A traffic accident analysis method and system based on multi-source data

Abstract

The present invention relates to the technical field of traffic accident analysis, and particularly to a traffic accident analysis method and system based on multi-source data. The method includes: determining the influencing factors of historical traffic accidents according to the correlation analysis results and forming an influencing factor set; determining the actual influencing factor set according to the principal component analysis method; determining the characterization tendency of the current traffic accident according to the actual influencing factor set to control the acquisition drone to collect traffic videos of the corresponding highway section, and determining the driving data sets of each driving vehicle to determine the driving characterization trend of any driving vehicle; locking the position of the driving vehicle with a recessive stable trend in the congested section according to the driving data, and controlling the communication drone to fly to a preset position to collect congestion videos in combination with the positional relationship of the congested section to determine the congestion cause; the present invention analyzes the accident influencing factors according to historical traffic accidents and combines real-time traffic characteristics to predict accident tendencies, thereby improving the timeliness of traffic accident occurrence.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic accident analysis, and in particular to a traffic accident analysis method and system based on multi-source data. Background Art

[0002] With the increase in highway traffic flow, the frequency of traffic accidents has also risen, posing a huge challenge to traffic safety and traffic management. Traditional traffic accident analysis methods often rely on a single data source or simple statistical analysis, and have the following limitations: relying only on limited data sources such as accident site reports and traffic police records, it is impossible to fully reflect the complex environment and various influencing factors of traffic accidents. For example, factors such as weather conditions, road facilities, and vehicle driving status may not be fully considered; it is difficult to obtain traffic information in real time and analyze potential accident risks in a timely manner. Usually, investigations and analyses are carried out after accidents occur, and effective preventive measures cannot be taken before accidents occur, resulting in the inability to avoid accidents in a timely manner or reduce the impact of accidents; the analysis methods used are relatively simple, unable to process complex multi-source data, difficult to accurately identify the degree of influence of different factors on traffic accidents, and also difficult to accurately predict and evaluate traffic conditions, thus affecting the traffic management department's formulation of scientific and effective traffic management strategies and emergency measures.

[0003] Chinese Patent Publication No. CN108417033B discloses a method for analyzing and predicting highway traffic accidents based on multi-dimensional factors, which includes the following steps: establishing a database based on historical traffic accident data and historical daily record data; selecting traffic accident types and corresponding daily record data from the database to obtain multi-dimensional influencing factor data of traffic accidents; establishing a Bayesian network for the multi-dimensional influencing factor data to obtain the influence probability of each factor on traffic accidents, and using it as a prediction model; predicting highway traffic accidents according to the prediction model and real-time data. This invention can preprocess and transform historical traffic accident data, analyze and utilize multi-dimensional influencing factors of traffic accidents to establish a corresponding Bayesian network to form a prediction model of traffic accidents, use data mining technology to find out the probability relationship between multi-dimensional factors affecting traffic accidents, and use real-time observation data based on the analysis results to predict whether an accident will occur. It can be seen that the prior art has the following problems:

[0004] The prior art cannot analyze accident influencing factors based on historical traffic accidents and cannot predict accident tendencies in combination with real-time traffic characteristics, resulting in the problem of untimely handling after traffic accidents occur. Summary of the Invention

[0005] To this end, the present invention provides a traffic accident analysis method and system based on multi-source data to overcome the problem in the prior art that accident influencing factors cannot be analyzed based on historical traffic accidents and accident tendencies cannot be predicted in combination with real-time traffic characteristics, resulting in untimely handling after traffic accidents occur.

[0006] To achieve the above object, on the one hand, the present invention provides a traffic accident analysis method based on multi-source data, including:

[0007] Obtain the historical traffic accident congestion rate of a highway section and its corresponding several traffic accident characteristics, and determine the influencing factors of the historical traffic accident according to the correlation analysis results between each traffic accident characteristic and the corresponding historical traffic accident congestion rate, and form an influencing factor set;

[0008] Determine the contribution rate and principal component characteristics of each influencing factor in the influencing factor set according to the principal component analysis method to determine the influencing factor contribution set and principal component characteristic set of the corresponding historical traffic accident;

[0009] Determine the actual influencing factors affecting traffic accidents on the highway section according to the intersection of the principal component characteristic sets of each historical traffic accident to determine the actual influencing factor set;

[0010] Obtain the specific values of each influencing factor in the influencing factor set in real time to determine the real-time contribution rate of each influencing factor, and determine the current traffic accident characterization tendency according to the real-time cumulative contribution rate of the actual influencing factor set;

[0011] Control several collecting drones to take off in sequence according to the determination result of the characterization tendency and respectively collect the traffic videos of the corresponding highway section;

[0012] Determine the driving data group of each driving vehicle according to a single traffic video to determine the driving characterization trend of any driving vehicle;

[0013] In response to the existence of a congested section, lock the positions of the driving vehicles with a hidden stable trend in the congested section according to the driving data of the driving vehicles with a hidden stable trend;

[0014] Control the communication drone to fly to a preset position to collect a congestion video according to the positional relationship between the position of the driving vehicle with a hidden stable trend and the position of the congested section to determine whether the cause of congestion is a traffic accident;

[0015] Wherein, the driving data group includes driving speed, the difference in driving speed from the vehicle in front during driving, and whether to overtake / change lanes;

[0016] The driving characterization trend includes an obvious stable trend and a hidden stable trend.

[0017] Further, the traffic accident characteristics include traffic accident environment characteristics and traffic accident types;

[0018] Among them, the traffic accident environmental characteristics include weather, visibility, time period, road section, road surface condition, lighting, whether it is a holiday, vehicle speed, and traffic flow;

[0019] The types of traffic accidents include rear-end accidents, lane-changing / overtaking accidents, single-vehicle accidents, and multi-vehicle chain accidents.

