Vehicle real-time accident detection method and system based on Beidou navigation technology

Abstract

The invention relates to a vehicle real-time accident detection method and system based on a Beidou navigation technology. The detection method comprises the following steps: S1, constructing a road section selection model to select a to-be-detected road section; s2, extracting navigation information of each vehicle in the road section to be detected through a Beidou navigation technology; s3, obtaining correlation degree information between the vehicles according to the space information and the time information; s4, establishing an accident detection model combining a cuckoo search algorithm and a random forest algorithm in combination with correlation information; and S5, inputting the correlation degree information into a trained accident detection model, and judging whether an accident occurs in the road section to be detected or not. Whether the vehicle has an accident or not is judged through the correlation degree information between the vehicles, learning optimization is performed on the detection result by establishing the accident detection model combining the cuckoo search algorithm and the random forest algorithm, and the accuracy and efficiency of accident detection are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle accident detection, and in particular to a vehicle real-time accident detection method based on Beidou navigation technology and a vehicle real-time accident detection system based on Beidou navigation technology. Background Art

[0002] Modern transportation refers to a transportation process that is convenient, safe, efficient and economical, based on modern technology and coordinated with multiple modes of transportation. The modern transportation system not only includes traditional road, rail, water, aviation and pipeline transportation, but also covers the application of emerging technologies such as intelligent transportation systems (ITS). Technological innovation is an important driving force for promoting transportation progress. In recent years, China has achieved remarkable results in scientific and technological innovation in the field of transportation, and key core technologies such as the Beidou navigation system, industrial Internet, 5G and blockchain have been widely used. The application of these technologies has not only improved the efficiency and safety of transportation, but also provided solid technical support for the construction of a strong transportation country. The development of intelligent transportation is an important direction for future transportation progress. Intelligent transportation systems have effectively solved problems such as traffic congestion, frequent accidents and air pollution through technologies such as artificial intelligence, edge computing and big data.

[0003] However, with the increasing number of vehicles on the road, it is difficult to avoid vehicle accidents. Human factors are one of the main causes of traffic accidents: driver's operating errors, non-compliance with traffic rules, safety uncertainty, distracted driving, and lack of concentration are all common reasons. In addition, the driver's psychological state will also affect driving behavior. For example, bad psychological states such as emotional excitement and confusion can also lead to traffic accidents. Other factors such as road factors and vehicle factors can also cause certain traffic accidents. Timely detection and response to traffic accidents can significantly reduce the number of casualties and property losses. Traditional accident reporting methods often rely on reports from witnesses or drivers, and the response time is long. Therefore, real-time vehicle accident detection is needed to quickly respond to vehicle accidents. Real-time vehicle accident detection technology is an important part of the intelligent transportation system (ITS), which aims to detect and respond to traffic accidents in a timely manner through various technical means to improve road safety and reduce losses caused by accidents. Although the existing accident detection methods have solved certain accident detection problems, they require the use of camera equipment, resulting in high detection costs, high computing costs, and low penetration rates. Summary of the invention

[0004] Based on this, it is necessary to propose a vehicle real-time accident detection method and system based on Beidou navigation technology to address the problems of high detection cost, high computing cost and low penetration rate caused by the use of camera equipment.

[0005] The present invention is implemented by the following technical scheme: A vehicle real-time accident detection method based on Beidou navigation technology comprises the following steps:

[0006] S1: Construct a road section selection model and select a road section to be tested from all road sections;

[0007] S2: Extract navigation information of each vehicle in the road section to be tested through Beidou navigation technology. The navigation information includes location information, time information and status information;

[0008] S3: Obtaining correlation information between each vehicle according to the navigation information; the correlation information includes: position correlation, time correlation and state correlation;

[0009] S4: Combine the location correlation, time correlation and state correlation to establish an accident detection model combining the cuckoo search algorithm and the random forest algorithm;

[0010] S40: Construct an accident detection model based on random forest algorithm;

[0011] S41: Optimizing hyperparameters of accident detection model using cuckoo search algorithm;

[0012] S42: training the accident detection model according to the position correlation, time correlation and state correlation;

[0013] S5: Input the correlation information into the trained accident detection model to determine whether an accident occurs in the road section to be tested.

