RB network abnormal vehicle speed early warning method based on gradient enhancement, equipment and medium

Abstract

The invention provides an RB network abnormal vehicle speed early warning method and device based on gradient enhancement and a medium, and relates to the field of machine learning technology vehicle networking safety, and the method comprises the steps: carrying out the data cleaning, abnormity elimination and noise removal of original data before an abnormal event occurs, and obtaining a training data set; calculating data item gradients to construct a sample library, classifying and sorting the sample library by adopting a random forest, and outputting variables and variable ranking; constructing a Bayesian network model for the associated variables to obtain an output result; according to an output result consistent with the variable ranking output by the random forest model, drawing an ROC curve function of the Bayesian network model, and carrying out prediction analysis on the vehicle speed to obtain an abnormal vehicle speed prediction result; according to the technical scheme of the invention, the prediction rate of special safety abnormal conditions in the Internet of Vehicles is improved, and the misjudgment rate of abnormal conditions is reduced.

Description

Technical Field

[0001] The present application relates to the field of vehicle network security using machine learning technology, and in particular to a method, device and medium for warning abnormal vehicle speed in a RB network based on gradient enhancement. Background Art

[0002] In the Internet of Vehicles, active safety technologies are designed to prevent accidents. The technologies that have been put into use include anti-lock braking systems (ABS), which prevent the tires from locking during emergency braking and maintain steering control. Electronic stability programs (ESP), which help vehicles stay stable when turning and prevent skidding. Blind spot monitoring systems (BSD), which monitor the blind spots around the vehicle through radar or cameras and warn the driver of approaching vehicles. Forward collision warning systems (FCW), which use lidar or cameras to monitor obstacles ahead and warn the driver of possible collisions. Automatic emergency braking systems (AEB), which automatically apply braking force to avoid or mitigate collisions when the driver fails to respond to FCW warnings. Lane departure warning systems (LDW) and lane keeping assist systems (LKA), which monitor whether the vehicle has deviated from the lane and provide warnings or slight corrections to the direction. Adaptive cruise control (ACC), which automatically adjusts the vehicle speed to maintain a safe distance from the vehicle in front. These technologies all rely on the vehicle's built-in electronic control and calculations, which are inevitably affected by human or environmental factors. With the application of a large number of roadside unit monitoring equipment and the birth of the Internet of Vehicles, vehicle driving safety has been improved. However, for vehicle safety warnings in the field of the Internet of Vehicles, considering the limited computing power of the vehicle and the real-time nature of traffic, the original roadside unit data is mostly input into a simple machine learning model to obtain analysis results. There are deficiencies in the automatic warning of abnormal vehicle speed based on the background of Internet of Vehicles safety machine learning, and there is a lack of a simple and efficient model for abnormal vehicle speed warning in the field of the Internet of Vehicles.

[0003] Bayesian networks use a graphical approach to represent the conditional independence assumptions between variables, making it possible to model and reason about complex probabilistic models. They are probabilistic graphical models used to represent a set of random variables and their conditional dependencies. Their modular nature makes them easy to expand and modify, and nodes can be added or deleted to adapt to new data or knowledge. Because the network structure intuitively reflects the causal relationship between variables, Bayesian networks provide high interpretability, which is particularly important for understanding and explaining model predictions. Using conditional independence, Bayesian networks can perform efficient reasoning and quickly calculate posterior probabilities even in large and complex networks.

[0004] The theoretical basis of Bayesian networks is Bayes' theorem, which describes how to update probability estimates based on prior knowledge and observed data. Therefore, Bayesian networks are very suitable for dealing with situations of uncertainty and incomplete information.

[0005] The existing Bayesian network and its variants only analyze and filter the raw data, which is not applicable to the current massive roadside unit data. They face the problems of insufficient real-time performance and insufficient analysis accuracy. After drawing the Bayesian network model, there is usually the problem of selecting a reasonable threshold. The simple statistical method used in traditional computational data analysis faces the problem of being unable to identify disturbed data and invalid data. Summary of the invention

[0006] The purpose of the present invention is to provide a RB network abnormal speed warning method, device and medium based on gradient enhancement in order to solve the problems of insufficient real-time performance, difficulty in balancing accuracy and false alarm rate of vehicle abnormal situation warning in existing vehicle networks.

