Highway situation awareness and accident risk early warning method fusing multi-source data

Abstract

The invention discloses a highway situation awareness and accident risk early warning method fusing multi-source data. The method comprises the following steps: S1, constructing multi-source data fusion and multi-scale traffic flow data; s2, studying and judging an accident influence factor mechanism, and excavating factors influencing the occurrence of a traffic accident; s3, early warning of accident risks is carried out, and the safety risk level of the road section is evaluated; and S4, researching the beforehand management and control system, and constructing a beforehand management and control strategy system based on deep mining analysis of traffic accidents and actual operation experience. According to the invention, on the basis of the high-speed portal data and the charging flow data, the detection data of the thunder-vision integrated roadside equipment are fused for fusion processing, so that fine traffic operation parameters such as traffic flow and vehicle speed before and after a historical accident are obtained; traffic flow characteristics causing accidents are mined, and a traffic accident risk dynamic early warning management and control model and a post-accident evaluation mechanism are established based on the traffic flow characteristics, so that a management department is helped to recognize potential risks in advance.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic risk warning, and in particular to a highway situation awareness and accident risk warning method integrating multi-source data. Background Art

[0002] As the level of intelligence and digitization of highways gradually improves, sensing equipment such as radar, video surveillance, and ETC toll gantries are widely used and generate a large amount of real-time highway operation time-space multi-source big data. How to fuse multi-source heterogeneous data according to business needs and identify accident risk-related factors, realize traffic situation perception and accident risk warning based on multi-source fusion big data, provide reliable early warning access for traffic safety management, and improve road safety levels, is the focus and purpose of this invention.

[0003] The accident risk on highways is affected by multiple factors, including driver behavior, vehicle performance, road conditions, and traffic flow. However, due to the challenges in obtaining real-time traffic data, traditional research can only rely on the frequency or ratio of traffic accidents to assess risks, and use long-term traffic parameters (such as annual average daily traffic volume) to establish a highway accident risk warning model. Although this method can provide guidance for long-term traffic safety management from a macro perspective, it cannot capture the impact of short-term traffic flow changes on accident risks, nor can it achieve real-time prediction of accident probability. Therefore, it is impossible to formulate effective management measures based on the immediate changes in traffic flow, which is inconsistent with the current demand for active prevention of traffic accidents. To address this issue, subsequent research uses traffic monitoring cameras, roadside sensors and other devices to collect traffic flow parameters such as vehicle flow, speed and acceleration; in addition, sensors are installed on test vehicles to collect driving behavior data of individual vehicles. These data are then used through statistical analysis and machine learning techniques to develop a risk prediction model that can assess accident probability in real time.

[0004] Nevertheless, the deployment and maintenance of these traffic sensing devices require high economic investment, and the accuracy of their data collection is easily affected by weather and lighting conditions. At the same time, the data collected by vehicle sensors is limited to a single vehicle and cannot cover the driving dynamics of other vehicles and changes in the overall traffic flow. Therefore, only the accident risk of a specific vehicle can be assessed. Summary of the invention

[0005] The purpose of the present invention is to provide a highway situation awareness and accident risk warning method that integrates multi-source data to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a highway situation awareness and accident risk warning method integrating multi-source data, the warning method comprising the following steps:

[0007] S1, construct multi-source data fusion and multi-scale traffic flow data, collect the gantry traffic flow data of the highway network toll collection system, the in and out traffic data of the toll station, and the roadside integrated radar and vision perception equipment data as the basic data for fusion analysis;

[0008] S2, to study the mechanism of factors affecting accidents, based on historical multi-scale, full-coverage, real-time and detailed traffic operation data, combined with the accident type, time of occurrence, area of ​​occurrence, location direction, lane occupation and congestion factors to conduct correlation analysis, and explore the factors affecting the occurrence of traffic accidents;

[0009] S3, early warning of accident risks, and by building a risk index assessment model, quantitatively analyzing the correlation between various influencing factors and accident risks, and taking each influencing factor as the characteristic parameter of the model, the assessment model outputs the accident risk index, and by setting the risk index threshold, the safety risk level of the road section is assessed;

[0010] S4, study the pre-emptive control system, build a pre-emptive control strategy system based on in-depth analysis of traffic accidents and actual operating experience, and ensure that highways have a high degree of response capabilities under different risk situations.