[0014] The above detection method uses a randomized selection algorithm to construct a road section selection model, making the road section retrieval faster and more efficient; it determines whether a vehicle has an accident through the correlation information between vehicles, and enhances the robustness of the model by establishing an accident detection model that combines the cuckoo search algorithm with the random forest algorithm. The detection results are learned and optimized through the model, thereby improving the accuracy and efficiency of accident detection.

[0015] Furthermore, the construction method of the road section selection model is as follows:

[0016] S10: number all road sections and sort them by number size to obtain an array S containing all sorts;

[0017] S11: Initialize the road segment selection model based on the random selection algorithm; define the random selection function, receive and input the road segment array S and the pre-detection value k;

[0018] S12: Based on the amount of data in array S, a pivot p is selected using a random number generator;

[0019] S13: Divide the segment array S into two parts according to the pivot p, namely Sleft and S right ; S left ≤ p ≤ S right ;

[0020] S14: Recursively call the data in the section array S, and determine the direction of recursion according to the value of the pre-detection value k;

[0021] S15: Take the section corresponding to the k value obtained as the section to be measured.

[0022] Furthermore, in step S14, the specific algorithm for the recursion direction is as follows:

[0023]

[0024] In the formula, S result is the final output value, S p is the value corresponding to the pivot p in the section array S, is the value in the section array S that is the same as the pre-detection value k when p > k; is the value in the section array S that is the same as the pre-detection value k when p < k; the pre-detection value k takes any value k random in the section array S, that is, k = k random ∈ S;

[0025] Furthermore, in step S3, the position correlation degree is the correlation of the position changes between two vehicles to determine whether there is a connection between the position changes of the two vehicles. The specific method is as follows:

[0026]

[0027] In the formula, D is the straight-line distance between two vehicles, R is the defined position correlation degree index, ΔV is the speed difference between two vehicles, angle is the angle between two vehicles, and β and m are weight parameters used to adjust the influence of different factors on the correlation degree.

[0028] Furthermore, in step S3, the time correlation degree is the consistency of the停留 time between a vehicle and its adjacent vehicle. The formula is as follows:

[0029]

[0030] In the formula, Δt is the time difference between two vehicles, ￡ is the vehicle time correlation degree index, and the smaller the ￡ value, the higher the similarity.

[0031] Furthermore, in step S3, the state correlation degree is the degree of correlation of the state changes between two vehicles. The calculation method is as follows:

[0032]

[0033] In the formula, I(C; D) represents the mutual information between random variables C and D; e and g represent different states of C and D, and S(C e ,D g ) Event C is in state C e And event D is in state D g The joint probability of S(C e ) is event C in state C e The marginal probability of S(D g ) is event D in state D g The marginal probability of .

[0034] Furthermore, in step S41, the specific steps of optimizing the hyperparameters of the accident detection model by the cuckoo search algorithm are as follows:

[0035] S410: Initialize the group: randomly generate a group of candidate solutions;

[0036] S411: Calculate the fitness of each solution according to the objective function;

[0037] S412: Generate new candidate solutions;

[0038] S413: Select a new solution to replace the old solution according to the fitness;

[0039] S414: Eliminate some poor solutions based on probability;

[0040] S415: Repeat steps S411 to S414 until the maximum number of iterations is reached;

[0041] S416: Select the solution with the best fitness from among all candidate solutions as the output result;

[0042] S417: Taking the optimal fitness solution as a hyperparameter of the accident detection model, the accident detection model is optimized.

[0043] Furthermore, in step S42, the accident detection model is trained as follows:

[0044] S420: Inputting the correlation information into the random forest algorithm model;

[0045] S421: Output the final prediction result of the random forest, which is the average of the prediction results of all features;

[0046] S422: Calculate the accuracy rate according to the predicted structure of each feature;

[0047] S423: Calculate the precision and recall rate according to the predicted number of samples;

[0048] S424: Calculate the F1 score to achieve a balance between precision and recall.