[0007] The above-mentioned purpose of the present application is achieved through the following technical solutions: S1: Construct a data set based on the original data before the abnormal event occurs; S2: Clean the data set, eliminate anomalies and remove noise to obtain a training data set; S3: Calculate the gradient value of the traffic volume, road occupancy rate and speed of each data item in the training data set with respect to the road section distance, and construct a sample library D; S4: Use the random forest model to classify the variables in the sample library D to obtain a preset number of variables with important attributes and variable rankings; the variables include: traffic flow, road section occupancy rate, speed, and gradient value; S5: Use the KMean method to convert the variables of the preset number of important attributes into discrete variables; determine whether there is a correlation between the discrete variables: If there is a correlation, a Bayesian network model is constructed to obtain the output result; If there is no association, repeat step S4; S6: Determine whether the output result is consistent with the variable ranking output by the random forest model; If not consistent, repeat step S4; If they are consistent, the ROC curve function of the Bayesian network model is drawn according to the output results to determine the threshold point; based on the threshold point, the vehicle speed is predicted and analyzed to obtain the abnormal speed prediction result and complete the abnormal speed warning.

[0008] Optionally, the raw data is selected from available data in a roadside unit sensor device data set, including: vehicle flow , road section occupancy rate and speed ; The original data includes: training data and comparative training data; Get the training data of the preset time period before the abnormal event of the Internet of Vehicles occurs, recorded as: , , ; Obtain comparative training data without abnormal events in the same preset time period on different dates, recorded as: , , .

[0009] Optionally, step S2 includes: S21: Delete invalid data items in the data set and construct a data item set D for abnormal warning of the road in the Internet of Vehicles; the invalid data items include: data items with a road section occupancy rate of 0 and a vehicle flow greater than 0, and data items with an average speed greater than 0 and a vehicle flow equal to 0; At every preset time interval T, the data items of the vehicle network road abnormal warning data item set D are averaged; S22: Determine whether the speed change rate of the abnormal data item in the vehicle network road abnormal warning data item set D is less than a preset change rate threshold: If it is less than the preset change rate threshold, the speed value in the abnormal data item is set to the minimum speed value, and the process jumps to step S21; S23: If it is greater than or equal to the preset change rate threshold, determine whether the road section distance of each sample in the data item set D is less than the preset value L; If it is greater than or equal to the preset value L, adjust the time interval to , jump to step S21; If it is less than the preset value L, the corresponding data item is retained to obtain the training data set.

[0010] Optionally, step S4 includes: S41: Establish a random forest model and set the number of trees and tree depth of the random forest; S42: Use the OOB error estimation method to evaluate the generalization error of the random forest model; Adjust the number of trees and tree depth of the random forest according to the generalization error; S43: Calculate the average decrease in the Gini index of each variable in the sample library D; based on the average decrease, select a preset number of variables with important attributes and determine the variable ranking.

[0011] Optionally, step S5 includes: S51: The sample library D contains discrete variables that cause abnormal vehicle speed events, and the discrete variables include: vehicle flow, road section occupancy rate, speed, gradient value of vehicle flow to road section distance, gradient value of road section occupancy rate to road section distance, and gradient value of speed to road section distance; S52: The attributes of the sample library D and the data items are taken as the cluster center points, and the cluster center is recorded as ; S53: Bring the cluster center into the loss function, and the loss function is defined as:

[0012] in, Representative samples, represent The cluster to which it belongs, Represents the center point of the cluster. represents the total number of samples; S54: For each sample , assign it to the nearest cluster center:

[0013] in, represents the cluster center of the φth iteration cluster update; Represents the φth iteration cluster The center point of represents the number of clusters generated after the φth iteration of clustering; Recompute the center of the class:

[0014] S55: Order is the number of iteration steps, and step S54 is repeated until the loss function of the Bayesian network model converges to obtain the output result.

[0015] Optionally, step S6 includes: Determine the threshold point from the interval where the ROC curve changes from steep to flat; According to the threshold point, the vehicle speed is predicted and analyzed to obtain the prediction result.