[0011] Preferably, the specific operation steps of the multi-source data fusion and multi-scale traffic flow data construction are:

[0012] S101, perform data governance, and directly delete the data that is repeatedly written, does not meet the time and space constraints, has poor quality, is redundant, and is abnormal in the toll station data, gantry data, and sensor data;

[0013] S102, supplement missing data. For missing data caused by integrated radar and vision, the average value of historical data for the same period is used for supplementation. For high-speed toll flow data and high-speed gantry flow data, the data is supplemented according to the topological structure in the highway network model.

[0014] S103, perform spatiotemporal correlation fusion of multi-source data, including space fusion and time field fusion. The fused multi-source big data can be used to mine speed differences in different gantry sections, forming a multi-scale, full-coverage, real-time, and precise traffic operation parameter monitoring database for the target highway, providing a data basis for accident factor mining.

[0015] Preferably, the spatial fusion performs spatial matching based on the latitude and longitude positions of highway gantries, toll booths and roadside equipment to obtain large-scale and small-scale traffic flow data of the highway.

[0016] Preferably, the time field fusion is resampled at a 5-minute granularity based on the data collected by highway gantries, toll stations and roadside equipment to achieve symmetry in its time granularity.

[0017] Preferably, the mining of factors causing traffic accidents is to mine the influence mechanism of single and multiple combined factors on the type and level of traffic accidents, to find the characteristics that induce accidents, including changes in traffic density and speed, and to analyze the changes in accident incidence under different traffic density, speed, flow and passenger-to-cargo ratio conditions.

[0018] Preferably, the analysis of the change in accident incidence rate includes analysis of flow interval and accident number, analysis of flow interval and accident ratio, and analysis of truck ratio and accident number.

[0019] Preferably, the flow interval and accident number analysis is based on the flow of the preceding gantry and roadside equipment integrated 5 minutes before the accident, dividing the flow into different intervals, counting the number of accidents in each flow interval, and analyzing the relationship between different flow levels and accident occurrence;

[0020] The flow interval and accident ratio analysis is to collect statistics on the flow data of each flow interval and analyze the ratio of flow to accidents in different intervals;

[0021] The analysis of truck ratio and accident number is based on the flow of the preceding gantry and roadside equipment integrated 5 minutes before the accident, and analyzes the frequency relationship of accidents under different truck ratios.

[0022] Preferably, the specific operation steps of the accident risk warning are:

[0023] S301, constructing an influencing factor feature vector, extracting multi-scale traffic operation parameters of the section according to the mechanism judgment result of the accident influencing factor, wherein the operation parameters include flow, density, speed and passenger-cargo ratio, and forming an influencing factor feature vector;

[0024] S302, predicting the traffic flow characteristics of the interval, based on the historical traffic flow characteristic data of the interval, using a long short-term memory network for prediction;

[0025] S303, perform accident risk prediction, build an accident risk prediction model based on the historical traffic characteristic data of the interval using the Xgboost method, and realize accident risk probability prediction by inputting the historical traffic characteristic data of the interval.

[0026] Preferably, the operation steps of the accident risk prediction are:

[0027] The first step is to construct a set of traffic accident-related feature vectors, expressed as:

[0028] (1);

[0029] in, represents the feature vector of the mth sample, and n represents the number of feature vectors;

[0030] In the second step, the feature vectors of m samples are input into Xgboost for training to establish the relationship between traffic accident characteristics and accident risks, which can be expressed as:

[0031] (2);

[0032] in, The optimal parameters of the Xgboost model are obtained through k-fold cross validation, which is used to characterize the mapping relationship from influencing factors to accident risks.