[0049] The present invention also provides a vehicle real-time accident detection system based on Beidou navigation technology, the detection system comprising: a Beidou vehicle-mounted terminal, an information preprocessing module, a calculation module, an accident detection module, an information sending module and a monitoring server.

[0050] The Beidou vehicle-mounted terminal is used to send and receive Beidou navigation and positioning information; the information preprocessing module is used to process the navigation information of vehicles on the road section to be tested; the calculation module is used to calculate the information correlation between vehicles in the road section to be tested; the accident detection module is used to improve the accuracy and efficiency of accident detection; the information sending module is used to send abnormal data information to the server; the monitoring server is used to receive abnormal data information and issue warnings to monitoring personnel.

[0051] Furthermore, the information preprocessing module includes: a positioning information storage module and an information retrieval module. The positioning information storage module is used to store all the road section positioning information and vehicle positioning information received by the Beidou vehicle-mounted terminal; and the randomization processing module is used to quickly retrieve the road section to be tested.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] The present invention uses a road section selection model based on a randomized selection algorithm to retrieve the road section to be tested, which can improve the retrieval efficiency of the road section information to be tested and improve the speed of accident detection; by establishing an accident detection model combining a cuckoo search algorithm with a random forest algorithm, the performance of the random forest is improved, overfitting is reduced, the generalization ability of the model is improved, and the robustness of the model is enhanced. The combination of the two makes the model more adaptable to different types of features, can handle a variety of complex problems, reduces computing costs, and thus saves time and resources. The detection results are learned and optimized through the model, thereby improving the accuracy and efficiency of accident detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a step diagram of a method for real-time vehicle accident detection based on Beidou navigation technology in Example 1 of the present invention;

[0055] Figure 2 This is a structural diagram of a vehicle real-time accident detection system based on Beidou navigation technology in Example 2 of the present invention. DETAILED DESCRIPTION

[0056] 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.

[0057] It should be noted that when a component is referred to as being "mounted on" another component, it may be directly on the other component or there may be a central component. When a component is considered to be "set on" another component, it may be directly set on the other component or there may be a central component at the same time. When a component is considered to be "fixed to" another component, it may be directly fixed on the other component or there may be a central component at the same time.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "or / and" used herein includes any and all combinations of one or more of the related listed items.

[0059] Example 1: Please refer to Figure 1 This embodiment provides a method for real-time vehicle accident detection based on Beidou navigation technology, comprising the following steps:

[0060] S1: Construct a road section selection model to select a random road section to be tested from all road sections. Using this method to select the road section to be tested optimizes the efficiency of road section selection, facilitates the subsequent retrieval of information on each road section, and improves the ability to analyze and handle vehicle accidents.

[0061] The road section selection model is constructed based on a random selection algorithm. The specific construction method is as follows:

[0062] S10: All road sections are numbered and sorted by number size to obtain an array S containing all sorts. Array S contains all road sections that need to be detected, which are integrated by array method to facilitate the screening of individual road sections later. This reduces working time and improves work efficiency.

[0063] S11: Construct a road segment selection model based on a random selection algorithm. Define a random selection function, receive and input a road segment array S and a pre-detection value k.

[0064] S12: Considering the amount of data in array S, a pivot p is selected using a random number generator.

[0065] S13: Divide the road segment array S into two parts according to the pivot p, namely S left and S right ; S left ≤ p ≤ S right .

[0066] S14: Recursively call the data in the road segment array S and determine the recursive direction according to the value of the pre-detection value k. The specific algorithm is as follows:

[0067]

[0068] In the formula, S result is the final output value, S p is the value corresponding to the pivot p in the road segment array S, is the value in the road segment array S that is the same as the pre-detection value k when p > k; is the value in the road segment array S that is the same as the pre-detection value k when p < k; the pre-detection value k takes any value k random in the road segment array S, that is, k = k random ∈ S.

[0069] S15: Take the road segment corresponding to the k value obtained as the road segment to be measured.

