaluated to determine the priority of the imputation strategy. For example, if road surface temperature data for a critical road segment is continuously missing with a high missing rate during heavy snowfall, it is given a high priority; while sporadic missing data in short-term, low-impact areas can be handled with a lower priority strategy. The multi-source data fusion model incorporates multiple imputation sub-modules, each suitable for different types and scenarios of missing data. Based on the determined data missing type and imputation priority, the most suitable imputation method is dynamically selected. For example, meteorological data with strong spatiotemporal continuity can be supplemented using Kriging interpolation based on nearby spatiotemporal points or time-series prediction models (such as LSTM); traffic data can be extrapolated using historical data from the same road segment or the similarity of traffic flow between adjacent road segments; for large-scale missing data caused by equipment failure, indirect reconstruction can be performed using other relevant features in the fusion model (such as inferring the probability of icing through visibility and humidity). The interpolation process fully utilizes the correlation and redundancy between multi-source data to improve the accuracy and physical plausibility of the supplemented results.

[0035] After data interpolation and supplementation, four key risk characteristic factors are extracted from the fused complete dataset: meteorological risk characteristic factors, including weather temperature, precipitation / snowfall, wind speed, visibility, road icing probability, and humidity, to reflect the direct impact of current weather on road safety; road risk characteristic factors, including road surface temperature, friction coefficient, water depth, snow thickness, and camera image recognition results (such as whether slippage, congestion, or abnormal obstacles are detected), to directly characterize the actual traffic conditions on the road surface; historical disaster risk characteristic factors, including the frequency of disasters on a certain road section in the past N years, the time since the most recent disaster (usually using exponential decay weighting), and the severity level of historical disasters, to measure the vulnerability of the road section under similar weather conditions; and traffic risk characteristic factors, including average vehicle speed, traffic volume, speed standard deviation (reflecting traffic flow stability), frequency of sudden braking events, and trajectory deviation rate (reflecting abnormal driver operation), to inversely determine the road risk level from the perspective of vehicle behavior.

[0036] In some embodiments of this application, determining the comprehensive risk score for each road area based on the key risk characteristic factors includes: constructing a risk assessment index system using the analytic hierarchy process (AHP), using meteorological risk characteristic factors, road risk characteristic factors, historical disaster risk characteristic factors, and traffic risk characteristic factors as primary indicators, and assigning corresponding weights to the characteristic factors under each primary indicator. These weights are determined comprehensively using an expert scoring method combined with an entropy weighting method. The key risk characteristic factors are standardized and converted into a unified scoring range. Specifically, the indicator type for each key risk characteristic factor is determined, including positive and negative indicators. For positive indicators, a standardization method favoring larger values ​​is adopted, while for negative indicators, a standardization method favoring smaller values ​​is adopted. The comprehensive risk score for each road area is calculated using a weighted summation method.

[0037] Understandably, based on the extracted multidimensional key risk characteristic factors, a scientific, objective, and quantifiable comprehensive risk scoring model is constructed to achieve a unified assessment and horizontal comparison of the risk levels of various road areas under severe weather conditions, providing a numerical basis for the subsequent generation of graded early warning and protection strategies.

[0038] Specifically, a hierarchical risk assessment indicator system is constructed: a structured risk assessment framework is established using the Analytic Hierarchy Process (AHP), with meteorological risk characteristic factors, road risk characteristic factors, historical disaster risk characteristic factors, and traffic risk characteristic factors as four primary indicators; each primary indicator contains several specific secondary characteristic factors (e.g., precipitation / snowfall and visibility are included under "meteorological risk"). This system clarifies the logical relationships and assessment levels of each risk dimension, making the risk calculation process clear and interpretable. To balance subjective experience with objective data distribution, a combination of expert scoring and entropy weighting is used to determine weights. Expert scoring reflects the prior knowledge of domain experts regarding the importance of different risk factors and is suitable for factors that are difficult to directly reflect from data but have practical significance (e.g., the severity level of historical disasters); entropy weighting, based on the information entropy of actual observation data, automatically assigns higher weights to factors with high variability and high information content, avoiding human bias. Finally, the results of the two methods are combined through weighted averaging or linear combination to obtain the comprehensive weight of each primary indicator and its subordinate characteristic factors, ensuring that the assessment results are both professional and data-driven.