[0033] Preferably, the pre-event control system automatically predicts the potential accident risk level through real-time dynamic monitoring of traffic flow data combined with the evaluation model, and automatically generates multi-level pre-event control plans and response measures according to different risk levels and accident probabilities, providing targeted action guidance, including sound and light intervention, dynamic adjustment of information boards and intelligent lane control;

[0034] The sound and light intervention uses sound and light warning devices along the highway to alert drivers to potential dangers;

[0035] The dynamic adjustment of the information board is that the upstream and downstream variable information boards dynamically display risk warnings and safe driving suggestions according to the real-time risk level, reminding the driver to slow down or change the route;

[0036] The intelligent lane control is to control the traffic flow in the high-risk traffic range by opening and closing the emergency lane and dynamically allocating lanes, so as to achieve the regulation of the traffic flow on the road section, and can also adjust the number of lanes according to the changes in the traffic flow.

[0037] Technical effects and advantages of the present invention:

[0038] The present invention proposes a highway accident warning method that integrates multi-source data. The method is based on highway gantry data and toll flow data, and integrates the detection data of integrated roadside equipment with radar and vision for fusion processing to obtain detailed traffic operation parameters such as traffic volume and speed before and after historical accidents. In combination with road line type and weather data, the influencing mechanism between traffic accidents and various factors is deeply studied, and the traffic characteristics that cause accidents are excavated. Based on this, a dynamic early warning management and control model for traffic accident risks and a post-accident evaluation mechanism are established to help management departments identify potential risks in advance and provide data support for various traffic control measures, thereby reducing the occurrence of accidents and improving the overall safety and traffic efficiency of highways. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Schematic diagram of the operation framework of the early warning method of the present invention.

[0040] Figure 2 It is a schematic diagram of the layout of the road section gantry of the present invention.

[0041] Figure 3 This is a schematic diagram of the relationship between the flow range and the number of accidents in the present invention.

[0042] Figure 4 This is a schematic diagram of the relationship between the accident rate and the interval flow rate of the present invention.

[0043] Figure 5 This is a schematic diagram of the relationship between the accident frequency and the truck ratio of the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] The present invention provides Figure 1 and Figure 2 The highway situation awareness and accident risk warning method integrating multi-source data shown includes the following steps:

[0046] S1, construct multi-source data fusion and multi-scale traffic flow data, collect the gantry traffic flow data of the highway network toll collection system, the in and out traffic data of the toll station, and the roadside integrated radar and vision perception equipment data as the basic data for fusion analysis;

[0047] Specifically, the specific steps for multi-source data fusion and multi-scale traffic flow data construction are as follows:

[0048] S101, perform data governance, and directly delete the data that is repeatedly written in the toll station data, gantry data, and sensor data, the data that does not meet the time and space constraints (too fast or too far away), the data of poor quality, redundant data, and abnormal data;

[0049] S102, supplement missing data. For missing data caused by integrated radar and vision, the average value of historical data for the same period is used for supplementation. For high-speed toll flow data and high-speed gantry flow data, the data is supplemented according to the topological structure in the highway network model.

[0050] That is, if there is a gap between the previous gantry and the next gantry, the road network topology structure is used to select the shortest path between the two gantries for inference and completion.

[0051] S103, perform spatiotemporal correlation fusion of multi-source data, including space fusion and time field fusion. The fused multi-source big data can be used to mine speed differences in different gantry sections, forming a multi-scale, full-coverage, real-time, and precise traffic operation parameter monitoring database for the target highway, providing a data basis for accident factor mining.

[0052] Table 1 Example of charging flow information

[0053]

[0054] Table 2 Example of gantry flow data

[0055]

[0056] It should be noted that a highway gantry refers to a structure installed on a highway to support traffic signs, signal lights and monitoring equipment. The design and construction of the gantry are used to improve traffic safety and management efficiency.