[0070] S2: Extract the navigation information of each vehicle in the road segment to be measured through Beidou navigation technology. The navigation information includes position information, time information, and status information. The Beidou navigation technology preferably uses the combination of PPP (Precise Point Positioning) positioning technology and RTK (Real-Time Kinematic) technology to achieve sub-meter or even centimeter-level positioning. The vehicle position information is the coordinate information of the position where the vehicle is located.

[0071] S3: Obtain the correlation information between each vehicle according to the spatial information and time information. The correlation information includes: position correlation, time correlation, and status correlation. Among them, the position correlation is the correlation of the position changes between two vehicles and is used as one of the judgment conditions when determining whether an accident has occurred. If the position correlation is strong, an accident may occur. The time correlation is the consistency of the停留 time between a vehicle and its adjacent vehicle and is a further condition for judging whether an accident has occurred. In the case of strong position correlation, a large time correlation will further increase the possibility of judging it as an accident. The status correlation is the degree of correlation of the status changes between two vehicles, that is, the degree of influence of the speed change, acceleration change, and direction change of the two vehicles on each other. The greater the degree of influence, the greater the possibility of judging it as an accident.

[0072] In step S3, the position correlation is the correlation of the position changes between two vehicles to determine whether there is a connection between the position changes of the two vehicles. The calculation method is as follows:

[0073] Calculate the straight-line distance between two vehicles using the following formula:

[0074]

[0075] In the formula, d is the position correlation, the result is the Euclidean distance between the two vehicles, (a1, b1) and (a2, b2) are the coordinates between the two vehicles. These coordinates are calculated based on the vehicle position coordinates obtained by Beidou navigation technology.

[0076] Calculate the angle between two vehicles. The specific formula is as follows:

[0077]

[0078] Where V A and V B are the direction vectors of the two vehicles, |V A | and |V B | is the magnitude of the vector.

[0079] The speed difference between the two vehicles is calculated to evaluate the correlation between them. The specific formula is as follows:

[0080] ΔV=|s1-s2|

[0081] Where ΔV is the speed difference between the two vehicles.

[0082] Combining the straight-line distance, angle and speed difference of the vehicles, a comprehensive correlation index is constructed. The specific method is as follows:

[0083]

[0084] In the formula, R is the defined position correlation index, and β and m are weight parameters used to adjust the impact of different factors on the correlation.

[0085] In step S3, the time correlation is the consistency of the stay time between the vehicle and the adjacent vehicles, which is specifically as follows:

[0086] Calculate the correlation between the speeds of two vehicles within the time window, the formula is as follows:

[0087]

[0088] Where n is the number of samples, s1 and s2 are the speeds of the two vehicles respectively.

[0089] Calculate the time difference between two vehicles using the following formula:

[0090] Δt=|t1-t2|

[0091] Where t1 and t2 are the event times of the two vehicles respectively.

[0092] Combining the speed correlation and time difference, the similarity of vehicle time is scored, and the formula is as follows:

[0093]

[0094] In the formula, ￡ is the vehicle time correlation index. The smaller the ￡ value, the higher the similarity.

[0095] In step S3, the state correlation is the correlation of the state changes between the two vehicles, and the calculation method is as follows:

[0096] Construct a state transfer matrix to record the state changes of the two vehicles in different time windows as follows:

[0097]

[0098] In the formula, a os It indicates the number of times the two vehicles are in the o and s states respectively within the time window.

[0099] Calculate the joint probability of two vehicles being in a certain state in the same time window as follows:

[0100]

[0101] Where S(C,D) is the joint probability, W is the number of times vehicle state o and state s occur, and K is the total number of time windows.

[0102] Calculate the marginal probability of the vehicle state, the formula is as follows:

[0103]

[0104] Where S(C) is the probability of state C occurring, S(D) is the probability of state D occurring, and W C is the number of times state C occurs, W D is the number of times state D occurs.