[0016] An electronic device includes a processor, a memory, a user interface and a network interface, wherein the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes a RB network abnormal vehicle speed warning method based on gradient enhancement.

[0017] A computer-readable storage medium stores instructions, and when the instructions are executed, a gradient-enhanced RB network abnormal vehicle speed warning method is executed.

[0018] The beneficial effects of the technical solution provided by this application are: 1. When selecting data sets from massive Internet of Vehicles data, the "mean cleaning data set" model is introduced to solve the problem that the existing Internet of Vehicles model does not take into account the data disturbance due to hardware or environmental conditions, and protects the validity of the data.

[0019] 2. When processing data sets, the gradient value of vehicle flow, road occupancy rate and speed with respect to road section distance of each data item is calculated; this solves the problem of insufficient data validity faced by the existing Internet of Vehicles data processing and analysis, which uses simple statistical analysis of raw data to ensure real-time performance.

[0020] 3. In the method of using machine learning for data analysis, the traditional machine learning model often leads to overfitting of the trained model due to the correlation between data. To address this problem, the present invention proposes a G-parameter random forest analysis method, which uses random forest to analyze the importance of variables, calculates the average decrease in the Gini index, evaluates the importance of data attributes, selects important attribute data, excludes irrelevant or extremely small correlation data, and ensures real-time performance in the Internet of Vehicles.

[0021] 4. The random forest model is used to classify the data, which solves the problem that the simple statistical method used in traditional computational data analysis cannot identify disturbed data and invalid data.

[0022] 5. Establish an R-type Bayesian network model and select a reasonable threshold value to improve the accuracy of warning abnormal events in the Internet of Vehicles and reduce the false alarm rate. A high threshold means low sensitivity, which will miss potential dangerous accident prediction alarms; a low threshold means a high false alarm rate, which will misreport normal situations as abnormal accidents. Establish a curve function according to different thresholds, and take the extreme threshold value under the ideal conditions of maximum accuracy and minimum false alarm rate. The present invention can predict and alarm special situations in the Internet of Vehicles, such as blind spots, slippery local roads, fatigue driving, non-compliance with traffic regulations, illegal parking and other abnormal behaviors, so as to establish a Bayesian network model for analyzing key Internet of Vehicles events, specifically analyze the connection between key events, improve the prediction rate of special safety abnormal situations in the Internet of Vehicles, and reduce the misjudgment rate of abnormal situations. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The present application will be further described below with reference to the accompanying drawings and embodiments, in which: Figure 1 is a flow chart in an embodiment of the present application; Figure 2 is a data preprocessing flow chart in an embodiment of the present application; Figure 3 is a calculation flow chart for establishing a random forest model in an embodiment of the present application; Figure 4is a modeling flow chart for establishing an R-type Bayesian network in an embodiment of the present application; Figure 5 It is a data comparison diagram of the present invention and the existing model in the embodiment of the present application; Figure 6 It is a schematic diagram of the structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to have a clearer understanding of the technical features, purposes and effects of the present application, the specific implementation methods of the present application are now described in detail with reference to the accompanying drawings.

[0025] An embodiment of the present application provides a RB network abnormal vehicle speed warning method based on gradient enhancement.

[0026] Please refer to Figure 1 , Figure 1 It is a flow chart of a RB network abnormal vehicle speed warning method based on gradient enhancement in an embodiment of the present application, including: S1: Construct a data set based on the original data before the abnormal event occurs; The raw data is selected from available data in the roadside unit sensor equipment data set, including: vehicle flow , road section occupancy rate and speed ; The original data includes: training data and comparative training data; Get the training data of the preset time period before the abnormal event of the Internet of Vehicles occurs, recorded as: , , ; Specifically, because the accident occurs later than the recording time, valid vehicle state changes can be recorded one hour before the accident.

[0027] Obtain comparative training data without abnormal events in the same preset time period on different dates, recorded as: , , .

[0028] Specifically, the comparative training data is used for comparative training, and also includes traffic volume, vehicle speed, and road occupancy rate.