[0057] Furthermore, spatial fusion performs spatial matching based on the latitude and longitude positions of highway gantries, toll stations and roadside equipment to obtain large-scale and small-scale traffic flow data on the highway. Temporal field fusion resamples the data collected by highway gantries, toll stations and roadside equipment at a granularity of 5 minutes to achieve symmetry in temporal granularity.

[0058] S2, to study the mechanism of factors affecting accidents, based on historical multi-scale, full-coverage, real-time and detailed traffic operation data, combined with the accident type, time of occurrence, area of ​​occurrence, location direction, lane occupation and congestion factors to conduct correlation analysis, and explore the factors affecting the occurrence of traffic accidents;

[0059] Specifically, refer to Figure 3 - Figure 5 As shown, the mining of factors causing traffic accidents is to explore the impact mechanism of single and multiple combined factors on the type and level of traffic accidents, to find the characteristics that induce accidents, including changes in traffic density and speed, and to analyze the changes in accident rates under different traffic density, speed, flow and passenger-to-cargo ratio conditions.

[0060] Furthermore, the analysis of the change in the accident rate includes the analysis of flow interval and the number of accidents, the analysis of flow interval and the proportion of accidents, and the analysis of the proportion of trucks and the number of accidents;

[0061] The analysis of flow intervals and the number of accidents is based on the flow of the preceding gantry and roadside equipment integrated 5 minutes before the accident. The flow is divided into different intervals, the number of accidents in each flow interval is counted, and the relationship between different flow levels and accidents is analyzed;

[0062] The flow interval and accident ratio analysis is to collect statistics on the flow data of each flow interval and analyze the ratio of flow to accidents in different intervals. Through this ratio, the probability of accidents in different flow intervals can be globally evaluated, and then the flow intervals with high accident risks can be identified;

[0063] The analysis of truck ratio and accident number is based on the integrated flow of preceding gantries and roadside equipment 5 minutes before the accident, and analyzes the frequency relationship of accidents under different truck ratios.

[0064] S3, early warning of accident risks, and by building a risk index assessment model, quantitatively analyzing the correlation between various influencing factors and accident risks, and taking each influencing factor as the characteristic parameter of the model, the assessment model outputs the accident risk index, and by setting the risk index threshold, the safety risk level of the road section is assessed;

[0065] Specifically, the specific operation steps of accident risk warning are:

[0066] S301, constructing a characteristic vector of influencing factors, extracting multi-scale traffic operation parameters of the interval according to the results of the mechanism of accident influencing factors, the operation parameters include flow, density, speed and passenger-cargo ratio, and forming a characteristic vector of influencing factors;

[0067] Table 3 Feature vector construction table

[0068]

[0069] S302, predicting the traffic flow characteristics of the interval, based on the historical traffic flow characteristic data of the interval, using a long short-term memory network for prediction;

[0070] It should be noted that the long short-term memory network (LSTM) model is a deep learning model suitable for sequence data, which can capture the long-term dependencies of data. By training and adjusting the parameters of the LSTM model, the time series characteristics of traffic flow are extracted to achieve short-term traffic flow prediction in intervals of 5 minutes, 10 minutes, and 15 minutes.

[0071] S303, perform accident risk prediction, build an accident risk prediction model based on the historical traffic characteristic data of the interval using the Xgboost method, and realize accident risk probability prediction by inputting the historical traffic characteristic data of the interval.

[0072] It should be noted that XGBoost is a machine learning algorithm that is applied to classification and regression problems. It is based on the gradient boosting framework and improves the predictive performance of the model by building multiple decision trees.

[0073] Furthermore, the operation steps of accident risk prediction are:

[0074] The first step is to construct a set of traffic accident-related feature vectors, expressed as:

[0075] (1);

[0076] in, represents the feature vector of the mth sample, and n represents the number of feature vectors;

[0077] In the second step, the feature vectors of m samples are input into Xgboost for training to establish the relationship between traffic accident characteristics and accident risks, which can be expressed as:

[0078] (2);

[0079] in, The optimal parameters of the Xgboost model are obtained through k-fold cross validation, which is used to characterize the mapping relationship from influencing factors to accident risks.