[0020] Further, determine the characteristic variables related to traffic accidents according to the Pearson correlation coefficient between the historical traffic accidents and each of the traffic accident characteristics, determine the relevant characteristic variables as the set of influencing factors of traffic accidents, and record the data corresponding to each influencing factor as the influencing factor data of traffic accidents;

[0021] Among them, if the absolute value of the Pearson correlation coefficient between a single traffic accident characteristic and a traffic accident is greater than a preset value, it is determined that the traffic accident characteristic is related to the traffic accident and is recorded as an influencing factor.

[0022] Further, determine the contribution rate of each influencing factor in the set of influencing factors according to the principal component analysis method, and determine all influencing factors with a cumulative contribution rate greater than or equal to the preset contribution rate as the principal component characteristics.

[0023] Further, obtain the specific values of each influencing factor in the set of influencing factors in real time to determine the real-time contribution rate of each influencing factor according to the principal component analysis method, and determine the characterization tendency of the current traffic accident according to the real-time cumulative contribution rate of the actual set of influencing factors. Among them,

[0024] If the real-time cumulative contribution rate is greater than or equal to the preset cumulative contribution rate, it is determined that the characterization tendency of the current traffic accident is the occurrence tendency;

[0025] If the real-time cumulative contribution rate is less than the preset cumulative contribution rate, it is determined that the characterization tendency of the current traffic accident is the non-occurrence tendency.

[0026] Further, according to the determination result that the characterization tendency is the occurrence tendency, control a number of acquisition drones to take off in sequence and respectively acquire the traffic videos of the corresponding highway sections;

[0027] Among them, the number of the acquisition drones is at least 3, and the take-off time intervals of each acquisition drone are equal and are all greater than the preset duration.

[0028] Further, the flight direction of the acquisition drone is opposite to the driving direction of the driving vehicle.

[0029] Further, determine the driving data of each driving vehicle according to a single traffic video to determine the driving characterization trend of any driving vehicle, including,

[0030] An artificial intelligence software is used to determine the driving data set of each driving vehicle according to the single traffic video;

[0031] Determine the difference degree of each driving data in the several driving data sets determined by a single driving vehicle according to each traffic video;

[0032] Determine the driving characterization trend of the corresponding driving vehicle according to the difference degree of each driving data, where,

[0033] If the difference degree of each driving data is less than or equal to the preset difference degree, it is determined that the driving characterization trend of the driving vehicle is an explicit stable trend;

[0034] If there is any driving data with a difference degree greater than the preset difference degree, it is determined that the driving characterization trend of the driving vehicle is a recessive stable trend.

[0035] Furthermore, according to the positional relationship between the position of the driving vehicle with a recessive stable trend and the position of the congested section, control the preset flight position of the communication drone and collect the congestion video at the preset position, including,

[0036] If the position of the driving vehicle with a recessive stable trend is before the target position of the congested section, control the preset flight position of the communication drone to be above the position of the driving vehicle;

[0037] If the position of the driving vehicle with a recessive stable trend is not before the target position of the congested section, control the preset flight position of the communication drone to be the target position of the congested section;

[0038] Wherein, the target position is at the quarter position of the congested section.

[0039] On the other hand, the present invention also provides a traffic accident analysis system based on multi-source data, including,

[0040] A data acquisition module, including a retrieval unit for obtaining the historical traffic accident congestion rate of the highway section and its corresponding several traffic accident characteristics, a collection unit for obtaining the specific values of each influencing factor in the influencing factor set in real time, and a networking unit for determining whether there is a congested section;

[0041] A data analysis module, which is connected to the data acquisition module, and is used to determine the influencing factors of the historical traffic accident and form an influencing factor set according to the correlation analysis result of each traffic accident characteristic and the corresponding historical traffic accident congestion rate, determine the contribution rate and principal component characteristics of each influencing factor in the influencing factor set according to the principal component analysis method to determine the influencing factor contribution set and principal component characteristic set of the corresponding historical traffic accident, and determine the actual influencing factors affecting the traffic accidents on the highway section according to the intersection of the principal component characteristic sets of each historical traffic accident to determine the actual influencing factor set;

[0042] A real-time analysis module, which is respectively connected to the data acquisition module and the data analysis module, is used to determine the real-time contribution rate of each influencing factor according to the specific values of each real-time influencing factor, determine the characterization tendency of the current traffic accident according to the real-time cumulative contribution rate of the actual influencing factor set, control several acquisition drones to take off in sequence and respectively collect traffic videos of corresponding highway sections according to the determination result of the characterization tendency, and determine the driving data group of each driving vehicle according to a single traffic video to determine the driving characterization trend of any driving vehicle;

[0043] An analysis and determination module, which is respectively connected to the real-time analysis module and the data acquisition module, is used to lock the positions of the driving vehicles with a recessive stable trend in the congested section in response to the existence of the congested section and according to the driving data of the driving vehicles with a recessive stable trend, and control the flight preset position of the communication drone to collect the congestion video according to the positional relationship between the positions of the driving vehicles with a recessive stable trend and the congested section, so as to determine whether the cause of the congestion is a traffic accident.