[0029] S2: Clean the data set, eliminate anomalies and remove noise to obtain a training data set; Figure 2 The figure shows the data preprocessing flow chart of this application.

[0030] Step S2 includes: S21: Delete invalid data items in the data set and construct a data item set D for abnormal warning of the road in the Internet of Vehicles; the invalid data items include: data items with a road section occupancy rate of 0 and a vehicle flow greater than 0, and data items with an average speed greater than 0 and a vehicle flow equal to 0; At every preset time interval T, the data items of the vehicle network road abnormal warning data item set D are averaged; S22: Determine whether the speed change rate of the abnormal data item in the vehicle network road abnormal warning data item set D is less than a preset change rate threshold: If it is less than the preset change rate threshold, the speed value in the abnormal data item is set to the minimum speed value, and the process jumps to step S21; S23: If it is greater than or equal to the preset change rate threshold, determine whether the road section distance of each sample in the data item set D is less than the preset value L; If it is greater than or equal to the preset value L, adjust the time interval to , jump to step S21; If it is less than the preset value L, the corresponding data item is retained to obtain the training data set.

[0031] Specifically, the roadside unit equipment will generate continuously disturbed traffic data due to hardware problems and environmental weather problems, so the present invention takes the average value every 5 minutes in the data set to reduce the noise in the data item set D and obtain the training data set D, where are the n data items in D.

[0032] S3: Calculate the gradient value of the traffic volume, road occupancy rate and speed of each data item in the training data set with respect to the road section distance, and construct a sample library D; S4: Use the random forest model to classify the variables in the sample library D to obtain a preset number of variables with important attributes and variable rankings; the variables include: traffic flow, road section occupancy rate, speed, and gradient value; Specifically, Figure 3 It is the computational flow chart of the random forest model.

[0033] Step S4 includes: S41: Establish a random forest model and set the number of trees and tree depth of the random forest; S42: Use the OOB error estimation method to evaluate the generalization error of the random forest model; Adjust the number of trees and tree depth of the random forest according to the generalization error; Specifically, the OOB strategy, ‌out-of-bag strategy, ‌ allows us to extract multiple sub-samples from the original dataset with replacement, ‌ each sub-sample is used to train a base learner. ‌ Since in each sub-sample, ‌ only about 63.2% of the samples are extracted, ‌ the remaining about 36.8% of the samples can be used as OOB samples for model evaluation. ‌

[0034] S43: Calculate the average decrease in the Gini index of each variable in the sample library D; based on the average decrease, select a preset number of variables with important attributes and determine the variable ranking.

[0035] Specifically, in the Internet of Vehicles, the real-time calculation time is shorter while ensuring good prediction results. The average decrease in the Gini index is calculated, which indicates the importance of the corresponding variable. The top ten important parameters are selected to construct the Bayesian network model. These ten parameters have a major impact on abnormal events in the Internet of Vehicles.

[0036] Specifically, in large data sets, the correlation between data often leads to overfitting of the trained model. Therefore, random forest is used to analyze the importance of variables, evaluate the importance of data attributes, and select important attribute data.

[0037] S5: Use the KMean method to convert the variables of the preset number of important attributes into discrete variables; determine whether there is a correlation between the discrete variables: If there is a correlation, a Bayesian network model is constructed to obtain the output result; If there is no association, repeat step S4; Step S5 includes: S51: The sample library D contains discrete variables that cause abnormal vehicle speed events, and the discrete variables include: vehicle flow, road section occupancy rate, speed, gradient value of vehicle flow to road section distance, gradient value of road section occupancy rate to road section distance, and gradient value of speed to road section distance; S52: The attributes of the sample library D and the data items are taken as the cluster center points, and the cluster center is recorded as ; S53: Bring the cluster center into the loss function, and the loss function is defined as:

[0038] in, Representative samples, represent The cluster to which it belongs, Represents the center point of the cluster. represents the total number of samples; S54: For each sample , assign it to the nearest cluster center:

[0039] in, represents the cluster center of the φth iteration cluster update; Represents the φth iteration cluster The center point of represents the number of clusters generated after the φth iteration of clustering; Recompute the center of the class:

[0040] S55: Order is the number of iteration steps, and step S54 is repeated until the loss function of the Bayesian network model converges to obtain the output result.