[0080] S4, study the pre-emptive control system, build a pre-emptive control strategy system based on in-depth analysis of traffic accidents and actual operating experience, and ensure that highways have a high degree of response capabilities under different risk situations.

[0081] Specifically, the pre-control system uses real-time dynamic monitoring of traffic flow data combined with the assessment model to automatically predict the potential accident risk level. At the same time, according to different risk levels and accident probabilities, it automatically generates multi-level pre-control plans and response measures, and provides targeted action guidance, including sound and light intervention, dynamic adjustment of information boards and intelligent lane control, to ensure the effectiveness and timeliness of response measures. Once the risk value output by the assessment model exceeds the set threshold, the corresponding control measures will be immediately initiated;

[0082] Sound and light intervention uses sound and light warning devices along the highway to alert drivers to potential dangers and reduce the risk of accidents caused by driver negligence;

[0083] Dynamic adjustment of information boards means that upstream and downstream variable information boards dynamically display risk warnings and safe driving suggestions based on real-time risk levels, reminding drivers to slow down or change routes;

[0084] Intelligent lane control is for traffic in high-risk traffic areas. By opening and closing emergency lanes and dynamically allocating lanes, it can regulate traffic flow on road sections and reduce the possibility of rear-end collisions. It can also adjust the number of lanes according to changes in traffic flow, improve road traffic efficiency, and alleviate road congestion.

[0085] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A highway situation awareness and accident risk warning method integrating multi-source data, characterized in that: The early warning method comprises the following steps: S1, construct multi-source data fusion and multi-scale traffic flow data, collect the gantry traffic flow data of the highway network toll collection system, the in and out traffic data of the toll station, and the roadside integrated radar and vision perception equipment data as the basic data for fusion analysis; S2, to study the mechanism of factors affecting accidents, based on historical multi-scale, full-coverage, real-time and detailed traffic operation data, combined with the accident type, time of occurrence, area of ​​occurrence, location direction, lane occupation and congestion factors to conduct correlation analysis, and explore the factors affecting the occurrence of traffic accidents; S3, early warning of accident risks, and by building a risk index assessment model, quantitatively analyzing the correlation between various influencing factors and accident risks, and taking each influencing factor as the characteristic parameter of the model, the assessment model outputs the accident risk index, and by setting the risk index threshold, the safety risk level of the road section is assessed; S4, study the pre-emptive control system, build a pre-emptive control strategy system based on in-depth analysis of traffic accidents and actual operating experience, and ensure that highways have a high degree of response capabilities under different risk situations.

2. The highway situation awareness and accident risk warning method integrating multi-source data according to claim 1 is characterized in that: The specific operation steps of multi-source data fusion and multi-scale traffic flow data construction are as follows: S101, perform data governance, and directly delete the data that is repeatedly written, does not meet the time and space constraints, has poor quality, is redundant, and is abnormal in the toll station data, gantry data, and sensor data; S102, supplement missing data. For missing data caused by integrated radar and vision, the average value of historical data for the same period is used for supplementation. For high-speed toll flow data and high-speed gantry flow data, the data is supplemented according to the topological structure in the highway network model. S103, perform spatiotemporal correlation fusion of multi-source data, including space fusion and time field fusion. The fused multi-source big data can be used to mine speed differences in different gantry sections, forming a multi-scale, full-coverage, real-time, and precise traffic operation parameter monitoring database for the target highway, providing a data basis for accident factor mining.

3. The highway situation awareness and accident risk warning method integrating multi-source data according to claim 2 is characterized in that: The spatial fusion performs spatial matching according to the latitude and longitude positions of highway gantries, toll stations and roadside equipment to obtain large-scale and small-scale traffic flow data of the highway.