[0039] The characteristic factors are standardized and their index types are distinguished: Due to the significant differences in the dimensions and numerical ranges of the various characteristic factors (e.g., traffic volume can reach thousands, while the friction coefficient is only between 0 and 1), they need to be converted to a unified scoring range (usually 0–1 or 0–100). In this process, the index type of each factor is first determined: Positive indicators: the larger the value, the higher the risk (e.g., precipitation / snowfall, frequency of sudden braking, snow depth), using a "larger is better" standardization formula (e.g., linear mapping or standardization of extremely large indicators); Negative indicators: the smaller the value, the higher the risk (e.g., visibility, friction coefficient, average vehicle speed in certain scenarios), using a "smaller is better" standardization method (e.g., reciprocal transformation or standardization of extremely small indicators). After standardization, all factors express risk contribution on the same scale, avoiding scoring distortion due to differences in dimensions.

[0040] Calculating the comprehensive risk score: Each standardized key risk characteristic factor is multiplied by its corresponding comprehensive weight, and then weighted and summed hierarchically (e.g., first calculate the sub-item scores under each primary indicator, then sum them into a total score), ultimately obtaining the comprehensive risk score for each road area. This score comprehensively reflects the overall risk level under the combined effects of current weather, road conditions, historical vulnerability, and real-time traffic behavior, providing a quantitative basis for subsequently setting warning levels.

[0041] In some embodiments of this application, determining the warning level of each road area based on the comprehensive risk score includes: pre-setting a first score, a second score, and a third score, wherein the first score, the second score, and the third score increase sequentially; setting the warning level of the corresponding road area based on the relationship between the comprehensive risk score and the first score, the second score, and the third score; if the comprehensive risk score is less than the first score, then setting the warning level of the corresponding road area to a low-risk level; if the comprehensive risk score is greater than or equal to the first score and less than the second score, then setting the warning level of the corresponding road area to a medium-risk level; if the comprehensive risk score is greater than or equal to the second score and less than the third score, then setting the warning level of the corresponding road area to a high-risk level; if the comprehensive risk score is greater than or equal to the third score, then setting the warning level of the corresponding road area to an extremely high-risk level.

[0042] It is understandable that the quantitative comprehensive risk score can be transformed into an early warning level with clear management significance and operational guidance, so as to realize the graded identification and differentiated response to the risk status of different road areas, thereby improving the accuracy and efficiency of traffic management under severe weather conditions.

[0043] Specifically, three increasing scoring thresholds are set in advance according to historical accident data, expert experience, and system operation objectives, denoted as the first score (T1), the second score (T2), and the third score (T3) in sequence, satisfying T1 < T2 < T3. These three thresholds divide the continuous value space of the comprehensive risk score into four intervals, corresponding to the four warning levels of low, medium, high, and extremely high respectively. For example, T1 = 30, T2 = 60, and T3 = 85 can be set (assuming the scoring range is 0–100). If the comprehensive risk score of a certain road area is 25 (less than T1 = 30), it is determined as the low-risk level, indicating that the current environment and road conditions are basically safe, and only routine monitoring is required; if the score is 45 (≥30 and <60), it is determined as the medium-risk level, suggesting the existence of potential safety hazards, and enhanced monitoring and preliminary intervention measures should be prepared; if the score is 75 (≥60 and <85), it is determined as the high-risk level, indicating that obvious risk factors have emerged, and targeted protection strategies should be initiated, such as issuing warning messages, adjusting speed limits, etc.; if the score is 90 (≥85), it is determined as the extremely high-risk level, meaning that a serious safety accident is very likely to occur, and coercive measures should be taken immediately, such as closing the road, guiding detours, etc.