[0105] Set the conditional probability to describe the probability that when one car is in a certain state, another car is in another state. The calculation method is as follows:

[0106]

[0107] Where S(D|C) is the probability of event D occurring under the condition that event C occurs; S(C,D) is the joint probability of event C and event D occurring simultaneously; S(C) is the marginal probability of event C occurring.

[0108] Calculate the position correlation between the two vehicles. The calculation method is as follows:

[0109]

[0110] In the formula, I(C; D) represents the mutual information between random variables C and D, which measures the degree of dependence between C and D. The larger the value, the stronger the correlation between C and D. e and g represent the different states S(C e ,D g ) Event C is in state C e And event D is in state D g The joint probability of C and D is: e ) is event C in state C e The marginal probability of event C in C e The probability of the state; S(D g ) Event D is in state D g The marginal probability of D is that in all possible cases, event D is in state D g probability.

[0111] S4: Combine the location correlation, time correlation and state correlation to establish an accident detection model combining the cuckoo search algorithm and the random forest algorithm.

[0112] S40: Construct an accident detection model based on random forest algorithm.

[0113] S41: Use the cuckoo search algorithm to optimize the hyperparameters of the accident detection model. The specific steps are as follows:

[0114] S410: Initialize group: randomly generate a set of candidate solutions X i , expressed as:

[0115] X i =[x1,x2,...,x i ](i=1,2,...,n)

[0116] Where n is the number of candidate solutions.

[0117] S411: Calculate the fitness of each solution according to the objective function: The fitness function formula is as follows:

[0118]

[0119] In the formula, is the objective function.

[0120] S412: Generate a new candidate solution in the following manner:

[0121] x' i =x i +α·Levy(λ)

[0122] Where α is the step size scaling factor and Levy distribution is used to generate random step sizes.

[0123] S413: Select a new solution to replace the old solution according to the fitness, in the following way:

[0124]

[0125] x' i =x i

[0126] S414: According to probability Eliminate some of the poorer solutions.

[0127] S415: Repeat steps S411 to S414 until the maximum number of iterations is reached.

[0128] S416: Select the solution with the best fitness from all candidate solutions as the output result.

[0129] S42: Construct a data set based on location correlation, time correlation and state correlation, and divide it into training set, validation set and test set to train the accident detection model.

[0130] The training method is as follows:

[0131] S420: Input the correlation information into the random forest algorithm model to obtain each tree T j The prediction formula is:

[0132]

[0133] In the formula, Y is the input feature and f() is the prediction function of the decision tree.

[0134] S421: Output the final prediction result of the random forest, which is the average of the prediction results of all features, in the following way:

[0135]

[0136] In the formula, is the prediction result of the jth feature.

[0137] S422: Calculate the accuracy according to the predicted structure of each feature in the following way:

[0138]

[0139] Where TP is the number of samples that the model correctly predicts as accidents. TN is the number of samples that the model correctly predicts as no accidents. FP is the number of samples that the model incorrectly predicts as accidents, but in fact they did not happen. FN is the number of samples that the model incorrectly predicts as no accidents, but in fact they did happen.

[0140] S423: Calculate the precision and recall according to the predicted number of samples in the following way:

[0141]

[0142]

[0143] In the formula, precision is the proportion of all samples predicted to be accidents that are actually accidents, and recall is the proportion of all samples that are actually accidents that are correctly predicted to be accidents.

[0144] S424: Calculate the F1 score in the following manner:

[0145]

[0146] Used to strike a balance between precision and recall.

[0147] Using the optimal fitness solution as the hyperparameter of the accident detection model to optimize the accident prediction model can significantly improve the performance of the accident detection model, reduce overfitting, improve the generalization ability of the model, and enhance the robustness of the model.

[0148] S5: Input the correlation information into the trained accident detection model to determine whether an accident has occurred in the road section to be tested. The accident information is sorted and sent to the monitoring server, and the operator is warned.