[0041] S6: Determine whether the output result is consistent with the variable ranking output by the random forest model; If not consistent, repeat step S4; If they are consistent, the ROC curve function of the Bayesian network model is drawn according to the output results to determine the threshold point; based on the threshold point, the vehicle speed is predicted and analyzed to obtain the abnormal speed prediction result and complete the abnormal speed warning.

[0042] Specifically, a modeling flowchart of the RB network (R-type Bayesian network, i.e., the Bayesian network enabled by random forest) is established.

[0043] Step S6 includes: Determine the threshold point from the interval where the ROC curve changes from steep to flat; According to the threshold point, the vehicle speed is predicted and analyzed to obtain the prediction result.

[0044] Aiming at the problems faced by the existing Internet of Vehicles data analysis, a RB network abnormal speed warning model based on gradient enhancement is proposed in the field of Internet of Vehicles by applying machine learning methods. The efficient Bayesian network is used to warn the abnormal speed risk of vehicles in real time, so as to describe the Bayesian network that selects the ten most likely events that lead to vehicle safety failure from a large number of Internet of Vehicles big data. Through ROC curve analysis, the accuracy of the Bayesian network warning is improved, the error rate of the warning is reduced, and specific optimization is carried out for vehicle speed while ensuring real-time performance.

[0045] The present invention improves the deficiencies of abnormal vehicle speed warning analysis in the field of Internet of Vehicles, and solves the problem of low speed warning accuracy of existing Internet of Vehicles models. Taking into account the stability of hardware and the validity of data during weather interference, the validity of the Internet of Vehicles data set is ensured by statistically analyzing roadside unit data and using "mean cleaning and gradient enhancement". Each tuple is represented by a data item, and a data gradient is constructed through statistical data. According to the roadside units of a specific road section, an efficient and useful data set is obtained to ensure the effective amount of information in the data. This is a rough processing of the metadata. After the data set is classified and analyzed using random forests, it is determined which specific data items are highly sensitive to abnormal vehicle speed events, which is of great significance to the safety of Internet of Vehicles vehicles.

[0046] Specifically, Figure 5 It is a comparison chart of the RB network abnormal vehicle speed warning model based on gradient enhancement of the present invention and the existing vehicle network warning model. Table 1 is a data set sample attribute description of an embodiment of the present invention. Table 2 is a comparison data table of the RB network abnormal vehicle speed warning model based on gradient enhancement established in an embodiment of the present invention and the existing vehicle network warning model.

[0047] Table 1

[0048] Table 2

[0049] The present application also discloses an electronic device. Figure 6 , Figure 6 The electronic device 500 may include: at least one processor 501 , at least one network interface 504 , a user interface 503 , a memory 505 , and at least one communication bus 502 .

[0050] The communication bus 502 is used to realize the connection and communication between these components.

[0051] The user interface 503 may include a display screen, and the optional user interface 503 may also include a standard wired interface or a wireless interface.

[0052] The network interface 504 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0053] The present application also discloses a computer-readable storage medium, which stores a plurality of instructions, and the instructions are suitable for a processor to load to execute the above-mentioned RB network abnormal vehicle speed warning method based on gradient enhancement.

[0054] The above are only exemplary embodiments of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure.

[0055] This application is intended to cover any variation, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art not described in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A RB network abnormal vehicle speed warning method based on gradient enhancement, characterized in that: The method comprises the following steps: S1: Construct a data set based on the original data before the abnormal event occurs; S2: Clean the data set, eliminate anomalies and remove noise to obtain a training data set; S3: Calculate the gradient value of the traffic volume, road occupancy rate and speed of each data item in the training data set with respect to the road section distance, and construct a sample library D; S4: Use the random forest model to classify the variables in the sample library D to obtain a preset number of variables with important attributes and variable rankings; the variables include: traffic flow, road section occupancy rate, speed, and gradient value; S5: Use the KMean method to convert the variables of the preset number of important attributes into discrete variables; determine whether there is a correlation between the discrete variables: If there is a correlation, a Bayesian network model is constructed to obtain the output result; If there is no association, repeat step S4; S6: Determine whether the output result is consistent with the variable ranking output by the random forest model; If not consistent, repeat step S4; If they are consistent, the ROC curve function of the Bayesian network model is drawn according to the output results to determine the threshold point; based on the threshold point, the vehicle speed is predicted and analyzed to obtain the abnormal speed prediction result and complete the abnormal speed warning.