[0044] This grading mechanism simplifies the complex multi-dimensional risk assessment results into clear and executable four-level response instructions, facilitating rapid decision-making by traffic management departments. At the same time, the setting of the thresholds is configurable and can be dynamically adjusted according to different regions, seasons, or road types, enhancing the adaptability and practicality of the system. In addition, this method avoids the "one-size-fits-all" type of warning, achieving a match between resource investment and risk level, ensuring safety while reducing unnecessary traffic interference.

[0045] In some embodiments of the present application, before determining the warning level of each road area according to the comprehensive risk score, it includes: setting the threshold of each key risk characteristic factor, comparing the corresponding value of the key risk characteristic factor with the threshold, and if the corresponding value of the key risk characteristic factor is greater than the threshold, directly determining the warning level of the corresponding road area as the extremely high-risk level.

[0046] It can be understood that before the conventional grading warning based on the comprehensive risk score, a "one-vote veto" type of threshold judgment mechanism for key risk characteristic factors is introduced to cope with certain extremely dangerous situations. In such situations, even if the comprehensive score has not reached the extremely high-risk threshold, but its single factor is already sufficient to directly threaten driving safety, and the highest-level warning must be triggered immediately to ensure the timeliness and safety of the response.

[0047] Specifically, safety thresholds are pre-set for several key risk characteristics with high hazard potential. These factors are usually closely related to sudden high-risk events, such as: visibility ≤ 50 meters; snow thickness ≥ 15 centimeters; road icing probability ≥ 90%; frequency of sudden braking events ≥ 10 times per minute; and coefficient of friction ≤ 0.2.

[0048] The above thresholds are determined based on traffic engineering standards, historical accident statistics, and expert experience, and represent the critical state that is highly likely to cause an accident once exceeded.

[0049] After calculating the comprehensive risk score and before formally classifying the warning level, the system checks whether the actual value of each key risk characteristic factor exceeds its corresponding threshold. If any factor exceeds the standard (e.g., the actual visibility is 40 meters, which is below the 50-meter threshold), regardless of the comprehensive risk score (even if it is only 70, which is below the third score of 85), the system immediately skips the regular classification process and directly sets the warning level of the road area to an extremely high risk level.

[0050] For example, suppose a mountain road section experiences heavy fog, with overall stable meteorological and traffic data, resulting in a comprehensive risk score of 78 (high-risk). However, real-time monitoring shows a sudden drop in visibility to 30 meters, below the preset 50-meter threshold. In this situation, the system does not wait for the scoring and classification results but immediately determines the road section as extremely high-risk and automatically triggers emergency measures such as road closures, electronic display warnings, and navigation platform-push detour suggestions, avoiding response delays due to scoring lag or weight dilution.

[0051] In some embodiments of this application, a protection strategy is generated based on the warning level and key risk characteristic factors, including: if the warning level is low risk, no protection strategy is generated and monitoring continues; if the warning level is medium risk or high risk, a protection strategy is generated; if the warning level is high risk, an emergency response mechanism is immediately activated, the road area is closed, and vehicles are clearly informed of the prohibition of passage through road closure warning signs and electronic displays.

[0052] Understandably, based on the warning level and key risk characteristic factors, protective strategies or emergency measures that match the degree of risk are dynamically generated, achieving precise connection from risk identification to response, avoiding excessive intervention in low-risk scenarios, and ensuring that mandatory safety control can be quickly implemented in high-risk situations.