4. The highway situation awareness and accident risk warning method integrating multi-source data according to claim 2 is characterized in that: The time field fusion is resampled at a 5-minute granularity based on the data collected by the highway gantries, toll stations and roadside equipment to achieve symmetry in its time granularity.

5. The highway situation awareness and accident risk warning method integrating multi-source data according to claim 1 is characterized in that: The mining of factors causing traffic accidents is to find out the influence mechanism of single and multiple combined factors on the types and levels of traffic accidents, to find the characteristics that induce accidents, including changes in traffic density and speed, and to analyze the changes in accident rates under different traffic density, speed, flow and passenger-cargo ratio conditions.

6. The highway situation awareness and accident risk warning method integrating multi-source data according to claim 5 is characterized in that: The analysis of the change in accident incidence rate includes analysis of flow range and accident number, analysis of flow range and accident ratio, and analysis of truck ratio and accident number.

7. The highway situation awareness and accident risk warning method integrating multi-source data according to claim 6 is characterized in that: The analysis of flow intervals and the number of accidents is based on the flow of the preceding gantry and roadside equipment integrated 5 minutes before the accident, dividing the flow into different intervals, counting the number of accidents in each flow interval, and analyzing the relationship between different flow levels and the occurrence of accidents; The flow interval and accident ratio analysis is to collect statistics on the flow data of each flow interval and analyze the ratio of flow to accidents in different intervals; The analysis of truck ratio and accident number is based on the flow of the preceding gantry and roadside equipment integrated 5 minutes before the accident, and analyzes the frequency relationship of accidents under different truck ratios.

8. The highway situation awareness and accident risk warning method integrating multi-source data according to claim 1 is characterized in that: The specific operation steps of the accident risk warning are: S301, constructing an influencing factor feature vector, extracting multi-scale traffic operation parameters of the section according to the mechanism judgment result of the accident influencing factor, wherein the operation parameters include flow, density, speed and passenger-cargo ratio, and forming an influencing factor feature vector; S302, predicting the traffic flow characteristics of the interval, based on the historical traffic flow characteristic data of the interval, using a long short-term memory network for prediction; S303, perform accident risk prediction, build an accident risk prediction model based on the historical traffic characteristic data of the interval using the Xgboost method, and realize accident risk probability prediction by inputting the historical traffic characteristic data of the interval.

9. The highway situation awareness and accident risk warning method integrating multi-source data according to claim 8 is characterized in that: The operation steps of the accident risk prediction are: The first step is to construct a set of traffic accident-related feature vectors, expressed as: （1）； in, represents the feature vector of the mth sample, and n represents the number of feature vectors; In the second step, the feature vectors of m samples are input into Xgboost for training to establish the relationship between traffic accident characteristics and accident risks, which can be expressed as: （2）； in, The optimal parameters of the Xgboost model are obtained through k-fold cross validation, which is used to characterize the mapping relationship from influencing factors to accident risks.

10. The highway situation awareness and accident risk warning method integrating multi-source data according to claim 1 is characterized in that: The pre-event control system uses real-time dynamic monitoring of traffic flow data combined with an assessment model to automatically predict potential accident risk levels. At the same time, it automatically generates multi-level pre-event control plans and response measures based on different risk levels and accident probabilities, and provides targeted action guidance, including sound and light intervention, dynamic adjustment of information boards, and intelligent lane control. The sound and light intervention uses sound and light warning devices along the highway to alert drivers to potential dangers; The dynamic adjustment of the information board is that the upstream and downstream variable information boards dynamically display risk warnings and safe driving suggestions according to the real-time risk level, reminding the driver to slow down or change the route; The intelligent lane control is to control the traffic flow in the high-risk traffic range by opening and closing the emergency lane and dynamically allocating lanes, so as to achieve the regulation of the traffic flow on the road section, and can also adjust the number of lanes according to the changes in the traffic flow.

Images