[0044] Compared with the prior art, the beneficial effects of the present invention are that the traffic accident analysis method based on multivariate data provided by the present invention comprehensively uses multivariate data and advanced technologies. First, the influencing factors and contribution sets are determined through correlation analysis and principal component analysis, then the accident characterization tendency is judged by combining real-time data, and then the drones are used to collect videos to obtain vehicle driving data, and then the vehicle positions are predicted and the positions of specific vehicles in the congested section are locked, and finally the congestion video is collected to analyze the cause of the accident; this method can accurately identify the influencing factors of traffic accidents, monitor the accident tendency in real time, efficiently obtain traffic information, provide a comprehensive and effective solution for the prevention and handling of traffic accidents, and improve the scientificity, accuracy and timeliness of traffic management;

[0045] Furthermore, by obtaining the specific values of the influencing factors in real time, using principal component analysis to determine the real-time contribution rate, and then comparing the real-time cumulative contribution rate of the actual influencing factor set with the preset cumulative contribution rate to judge the characterization tendency of the current traffic accident, it is possible to accurately and timely evaluate the risk of traffic accident occurrence based on historical data and real-time dynamic information, provide strong data support for traffic management decision-making, and effectively improve the foresight and scientificity of traffic management;

[0046] Further, according to the determination result of the traffic accident proneness, at least 3 collecting drones are controlled to take off in sequence at equal time intervals greater than a preset duration to collect traffic videos of the corresponding highway section; meanwhile, a plurality of parking points with at least two drones are reasonably distributed along the highway section, and the distance between the parking points is adjusted according to the accident occurrence frequency of the section; such an arrangement can utilize the monitoring of multiple drones at different time points, and use advanced models to predict and analyze the traffic conditions, so as to more accurately grasp the current traffic situation, predict and respond to traffic accidents in a timely manner, and improve the timeliness and effectiveness of traffic management;

[0047] Further, an artificial intelligence software is used to determine a vehicle driving data set from a single traffic video, calculate the difference degree of each driving data, and determine the driving representation trend of the vehicle according to the difference degree and the speed difference from the vehicle in front. This process can clearly distinguish the driving state of the vehicle, accurately identify the vehicles with potential traffic impact risks, provide a basis for refined vehicle behavior analysis for traffic management, help predict traffic congestion and accident risks in advance, and improve the accuracy and effectiveness of traffic management;

[0048] Further, according to the positional relationship between the vehicles driving with a latent stable trend and the target position of the congested section, the flight preset position of the communication drone is flexibly controlled to collect the congestion video and transmit it to the traffic management center in real time. This measure can quickly provide clear and targeted congestion site information for traffic management personnel, help them accurately grasp the traffic conditions, and thus formulate response strategies in a timely and effective manner to improve the efficiency of traffic congestion governance. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a step diagram of the traffic accident analysis method and system based on multi-source data according to the embodiment of the present invention;

[0050] Figure 2 It is a flowchart for determining the actual influencing factor set according to the embodiment of the present invention;

[0051] Figure 3 It is a flowchart for determining the driving representation trend of the driving vehicle according to the embodiment of the present invention;

[0052] Figure 4 It is a connection diagram of the traffic accident analysis system based on multi-source data according to the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] In order to make the objectives and advantages of the present invention clearer, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0054] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and do not limit the protection scope of the present invention.

[0055] It should be noted that in the description of the present invention, the terms indicating directions or positional relationships such as "upper", "lower", "left", "right", "inner", "outer", etc. are based on the directions or positional relationships shown in the drawings. This is only for convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of the present invention.

[0056] In addition, it should also be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installation", "connection", and "connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0057] Please refer to Figure 1 as shown, which is a step diagram of the traffic accident analysis method and system based on multi-source data in the embodiments of the present invention. The embodiments of the present invention provide a traffic accident analysis method based on multi-source data, including:

[0058] Step S1, obtain the historical traffic accident congestion rate of the highway section and its corresponding several traffic accident characteristics, and determine the influencing factors of the historical traffic accident according to the correlation analysis results of each traffic accident characteristic and the corresponding historical traffic accident congestion rate, and form an influencing factor set; it can be understood that according to the traffic accident characteristics that can be thought of, the correlation between each characteristic and the traffic accident can be determined, and the relevant characteristics and irrelevant characteristics can be determined. The relevant characteristics are recorded as influencing factors, and the corresponding data can be collected in real time for subsequent analysis of the occurrence tendency of traffic accidents;

[0059] Step S2, determine the contribution rate and principal component characteristics of each influencing factor in the influencing factor set according to the principal component analysis method (determine the principal component characteristics according to the contribution rate of each influencing factor), so as to determine the influencing factor contribution set and principal component characteristic set of the corresponding historical traffic accident; it can be understood that according to the principal component analysis method, the contribution rate of each influencing factor can be determined, and the contribution rate size of each influencing factor can be determined; in practice, the principal component analysis method is a prior art, so how to calculate it will not be elaborated.

[0060] Step S3: Determine the actual influencing factors affecting traffic accidents on this highway section by identifying the intersection of the principal component feature sets of each historical traffic accident to determine the set of actual influencing factors. It can be understood that there may be slight differences in the few principal component features with relatively low contribution rates in the principal component feature sets of each historical traffic accident, indicating that several principal component features have a relatively weak actual impact on each traffic accident. Therefore, the intersection of each principal component feature set is determined as the set of actual influencing factors.

[0061] Step S4: Obtain the specific values of each influencing factor in the set of influencing factors in real time (denoted as real-time influencing factor data) to determine the real-time contribution rate of each influencing factor, and determine the characterization tendency of the current traffic accident based on the real-time cumulative contribution rate of the set of actual influencing factors.

[0062] Step S5: Control several acquisition drones to take off in sequence according to the determination result of the characterization tendency and respectively collect traffic videos of the corresponding highway sections.

[0063] Step S6: Determine the driving data group of each driving vehicle based on a single traffic video to determine the driving characterization trend of any driving vehicle.

[0064] Step S7: In response to the existence of a congested section (in practice, it can be determined whether there is congestion according to any navigation software), lock the positions of the driving vehicles with a recessive stable trend in the congested section based on the driving data of the driving vehicles with a recessive stable trend. It can be understood that based on the videos taken by each drone and the position and speed of a single driving vehicle at the time of shooting, the driving vehicle position of this vehicle at a future time point can be predicted through AI, GPT, and trained machine learning models; this is prior art, so it will not be elaborated here.