[0053] Specifically, when a road area is classified as low-risk, it indicates that the current weather, road conditions, and traffic operations are generally within a safe and controllable range. At this time, the system does not trigger any proactive protection strategies, but only maintains real-time data collection and monitoring to promptly detect risk trends and reserve time for potential escalation responses. If the warning level is medium-risk or high-risk, it indicates that there are identifiable safety hazards, but not yet to the point of requiring road closure. The system will generate corresponding protection strategies based on specific key risk characteristics (such as precipitation / snowfall, visibility, humidity, and friction coefficient). For example, when increased snowfall is detected but does not reach the closure standard, snow removal equipment can be preheated or a speed reduction warning can be issued; when visibility drops to a critical value but is above the meltdown threshold, lighting can be enhanced or fog lights can be activated for guidance. These strategies aim to "mitigate risk and maintain traffic flow," emphasizing a balance between proactive intervention and traffic efficiency. When the warning level is extremely high-risk (whether triggered by a comprehensive score exceeding the third score or by a single key factor exceeding a threshold), the system will no longer generate conventional protection strategies but will immediately activate the emergency response mechanism. Specific measures include: automatically raising road closure warning signs, displaying "No Entry" information on electronic screens, sending road closure instructions to navigation platforms, and coordinating with traffic police to forcibly prevent vehicles from entering dangerous areas and minimize the occurrence of accidents.

[0054] In some embodiments of this application, a protection strategy is generated based on the warning level and key risk characteristic factors, including: preset precipitation / snowfall limits, visibility limits, and humidity limits; comparing the precipitation / snowfall with the precipitation / snowfall limits, comparing the visibility with the visibility limits, and comparing the humidity with the humidity limits; if the precipitation / snowfall is greater than the precipitation / snowfall limits, a shielding strategy is generated; if the visibility is less than the visibility limits, or the humidity is greater than the humidity limits, a warm wind defogging strategy is generated.

[0055] Understandably, when the warning level reaches medium or high risk, further analysis of the comparison results between specific key meteorological risk characteristics (such as precipitation / snowfall, visibility, and humidity) and preset limits can generate targeted and executable physical protection strategies, so that the protection measures are accurately matched with the actual environmental threats and the effectiveness of proactive prevention and control can be improved.

[0056] Specifically, the system pre-sets safe operating limits for three types of meteorological factors as the basis for triggering specific protective strategies. For example: precipitation / snowfall limit: set at 10 mm / hour (representing moderate to heavy rain or moderate snow); visibility limit: set at 200 meters (below this value is prone to rear-end collisions); humidity limit: set at 90% (high humidity environments are prone to fog or condensation on low-temperature road sections). These limits are determined comprehensively based on road engineering specifications, equipment capabilities, and historical accident data, and have clear engineering significance.

[0057] If the monitored precipitation / snowfall exceeds the preset limit (e.g., the measured snowfall rate reaches 12 mm / hour), the system determines that there is a risk of rain erosion, snow cover, or obstructed visibility. At this time, a shielding strategy is generated to reduce the direct impact of precipitation on the road surface and the driver's vision. If the visibility is less than the visibility limit, or the humidity is higher than 90%, it indicates that the air is saturated with water vapor, which is likely to form dense fog or condensation on bridges, tunnel entrances, or low-temperature road sections at night, affecting visibility and road surface adhesion. At this time, the system generates a warm air defogging strategy to disperse the fog or prevent condensation by heating the airflow.

[0058] For example, at a highway bridge section, the system determines the warning level to be high-risk. Simultaneously, monitoring data shows: current snowfall is 15 mm / hour (exceeding the 10 mm limit), visibility is 180 meters (below the 200 meter limit), and humidity is 92%. According to the rules, the system simultaneously triggers a shielding strategy (to cope with heavy snowfall) and a warm wind defogging strategy (to cope with low visibility and high humidity). These two strategies will control the activation of corresponding physical facilities to achieve multi-dimensional collaborative protection.

[0059] In some embodiments of this application, the working state of the corresponding physical protection facility is controlled based on the protection strategy, including: the physical protection facility includes: a protective cover, a lighting unit, and a warm air defogging unit; if the protection strategy includes a shielding strategy, the protective cover and the lighting unit are activated; if the protection strategy includes a warm air defogging strategy, the warm air defogging unit and the lighting unit are activated.