2. The RB network abnormal vehicle speed warning method based on gradient enhancement as claimed in claim 1 is characterized in that: The raw data is selected from available data in the roadside unit sensor equipment data set, including: vehicle flow , road section occupancy rate and speed ; The original data includes: training data and comparative training data; Get the training data of the preset time period before the abnormal event of the Internet of Vehicles occurs, recorded as: , , ; Obtain comparative training data without abnormal events in the same preset time period on different dates, recorded as: , , .

3. The RB network abnormal vehicle speed warning method based on gradient enhancement as claimed in claim 2 is characterized in that: Step S2 includes: S21: Delete invalid data items in the data set and construct a data item set D for abnormal warning of the road in the Internet of Vehicles; the invalid data items include: data items with a road section occupancy rate of 0 and a vehicle flow greater than 0, and data items with an average speed greater than 0 and a vehicle flow equal to 0; At every preset time interval T, the data items of the vehicle network road abnormal warning data item set D are averaged; S22: Determine whether the speed change rate of the abnormal data item in the vehicle network road abnormal warning data item set D is less than a preset change rate threshold: If it is less than the preset change rate threshold, the speed value in the abnormal data item is set to the minimum speed value, and the process jumps to step S21; S23: If it is greater than or equal to the preset change rate threshold, determine whether the road section distance of each sample in the data item set D is less than the preset value L; If it is greater than or equal to the preset value L, adjust the time interval to , jump to step S21; If it is less than the preset value L, the corresponding data item is retained to obtain the training data set.

4. The method for warning abnormal vehicle speed based on gradient enhancement of RB network as claimed in claim 1, characterized in that: Step S4 includes: S41: Establish a random forest model and set the number of trees and tree depth of the random forest; S42: Use the OOB error estimation method to evaluate the generalization error of the random forest model; Adjust the number of trees and tree depth of the random forest according to the generalization error; S43: Calculate the average decrease in the Gini index of each variable in the sample library D; based on the average decrease, select a preset number of variables with important attributes and determine the variable ranking.

5. The method for warning abnormal vehicle speed based on gradient enhancement of RB network as claimed in claim 1, characterized in that: Step S5 includes: S51: The sample library D contains discrete variables that cause abnormal vehicle speed events, and the discrete variables include: vehicle flow, road section occupancy rate, speed, gradient value of vehicle flow to road section distance, gradient value of road section occupancy rate to road section distance, and gradient value of speed to road section distance; S52: The attributes of the sample library D and the data items are taken as the cluster center points, and the cluster center is recorded as ; S53: Bring the cluster center into the loss function, and the loss function is defined as: in, Representative samples, represent The cluster to which it belongs, Represents the center point of the cluster. represents the total number of samples; S54: For each sample , assign it to the nearest cluster center: in, represents the cluster center of the φth iteration cluster update; Represents the φth iteration cluster The center point of represents the number of clusters generated after the φth iteration of clustering; Recompute the center of the class: S55: Order is the number of iteration steps, and step S54 is repeated until the loss function of the Bayesian network model converges to obtain the output result.

6. The method for warning abnormal vehicle speed based on gradient enhancement of RB network as claimed in claim 1, characterized in that: Step S6 includes: Determine the threshold point from the interval where the ROC curve changes from steep to flat; According to the threshold point, the vehicle speed is predicted and analyzed to obtain the prediction result.

7. An electronic device, characterized in that: The electronic device comprises a processor (501), a memory (505), a user interface (503) and a network interface (504), wherein the memory (505) is used to store instructions, the user interface (503) and the network interface (504) are used to communicate with other devices, and the processor (501) is used to execute the instructions stored in the memory (505) so that the electronic device executes the method according to any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed by a computer, the method according to any one of claims 1 to 6 is executed.

Images