[0065] Step S8: Control the communication drone to fly to a preset position to collect a congestion video based on the positional relationship between the position of the driving vehicle with a recessive stable trend and the position of the congested section to determine whether the cause of congestion is a traffic accident.

[0066] Among them, the driving data group includes driving speed, the difference in driving speed from the vehicle in front during driving, and whether there is continuous overtaking / lane changing. It can be understood that the difference in driving speed from the vehicle in front = the driving speed of the vehicle in front - the driving speed of this vehicle. It can be understood that the difference in driving speed from the vehicle in front can be positive or negative. A positive value represents that the driving speed of the vehicle in front > the driving speed of this vehicle, and a negative value represents that the driving speed of the vehicle in front < the driving speed of this vehicle. It can be understood that if the driving speed of this vehicle is too slow, it is very easy to cause the vehicle behind to change lanes and is very likely to cause congestion.

[0067] The driving characterization trend includes a dominant stable trend and a recessive stable trend.

[0068] It can be understood that: (1) By analyzing the correlation between the historical traffic accident congestion rate on the highway section and multiple traffic accident characteristics, the factors that truly affect traffic accidents can be effectively screened out to form a set of influencing factors. Then, using principal component analysis to determine the contribution rate and principal component characteristics of each influencing factor, and further finding the actual set of influencing factors can avoid the interference of irrelevant factors, making the analysis results more targeted and accurate, and helping to deeply understand the causes of traffic accidents; (2) Obtaining the specific values of each influencing factor in the set of influencing factors in real time, calculating the real-time contribution rate, and determining the current traffic accident characterization tendency according to the real-time cumulative contribution rate of the actual set of influencing factors, which enables the traffic management department to timely grasp the possibility of accidents and take preventive measures in advance to reduce the occurrence probability of traffic accidents; (3) Using a collection drone to collect traffic videos of the highway section, and determining the driving data group and driving characterization trend of the driving vehicles according to the videos. This method can quickly and comprehensively obtain traffic information, avoid the limitations of traditional manual monitoring methods, and improve the efficiency and accuracy of information collection; (4) After discovering a congested section, it is possible to predict the future position of the driving vehicle based on the driving data of the vehicle, and lock the position of the driving vehicle with a hidden stable trend in the congested section, which helps to quickly find the source vehicle that may cause congestion and provides strong support for subsequent traffic guidance and accident handling; (5) By using a communication drone to collect congestion videos at a preset position, combined with vehicle driving data and driving characterization trends, it is possible to accurately determine whether the cause of congestion is a traffic accident and deeply analyze the cause of the accident. This provides a basis for formulating targeted traffic management strategies and accident handling plans, and improves the scientificity and effectiveness of traffic management.

[0069] Specifically, in step S1, the traffic accident characteristics include traffic accident environment characteristics and traffic accident types;

[0070] Among them, the traffic accident environment characteristics include weather, visibility, time period, road section, road surface condition, lighting, whether it is a holiday, vehicle speed, and traffic flow;

[0071] The traffic accident types include rear-end accidents, lane-changing / overtaking accidents, single-vehicle accidents, and multi-vehicle chain accidents.

[0072] Please refer to Figure 2 As shown, it is a flowchart for determining the actual set of influencing factors in an embodiment of the present invention. Specifically, in step S1, it includes: step S11, determining the characteristic variables related to traffic accidents according to the Pearson correlation coefficient between the historical traffic accidents and each of the traffic accident characteristics; step S12, determining the related characteristic variables as the set of influencing factors of traffic accidents and recording the data corresponding to each influencing factor as the influencing factor data of traffic accidents;

[0073] Among them, if the absolute value of the Pearson correlation coefficient between a single traffic accident feature and the traffic accident is greater than a preset value, it is determined that the traffic accident feature is related to the traffic accident and is recorded as an influencing factor.

[0074] It can be understood that the calculation method of the Pearson correlation coefficient is a prior art, so it will not be elaborated here.

[0075] In implementation, the preset value ≥ 0.5; it can be understood that when the preset value is about 0.3, it is generally considered a low correlation, indicating that there is a certain linear relationship between variables, but the relationship is relatively weak; if the research has not very high requirements for correlation, or is just a preliminary exploratory analysis, 0.3 may be used as the judgment threshold; about 0.5 is usually regarded as a moderate correlation, at this time the linear relationship between variables is more obvious, in some studies with slightly higher requirements, 0.5 is used as the boundary, and only when it is greater than 0.5 is it determined that the traffic accident feature is an influencing factor, so that factors with stronger correlation with the accident can be screened out; 0.7 and above are considered highly correlated, indicating that there is a strong linear relationship between variables; taking 0.7 as the preset value, only when the Pearson correlation coefficient reaches this level will it be determined that the feature is significantly related to the traffic accident and is recorded as an influencing factor.

[0076] It can be understood that the larger the preset value is, the stronger the correlation between the determined influencing factor and the congestion rate is. However, the contribution rate of each influencing factor needs to be determined by the principal component analysis method later, and a secondary determination will also be carried out. Therefore, the preset value here does not need to be too large; preferably, the preset value = 0.6.

[0077] Specifically, in step S2, according to the principal component analysis method, the contribution rate of each influencing factor in the set of influencing factors is determined, and each influencing factor with a cumulative contribution rate greater than or equal to the preset contribution rate is determined as the principal component feature.