[0060] Understandably, the generated protection strategies are transformed into specific control instructions for physical protection facilities, achieving a closed-loop linkage from digital decision-making to physical execution. This ensures that protection measures can be applied to the actual road environment in a timely and accurate manner, thereby effectively reducing the driving safety risks brought about by severe weather.

[0061] Specifically, the physical protection facilities associated with the system include three core types of equipment: Protective covers: These can be deployed above bridges, curves, or accident-prone road sections to provide partial shelter during heavy rainfall or snowfall, reducing the impact of rainwater erosion or snow cover on the road surface and driver visibility; Lighting units: These include high-brightness LED streetlights, contour lights, or fog lights to enhance road visibility in reduced visibility or nighttime conditions, assisting drivers in identifying lane lines and obstacles; Warm air defogging units: Installed in areas prone to fogging (such as tunnel entrances and exits, and bridges across rivers), these units suppress or disperse fog and condensation on the road surface by blowing heated airflow, improving visibility and road surface friction performance.

[0062] When the system-generated protection strategy includes a shielding strategy (usually triggered by precipitation / snowfall exceeding the limit), it automatically sends a command to the control system of the corresponding road segment to activate the protective shield deployment and simultaneously turn on the lighting unit. The lighting unit is activated to compensate for the reduced light caused by the shielding structure and ensure sufficient driver visibility. When the protection strategy includes a heated air defogging strategy (triggered by visibility below the limit or excessive humidity), the heated air defogging unit is activated to deliver heated air, and the lighting unit is turned on simultaneously to collaboratively improve visual conditions and road surface conditions in low-visibility environments.

[0063] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0064] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0065] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0066] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for multi-source perception and graded early warning of road risks under severe weather conditions, characterized in that, The method includes: Real-time collection of meteorological and road condition data for each road area, and preprocessing of the meteorological and road condition data; A multi-source data fusion model is established based on preprocessed meteorological data and road condition data. The multi-source data fusion model is used for data interpolation and supplementation, and key risk characteristic factors are extracted. A comprehensive risk score is determined for each road area based on the key risk characteristic factors, and a warning level is determined for each road area based on the comprehensive risk score. A protection strategy is generated based on the warning level and key risk characteristic factors, and the working status of the corresponding physical protection facilities is controlled based on the protection strategy.

2. The method for multi-source perception and graded early warning of road risks under severe weather conditions according to claim 1, characterized in that, When preprocessing the meteorological data and road condition data, the preprocessing includes data cleaning, standardization, and outlier removal.

3. The method for multi-source perception and graded early warning of road risks under severe weather conditions according to claim 1, characterized in that, A multi-source data fusion model is established based on preprocessed meteorological data and road condition data, including: Historical disaster data and vehicle traffic data are acquired, and the historical disaster data, vehicle traffic data, meteorological data, and road condition data are time-aligned and spatially standardized. Data features are extracted for each type of data, and the data features are fused through intermediate fusion and attention mechanisms to obtain a multi-source data fusion model.

4. The method for multi-source perception and graded early warning of road risks under severe weather conditions according to claim 3, characterized in that, The multi-source data fusion model is used for data interpolation and supplementation, and key risk characteristic factors are extracted, including: The missing data type, missing spatial region, and missing time window are determined, and the missing rate is calculated. Based on the missing spatial region, missing time window, and missing rate, the priority of the imputation strategy is determined. Based on the missing data type and the priority of the imputation strategy, the imputation submodule of the multi-source data fusion model is called to interpolate and supplement the data. After data interpolation and supplementation, key risk characteristic factors are extracted from the fused data. These key risk characteristic factors include meteorological risk characteristic factors, road risk characteristic factors, historical disaster risk characteristic factors, and traffic risk characteristic factors. Meteorological risk characteristic factors include: temperature, precipitation / snowfall, wind speed, visibility, road icing probability, and humidity. Road risk characteristic factors include: road surface temperature, friction coefficient, water depth, snow thickness, and camera image recognition results. Historical disaster risk characteristic factors include: frequency of disasters occurring on a certain road section in the past N years, time since the most recent disaster, and disaster severity level. Traffic risk characteristic factors include: average vehicle speed, traffic volume, speed standard deviation, frequency of sudden braking events, and trajectory deviation rate.