[0149] In summary, using a road section selection model based on a randomized selection algorithm to retrieve the road sections to be tested can improve the retrieval efficiency of the road section information to be tested and increase the speed of accident detection; by establishing an accident detection model that combines the cuckoo search algorithm with the random forest algorithm, the performance of the random forest is improved, overfitting is reduced, the generalization ability of the model is improved, and the robustness of the model is enhanced. The combination of the two makes the model more adaptable to different types of features, able to handle a variety of complex problems, and reduces computing costs, thereby saving time and resources. The detection results are learned and optimized through the model, which improves the accuracy and efficiency of accident detection.

[0150] Example 2: Please refer to Figure 2This embodiment provides a vehicle real-time accident detection system based on Beidou navigation technology, which can be controlled by the vehicle real-time accident detection method based on Beidou navigation technology in Example 1, including a Beidou vehicle-mounted terminal, an information preprocessing module, a calculation module, an accident detection module, an information sending module and a monitoring server.

[0151] Beidou vehicle-mounted terminal, which is used to send and receive Beidou navigation positioning information. Installed in each vehicle, it is used to send the anchor point of Beidou navigation and record the location information, time information and status information obtained by Beidou navigation technology.

[0152] The information preprocessing module is used to process the navigation information of the vehicles on the road section to be tested. The navigation information recorded by the Beidou vehicle terminal is sorted and summarized, and the data required by the accident detection model is preprocessed. The information preprocessing module includes: a positioning information storage module and an information retrieval module. The positioning information storage module is used to store all road section positioning information and vehicle positioning information received by the Beidou vehicle terminal. The randomization processing module is used to quickly retrieve the road section to be tested.

[0153] The calculation module is used to calculate the information correlation between vehicles in the road section to be tested. Combined with the navigation information received by the Beidou vehicle terminal, the position correlation, time correlation and state correlation are calculated.

[0154] The accident detection module is used to establish an accident detection model that combines the cuckoo search algorithm with the random forest algorithm, and outputs the detection result of whether an accident has occurred, combined with the information correlation obtained in the calculation module. The cuckoo search algorithm is used to optimize the accident detection model, which improves the performance of the random forest. The detection results are learned and optimized through the model, which improves the accuracy and efficiency of accident detection.

[0155] The information sending module is used to send abnormal data information to the server and send feedback information of the detected accident results to the server through mobile communication technology.

[0156] The monitoring server is used to receive abnormal data information and issue warnings to monitoring personnel.

[0157] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0158] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A vehicle real-time accident detection method based on Beidou navigation technology, characterized in that: The steps include: S1: Construct a road section selection model and select a road section to be tested from all road sections; S2: extracting navigation information of each vehicle in the road section to be tested by Beidou navigation technology, wherein the navigation information includes position information, time information and status information; S3: Acquire the correlation information between each vehicle according to the navigation information; The correlation information includes: position correlation, time correlation and state correlation; S4: combining the position correlation, the time correlation and the state correlation, and establishing an accident detection model combining a cuckoo search algorithm and a random forest algorithm; S40: Construct an accident detection model based on random forest algorithm; S41: Optimizing hyperparameters of accident detection model using cuckoo search algorithm; S42: training an accident detection model according to the position correlation, the time correlation and the state correlation; S5: Input the correlation information into the trained accident detection model to determine whether an accident occurs in the road section to be tested.

2. The method for real-time vehicle accident detection based on Beidou navigation technology according to claim 1, characterized in that: In step S1, the road section selection model is constructed based on a random selection algorithm, and the construction method is as follows: S10: number all road sections and sort them by number size to obtain an array S containing all sorts; S11: Initialize the road segment selection model based on the random selection algorithm; define the random selection function, receive and input the road segment array S and the pre-detection value k; S12: selecting a pivot p using a random number generator based on the amount of data in the road segment array; S13: Divide the segment array S into two parts according to the pivot p, namely S left and S right ; The S left ≤p≤S right ; S14: recursively calling the data in the road segment array S, and determining the recursive direction according to the value of the pre-detection value k; S15: The obtained road section corresponding to the k value is used as the road section to be tested.