[0078] In implementation, the preset contribution rate ≥ 80%; (1) When the cumulative contribution rate is selected as 80%, it means that the extracted principal components can explain 80% of the information of the original data. This value is relatively loose and is suitable for those who hope to retain most of the data information while minimizing the number of principal components as much as possible to simplify data analysis and interpretation. In some preliminary exploratory studies, the requirement for data dimensionality reduction is not particularly strict, and the 80% cumulative contribution rate can help quickly find the main influencing factors and provide a direction for subsequent in-depth research; (2) 85% is a relatively moderate choice, achieving a good balance between retaining data information and reducing the number of principal components. In most practical applications, this value can ensure that the extracted principal components can fully reflect the main characteristics of the original data without making the number of principal components too large, facilitating data analysis and understanding. In fields such as market research data analysis and traffic flow characteristic analysis, the 85% cumulative contribution rate usually meets the research requirements and helps researchers identify key influencing factors or characteristics; (3) The 90% cumulative contribution rate is more stringent, requiring the extracted principal components to explain 90% of the information of the original data. This value is applicable to situations with high requirements for retaining data information, such as in some high-precision scientific research, engineering data analysis, or fields with strict requirements for result accuracy, such as aerospace and medical image analysis, where it is necessary to retain the information of the original data as comprehensively as possible to ensure the reliability and accuracy of the analysis results. Therefore, preferably, the preset contribution rate = 85%.

[0079] Specifically, in step S4, the specific values of each influencing factor in the influencing factor set are obtained in real time to determine the real-time contribution rate of each influencing factor according to the principal component analysis method, and the characterization tendency of the current traffic accident is determined according to the real-time cumulative contribution rate of the actual influencing factor set, where

[0080] if the real-time cumulative contribution rate is greater than or equal to the preset cumulative contribution rate, it is determined that the characterization tendency of the current traffic accident is the occurrence tendency;

[0081] if the real-time cumulative contribution rate is less than the preset cumulative contribution rate, it is determined that the characterization tendency of the current traffic accident is the non-occurrence tendency.

[0082] It can be understood that each historical traffic accident will have its own actual cumulative contribution rate (all greater than the preset contribution rate) when performing principal component analysis. The value of the preset cumulative contribution rate is determined according to the average value and standard deviation of the cumulative contribution rates of each traffic accident. The difference between the average value and the standard deviation of the cumulative contribution rates of each traffic accident ≤ the preset cumulative contribution rate ≤ the sum of the average value and the standard deviation of the cumulative contribution rates of each traffic accident. Generally, the preset cumulative contribution rate = the average value of the cumulative contribution rates of each traffic accident.

[0083] It can be understood that: (1) Obtaining the specific values of each influencing factor in the influencing factor set in real time can closely follow the dynamic changes of the traffic conditions. Compared with simply relying on historical data, this method can timely capture the real-time states of various factors affecting traffic accidents in the current traffic environment, providing the latest information for real-time assessment of traffic accident risks and making the assessment results more in line with the actual traffic scenarios; (2) Principal component analysis is a scientific and effective data dimension reduction and analysis method. It can extract key information from complex multivariate data and quantify the influence degree of each factor on traffic accidents, providing a scientific and quantitative basis for the determination of the characterization tendency of traffic accidents and avoiding the deviation of subjective judgment; (3) Determining the preset cumulative contribution rate according to the average value and standard deviation of the cumulative contribution rates of each historical traffic accident, and usually setting the preset cumulative contribution rate as the average value of the cumulative contribution rates of each traffic accident. This setting method not only refers to the overall characteristics of historical data but also takes into account the degree of dispersion of the data (reflected by the standard deviation), making the preset cumulative contribution rate reasonable and representative; The preset value set based on historical data provides a reliable reference standard for the comparison of real-time cumulative contribution rates and can more accurately determine the characterization tendency of current traffic accidents.

[0084] Specifically, in step S5, according to the determination result that the characterization tendency is the occurrence tendency, control several acquisition drones to take off in sequence and respectively acquire the traffic videos of the corresponding highway sections.

[0085] Among them, the number of the acquisition drones is at least 3, and the takeoff time intervals of each acquisition drone are equal and greater than the preset duration; In implementation, 5min ≤ preset duration ≤ 10min.

[0086] It can be understood that each drone can record the corresponding traffic driving video during flight. Multiple drones taking off at the same time interval can predict and analyze the traffic process during this period according to models such as AI, machine learning, and GPT, which can more clearly understand the current traffic conditions and can also predict and respond to traffic accidents in a timely manner.

[0087] In implementation, there are several acquisition drone parking points distributed along the corresponding highway section, and there are at least two drones at each parking point; The distance between any two drone parking points in the low-traffic-accident section is greater than 2 km, and the distance between any two drone parking points in the high-traffic-accident section is greater than 1 km.

[0088] It can be understood that: (1) Multiple collection drones take off sequentially at equal time intervals greater than a preset duration, enabling traffic video collection on the highway section at different time points. This makes it possible to obtain dynamic change information on traffic conditions over a period of time. Compared with monitoring at a single time point, it can provide a more comprehensive and in-depth understanding of the operation of traffic flow, including changes in vehicle driving speed, lane change frequency, increase or decrease in traffic flow, etc., providing a rich data basis for accurately analyzing traffic conditions; (2) Using advanced models such as AI, machine learning, and GPT to analyze traffic videos at different time points collected by multiple drones can more accurately predict traffic development trends. For example, by analyzing the driving trajectories and speed changes of vehicles over a period of time, potential traffic congestion points or accident hazards can be detected in advance, and corresponding measures can be taken in a timely manner to effectively prevent the occurrence of traffic accidents and improve the safety and efficiency of traffic operation; (3) Distributing multiple stop points with at least two drones along the highway section and adjusting the distance between stop points according to the accident occurrence frequency of the section, narrowing the distance between stop points in high-accident sections can ensure denser monitoring coverage in these areas and promptly capture the occurrence of accidents. While appropriately increasing the distance between stop points in low-accident sections can not only meet the basic monitoring needs but also rationally allocate resources, avoiding excessive waste of resources and achieving optimized utilization of resources; (4) Each stop point is equipped with at least two drones. When one drone fails or runs out of power, the other can promptly take over the task, ensuring the continuity and stability of the monitoring work. This greatly improves the reliability of traffic management, ensuring continuous acquisition of traffic information under any circumstances and providing strong support for traffic management decisions.