5. The method for multi-source perception and graded early warning of road risks under severe weather conditions according to claim 4, characterized in that, The comprehensive risk score for each road area is determined based on the aforementioned key risk characteristic factors, including: A risk assessment index system was constructed using the analytic hierarchy process (AHP). Meteorological risk characteristic factors, road risk characteristic factors, historical disaster risk characteristic factors, and traffic risk characteristic factors were used as primary indicators. Each characteristic factor under a primary indicator was assigned a corresponding weight, which was determined by a combination of expert scoring and entropy weighting. Each key risk characteristic factor is standardized and converted into a unified scoring range. The indicator type for each key risk characteristic factor is determined, including positive and negative indicators. For positive indicators, a standardization method of "the larger the better" is adopted, and for negative indicators, a standardization method of "the smaller the better" is adopted. The comprehensive risk score for each road area is calculated by weighted summation.

6. The method for multi-source perception and graded early warning of road risks under severe weather conditions according to claim 3, characterized in that, The warning level for each road area is determined based on the comprehensive risk score, including: A first score, a second score, and a third score are preset, with the first score, the second score, and the third score increasing sequentially. The warning level for the corresponding road area is set according to the relationship between the comprehensive risk score and the first score, the second score, and the third score. If the comprehensive risk score is less than the first score, the warning level for the corresponding road area is set to low risk. If the comprehensive risk score is greater than or equal to the first score and the comprehensive risk score is less than the second score, then the warning level for the corresponding road area is set to medium risk. If the comprehensive risk score is greater than or equal to the second score, and the comprehensive risk score is less than the third score, then the warning level for the corresponding road area is set to a high-risk level. If the comprehensive risk score is greater than or equal to the third score, then the warning level for the corresponding road area is set to an extremely high risk level.

7. The method for multi-source perception and graded early warning of road risks under severe weather conditions according to claim 6, characterized in that, Before determining the warning level for each road area based on the comprehensive risk score, the following steps are included: A threshold is set for each key risk characteristic factor. The corresponding value of the key risk characteristic factor is compared with the threshold. If the corresponding value of the key risk characteristic factor is greater than the threshold, the warning level of the corresponding road area is directly determined to be an extremely high risk level.

8. The method for multi-source perception and graded early warning of road risks under severe weather conditions according to claim 7, characterized in that, Based on the aforementioned warning level and key risk characteristic factors, a protection strategy is generated, including: If the warning level is low risk, no protection strategy will be generated, and monitoring will continue. If the warning level is medium risk or high risk, a protection strategy is generated. If the warning level is high-risk, the emergency response mechanism will be activated immediately, the road area will be closed, and vehicles will be clearly informed of the prohibition of passage through road closure warning signs and electronic displays.

9. The method for multi-source perception and graded early warning of road risks under severe weather conditions according to claim 8, characterized in that, Based on the aforementioned warning level and key risk characteristic factors, a protection strategy is generated, including: Preset limits for precipitation / snowfall, visibility, and humidity; The precipitation / snowfall is compared with the precipitation / snowfall limit, the visibility is compared with the visibility limit, and the humidity is compared with the humidity limit. If the precipitation / snowfall exceeds the precipitation / snowfall limit, a shading strategy is generated. If the visibility is less than the visibility limit, or the humidity is greater than the humidity limit, a warm air defogging strategy is generated.

10. The method for multi-source perception and graded early warning of road risks under severe weather conditions according to claim 9, characterized in that, Controlling the operational status of the corresponding physical protection facilities based on the aforementioned protection strategy includes: The physical protection facilities include: a protective cover, a lighting unit, and a warm air defogging unit; If the protection strategy includes a shielding strategy, then the protective shield and lighting unit are activated; If the protection strategy includes a warm air defogging strategy, then the warm air defogging unit and the lighting unit are activated.

Images