3. The method for real-time vehicle accident detection based on Beidou navigation technology according to claim 2 is characterized in that: In step S14, the specific algorithm of the recursive direction is as follows: Wherein, S result is the final output value, and S p is the value corresponding to the pivot p in the section array S, is the value in the section array S that is the same as the pre-detection value k when p > k; is the value in the section array S that is the same as the pre-detection value k when p < k; the pre-detection value k takes any value k random in the section array S, that is, k = k random ∈ S.

4. The method for real-time vehicle accident detection based on Beidou navigation technology according to claim 1, characterized in that: In step S3, the position correlation is the correlation of the position changes between the two vehicles, so as to determine whether there is a connection between the position changes of the two vehicles, and the specific method is as follows: Where D is the straight-line distance between the two vehicles, R is the defined position correlation index, ΔV is the speed difference between the two vehicles, angle is the angle between the two vehicles, and β and m are weight parameters used to adjust the impact of different factors on the correlation.

5. The method for real-time vehicle accident detection based on Beidou navigation technology according to claim 1, characterized in that: In step S3, the time correlation is the consistency of the stay time between the vehicle and the adjacent vehicles, and the formula is as follows: In the formula, Δt is the time difference between the two vehicles, ￡ is the vehicle time correlation index, and the smaller the ￡ value, the higher the similarity.

6. The method for real-time vehicle accident detection based on Beidou navigation technology according to claim 1, characterized in that: In step S3, the state correlation is the correlation degree of the state changes between the two vehicles, and the calculation method is as follows: In the formula, I(C; D) represents the mutual information between random variables C and D; e and g represent different states of C and D, and S(C e ,D g ) Event C is in state C e And event D is in state D g The joint probability of S(C e ) is event C in state C e The marginal probability of S(D g ) is event D in state D g The marginal probability of .

7. The method for real-time vehicle accident detection based on Beidou navigation technology according to claim 1, characterized in that: In step S41, the specific steps of optimizing the hyperparameters of the accident detection model by the cuckoo search algorithm are as follows: S410: Initialize the group: randomly generate a group of candidate solutions; S411: Calculate the fitness of each solution according to the objective function; S412: Generate new candidate solutions; S413: Select a new solution to replace the old solution according to the fitness; S414: Eliminate some poor solutions based on probability; S415: Repeat steps S411 to S414 until the maximum number of iterations is reached; S416: Select the solution with the best fitness from among all candidate solutions as the output result; S417: Using the optimal fitness solution as a hyperparameter of the accident detection model to optimize the accident detection model.

8. The method for real-time vehicle accident detection based on Beidou navigation technology according to claim 1, characterized in that: In step S42, the training method of the accident detection model is as follows: S420: Inputting the correlation information into a random forest algorithm model; S421: Output the final prediction result of the random forest, which is the average of the prediction results of all features; S422: Calculate the accuracy rate according to the predicted structure of each feature; S423: Calculate the precision and recall rate according to the predicted number of samples; S424: Calculate the F1 score to achieve a balance between precision and recall.

9. A vehicle real-time accident detection system based on Beidou navigation technology, which adopts a vehicle real-time accident detection method based on Beidou navigation technology as claimed in any one of claims 1 to 8, characterized in that: include: Beidou vehicle-mounted terminal, which is used to send and receive Beidou navigation and positioning information; An information preprocessing module, which is used to process the navigation information of the vehicle on the road section to be tested; A calculation module, which is used to calculate the information correlation between vehicles in the road section to be tested; The accident detection module is used to establish an accident detection model combining the cuckoo search algorithm and the random forest algorithm, and output the detection result of whether an accident has occurred in combination with the information correlation obtained in the calculation module; An information sending module is used to send abnormal data information to the server; The monitoring server is used to receive abnormal data information and issue warnings to monitoring personnel.

10. A vehicle real-time accident detection system based on Beidou navigation technology according to claim 9, characterized in that: The information preprocessing module comprises: A positioning information storage module, which is used to store all road section positioning information and vehicle positioning information received by the Beidou vehicle-mounted terminal; The information retrieval module is used to quickly retrieve the road section to be tested.

Images