[0089] Specifically, in step S5, the flight direction of the collection drone is opposite to the driving direction of the driving vehicle. It can be understood that for two highway sections in opposite directions (location a → location b, location b → location a), they are two sets of the same analysis processes. This application only considers the highway section in one direction; the traffic accident analysis process in the other direction is the same, but the historical traffic accidents and their corresponding historical traffic accident congestion rates and traffic accident characteristics determined in step S1 are different.

[0090] Please refer to Figure 3 as shown, which is a flowchart for determining the driving characterization trend of a driving vehicle in an embodiment of the present invention. Specifically, in step S6, driving data of each driving vehicle is determined based on a single traffic video to determine the driving characterization trend of any driving vehicle, including,

[0091] Step S61: Use artificial intelligence software to determine the driving data sets of each moving vehicle based on a single traffic video. It can be understood that three acquisition drones can obtain three traffic videos, that is, three driving data sets of each moving vehicle can be determined. In implementation, each drone is equipped with a wide-angle camera. Therefore, during the flight of the drone and the driving of the vehicle, a continuous video segment of a vehicle can also be captured.

[0092] Step S62: Determine the difference degree of each driving data in the several driving data sets determined for a single moving vehicle according to each traffic video. It can be understood that three driving data sets mean that there are three of any driving data of each moving vehicle, so there should be two difference degrees for any driving data. It can be understood that if there is continuous overtaking / lane changing (that is, lane changing / overtaking more than 3 times within 200 m), then it is determined that the difference degree of this driving data (whether there is continuous overtaking / lane changing) is greater than the preset difference degree.

[0093] Step S63: Determine the driving characterization trend of the corresponding moving vehicle according to the difference degree of each driving data and the driving speed difference from the vehicle in front, where

[0094] If the difference degree of each driving data is less than or equal to the preset difference degree and the driving speed difference from the vehicle in front is less than or equal to the preset speed difference, then it is determined that the driving characterization trend of this moving vehicle is an explicit stable trend.

[0095] If there is a difference degree of any driving data greater than the preset difference degree and / or the driving speed difference from the vehicle in front is greater than the preset speed difference, then it is determined that the driving characterization trend of this moving vehicle is an implicit stable trend.

[0096] In implementation, the difference degree is expressed as a percentage. The difference degree of the driving speed = (the adjacent and subsequent driving speed - the adjacent and previous driving speed) ÷ the adjacent and previous driving speed × 100%. The difference degree of the driving speed difference = (the adjacent and subsequent driving speed difference - the adjacent and previous driving speed difference) ÷ the adjacent and previous driving speed difference × 100%. Generally, the preset difference degree ∈ [10%, 20%], and preferably it is set to 15%.

[0097] In implementation, if the driving speed of the vehicle itself is too slow, it is very easy to cause the vehicle behind to change lanes, and it is very easy to cause congestion. And if the difference from the vehicle in front is too large, it means that the congestion may start from this vehicle. The preset speed difference is 10 km / h to 20 km / h; preferably it is set to 15 km / h.

[0098] It is understandable that: (1) By processing the traffic videos captured by the collection drones with the aid of artificial intelligence software, the driving data sets of each moving vehicle can be determined quickly and accurately. In particular, for the continuous video segments obtained by multiple drones equipped with wide-angle cameras, it can ensure the comprehensive capture of vehicle driving data, greatly improving the data collection efficiency and providing a rich and accurate data basis for subsequent analysis. Compared with the traditional manual data collection method, it significantly saves labor and time costs; (2) By calculating the difference degrees of each driving data, the changes in the vehicle driving state are quantified, and a reasonable preset difference degree range is set, which can be used as a quantitative standard for judging whether the vehicle driving state is stable, providing a scientific basis for accurately judging the trend of driving characteristics; (3) Based on the magnitudes of the difference degrees of each driving data and the speed difference from the vehicle in front, clear rules for judging the trend of driving characteristics are formulated. When the difference degrees of each driving data are small and the speed difference from the vehicle in front is small, it is judged as an obvious stable trend, indicating that the vehicle is driving smoothly; while when there is a large difference degree of data or a large speed difference, it is judged as a hidden stable trend, meaning that there are unstable factors in the vehicle driving, which may affect traffic. This precise judgment method helps to timely discover potential traffic risk sources; (4) Anticipate traffic congestion and accident risks in advance according to the trend of the vehicle's driving characteristics. When a large number of vehicles show a hidden stable trend, especially when the driving speed of the vehicle itself is too slow or the speed difference from the vehicle in front is too large, it can be inferred that it may cause frequent lane changes of the following vehicles, thus leading to traffic congestion or even accidents; the traffic management department can take measures in advance based on this.

[0099] Specifically, in step S8, according to the positional relationship between the position of the moving vehicle with a hidden stable trend and the position of the congested section, control the flight preset position of the communication drone and collect congested videos at the preset position, including,

[0100] If the position of the moving vehicle with a hidden stable trend is before the target position of the congested section (i.e., between the start position of the congestion and the target position), then control the flight preset position of the communication drone to be above the position of the moving vehicle;

[0101] If the position of the moving vehicle with a hidden stable trend is not before the target position of the congested section (i.e., after the target position), then control the flight preset position of the communication drone to be the target position of the congested section;

[0102] Among them, the target position is located at the quarter position of the congested section; that is, record the distance between the start position of the congested section and the target position as the first distance, and record the distance between the end position of the congested section and the target position as the second distance, and the first distance: the second distance = 1:3.

[0103] It is understandable that the communication drones collect congestion videos on-site and transmit them in real time to the traffic management center, which is connected to any traffic management personnel, enabling the traffic management personnel to quickly understand the current traffic problems and take countermeasures.

[0104] In implementation, each drone parking point also includes at least two communication drones, and the communication drones for takeoff are determined according to the principle of proximity.

[0105] It is understandable that: (1) By setting a quarter of the congested section as the target position and determining the preset flight position of the communication drone based on the front-back relationship between the vehicles traveling in the implicit stable trend and the target position: If the vehicle is before the target position, the drone is positioned above the vehicle to focus on the key vehicles that may cause congestion and capture their driving states and the impact on the surrounding traffic flow; If the vehicle is after the target position, the drone is positioned at the target position of the congested section to comprehensively capture the core area conditions of the congested section, ensuring the collection of the most valuable congestion information and improving the pertinence and accuracy of video collection; (2) The communication drone collects congestion videos in real time and transmits them to the traffic management center, which is directly connected to the traffic management personnel, greatly shortening the information acquisition time, enabling the traffic management personnel to quickly understand the current traffic problems without having to be on-site, breaking through the time and space limitations, providing strong support for timely decision-making, and being able to respond to traffic congestion faster than traditional manual reporting or information collection methods with high latency; (3) After obtaining real-time and accurate congestion videos, the traffic management personnel can intuitively and clearly observe the actual situation of the congested section, such as details of vehicle density, driving direction, road traffic conditions, etc.; Based on these detailed information, they can quickly and accurately judge the causes of congestion, and then formulate more targeted countermeasures, such as arranging police to direct traffic, adjusting signal timings, issuing traffic guidance information, etc., effectively improving the efficiency and effect of traffic congestion governance and ensuring smooth road traffic.

[0106] Please refer to Figure 4 as shown, which is the connection diagram of the traffic accident analysis system based on multi-source data in the embodiment of the present invention. On the other hand, the present invention also provides a traffic accident analysis system based on multi-source data, including,

[0107] A data collection module, including a retrieval unit for obtaining the historical traffic accident congestion rate of the highway section and its corresponding several traffic accident characteristics, a collection unit for obtaining the specific values of each influencing factor in the influencing factor set in real time, and a networking unit for determining whether there is a congested section;

[0108] A data analysis module, which is connected to the data acquisition module, is used to determine the influencing factors of the historical traffic accident and form a set of influencing factors according to the correlation analysis results between each of the traffic accident characteristics and the corresponding historical traffic accident congestion rate, determine the contribution rate and principal component characteristics of each influencing factor in the set of influencing factors according to the principal component analysis method to determine the contribution set of influencing factors and the principal component characteristic set of the corresponding historical traffic accident, and determine the actual influencing factors affecting the traffic accidents on the highway section according to the intersection of the principal component characteristic sets of each historical traffic accident to determine the set of actual influencing factors;

[0109] A real-time analysis module, which is respectively connected to the data acquisition module and the data analysis module, is used to determine the real-time contribution rate of each influencing factor according to the specific values of each real-time influencing factor, determine the characterization tendency of the current traffic accident according to the real-time cumulative contribution rate of the set of actual influencing factors, control a number of acquisition drones to take off in sequence and respectively collect the traffic videos of the corresponding highway section according to the determination result of the characterization tendency, and determine the driving data group of each driving vehicle according to a single traffic video to determine the driving characterization trend of any driving vehicle;

[0110] An analysis and determination module, which is respectively connected to the real-time analysis module and the data acquisition module, is used to respond to the existence of a congested section and lock the positions of the driving vehicles with a recessive stable trend in the congested section according to the driving data of the driving vehicles with a recessive stable trend, and control the flight preset positions of communication drones to collect congestion videos according to the positional relationship between the positions of the driving vehicles with a recessive stable trend and the congested section to determine whether the cause of congestion is a traffic accident.

[0111] So far, the technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0112] The above are only the preferred embodiments of the present invention and are not used to limit the present invention; for those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A traffic accident analysis method based on multivariate data, characterized in that: include: Obtaining the historical traffic accident congestion rate of the expressway section and its corresponding traffic accident characteristics, and determining the influencing factors of the historical traffic accident according to the correlation analysis results of each of the traffic accident characteristics and the corresponding historical traffic accident congestion rate, and forming an influencing factor set; Determine the contribution rate and principal component characteristics of each influencing factor in the influencing factor set according to the principal component analysis method, so as to determine the influencing factor contribution set and principal component characteristic set corresponding to the historical traffic accidents; The actual influencing factors affecting the traffic accidents on the expressway section are determined according to the intersection of the principal component feature sets of each historical traffic accident to determine the actual influencing factor set; Obtain the specific value of each influencing factor in the influencing factor set in real time to determine the real-time contribution rate of each influencing factor, and determine the characterization tendency of the current traffic accident according to the real-time cumulative contribution rate of the actual influencing factor set; According to the determination result of the characterization tendency, a plurality of collecting drones are controlled to take off in sequence and collect traffic videos of corresponding highway sections respectively; Determine a driving data group of each driving vehicle based on a single traffic video to determine a driving characteristic trend of any driving vehicle; In response to the existence of a congested road section, the position of the vehicle with an implicit stable trend in the congested road section is locked according to the driving data of the vehicle with an implicit stable trend; According to the position relationship between the driving vehicle position of the implicit stable trend and the congested road section, the communication drone is controlled to fly to a preset position to collect congestion video to determine whether the cause of the congestion is a traffic accident; The driving data group includes the driving speed, the driving speed difference with the preceding vehicle during driving, and whether the vehicle is overtaking / changing lanes; The driving characterization trend includes an explicit stable trend and an implicit stable trend.

2. The traffic accident analysis method based on multivariate data according to claim 1, characterized in that: The traffic accident characteristics include traffic accident environment characteristics and traffic accident types; The traffic accident environment characteristics include weather, visibility, time period, road section, road condition, lighting, whether it is a holiday, vehicle speed and traffic volume; The types of traffic accidents include rear-end collisions, lane change / overtaking accidents, single-vehicle accidents and multi-vehicle chain accidents.

3. The traffic accident analysis method based on multivariate data according to claim 1, characterized in that: Determine characteristic variables related to the traffic accident according to the Pearson correlation coefficient between the historical traffic accident and each of the traffic accident characteristics, determine the relevant characteristic variables as a set of influencing factors of the traffic accident and record the data corresponding to each influencing factor as influencing factor data of the traffic accident; Among them, if the absolute value of the Pearson correlation coefficient between a single traffic accident feature and the traffic accident is greater than a preset value, it is determined that the traffic accident feature is related to the traffic accident and is recorded as an influencing factor.

4. The traffic accident analysis method based on multivariate data according to claim 1, characterized in that: The contribution rate of each influencing factor in the influencing factor set is determined according to the principal component analysis method, and each influencing factor whose cumulative contribution rate is greater than or equal to the preset contribution rate is determined as the principal component feature.

5. The traffic accident analysis method based on multivariate data according to claim 1, characterized in that: The specific value of each influencing factor in the influencing factor set is obtained in real time to determine the real-time contribution rate of each influencing factor according to the principal component analysis method, and the characterization tendency of the current traffic accident is determined according to the real-time cumulative contribution rate of the actual influencing factor set, where: If the real-time cumulative contribution rate is greater than or equal to the preset cumulative contribution rate, it is determined that the current traffic accident representation tendency is an occurrence tendency; If the real-time cumulative contribution rate is less than the preset cumulative contribution rate, it is determined that the current traffic accident characterization tendency is a non-occurrence tendency.

6. The traffic accident analysis method based on multivariate data according to claim 1, characterized in that: According to the determination result that the characteristic tendency is an occurrence tendency, a plurality of collecting drones are controlled to take off in sequence and collect traffic videos of corresponding highway sections respectively; The number of the collection drones is at least 3, and the take-off time intervals of the collection drones are equal and greater than a preset time length.

7. The traffic accident analysis method based on multivariate data according to claim 1, characterized in that: The flying direction of the collection drone is opposite to the driving direction of the moving vehicle.

8. The traffic accident analysis method based on multivariate data according to claim 1, characterized in that: Determining the driving data of each driving vehicle based on the single traffic video to determine the driving characteristic trend of any driving vehicle, including: Using artificial intelligence software to determine the driving data group of each driving vehicle based on the single traffic video; Determine the difference between each driving data in a plurality of driving data groups determined by a single driving vehicle according to each traffic video; The driving characterization trend of the corresponding vehicle is determined according to the difference between each driving data, where: If the differences of the driving data are all less than or equal to the preset differences, it is determined that the driving characteristic trend of the driving vehicle is a dominant stable trend; If there is any driving data whose difference is greater than a preset difference, it is determined that the driving characteristic trend of the driving vehicle is a hidden stable trend.

9. The traffic accident analysis method based on multivariate data according to claim 1, characterized in that: According to the position relationship between the driving vehicle position of the implicit stable trend and the congested road section, the communication drone is controlled to fly to a preset position and collect congestion video at the preset position, including: If the position of the moving vehicle with an implicit stable trend is before the target position of the congested road section, the preset flight position of the control communication drone is above the position of the moving vehicle; If the position of the driving vehicle with an implicit stable trend is not before the target position of the congested road section, the preset flight position of the control communication drone is the target position of the congested road section; The target location is located at one quarter of the congested road section.

10. A traffic accident analysis system based on multivariate data using the traffic accident analysis method based on multivariate data according to any one of claims 1 to 9, characterized in that: include, The data collection module includes a retrieval unit for obtaining the historical traffic accident congestion rate of the expressway section and its corresponding traffic accident characteristics, a collection unit for obtaining the specific values ​​of each influencing factor in the influencing factor set in real time, and a networking unit for determining whether there is a congested section; A data analysis module, which is connected to the data acquisition module, is used to determine the influencing factors of the historical traffic accident and form an influencing factor set according to the correlation analysis results of each of the traffic accident characteristics and the corresponding historical traffic accident congestion rate, determine the contribution rate and principal component characteristics of each influencing factor in the influencing factor set according to the principal component analysis method to determine the influencing factor contribution set and principal component characteristic set of the corresponding historical traffic accident, and determine the actual influencing factors that affect the traffic accident on the expressway section according to the intersection of the principal component characteristic sets of each historical traffic accident to determine the actual influencing factor set; A real-time analysis module, which is connected to the data acquisition module and the data analysis module respectively, and is used to determine the real-time contribution rate of each influencing factor according to the specific value of each real-time influencing factor, determine the characterization tendency of the current traffic accident according to the real-time cumulative contribution rate of the actual influencing factor set, control a number of collection drones to take off in sequence and collect traffic videos of corresponding highway sections respectively according to the determination result of the characterization tendency, and determine the driving data group of each driving vehicle according to a single traffic video to determine the driving characterization trend of any driving vehicle; The analysis and determination module is respectively connected to the real-time analysis module and the data acquisition module, and is used to respond to the existence of a congested road section and lock the position of the vehicle with an implicit stable trend in the congested road section according to the driving data of the vehicle with an implicit stable trend, and control the communication drone to fly at a preset position to collect congestion video according to the positional relationship between the position of the vehicle with the implicit stable trend and the congested road section, so as to determine whether the cause of the congestion is a traffic accident.

Images