Serious conflict and collision prediction method, system and device considering driving distraction and storage medium

Abstract

The invention provides a serious conflict and collision prediction method, system and device considering driving distraction and a storage medium. The prediction method comprises the steps of obtaining driving data of a vehicle; inputting the driving data into a driving distraction recognition model, and recognizing to obtain driving distraction state features; and inputting the driving distraction state characteristics and the driving data into a serious conflict and collision prediction model, predicting a serious conflict event and a collision event, and outputting a prediction result. According to the method, conflict types are added into a consideration range of traffic conflict recognition, a distraction recognition model is built based on machine learning, and a driving distraction state is used as input of a serious conflict and collision prediction model; driving distraction features are considered in severe conflict and collision prediction, so that the prediction accuracy and sensitivity of the model are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of road traffic safety, and relates to a method, system, device and storage medium for predicting serious conflicts and collisions considering driver distraction. Background Art

[0002] With the increasing number of automobiles year by year, the road traffic environment has become more complex and changeable, resulting in more frequent traffic conflicts. As the primary form of traffic accidents, traffic conflicts can evolve into traffic accidents due to inappropriate avoidance behaviors, and their essence is the unsafe state of the "human-vehicle-road" system. Reasonably classifying the severity of traffic conflicts can enrich conflict theory and further study the internal relationship between traffic conflicts and traffic accidents. In most studies, the severity of traffic conflicts only considers the possibility of collision and does not consider the severity of the accident after collision.

[0003] In the human-vehicle-road traffic system, the motor vehicle driver is the weakest link. Driver distraction will further exacerbate traffic conflicts and lead to an increase in traffic accidents. Driver distraction significantly affects driving performance, reduces the ability to obtain road information, weakens the ability to control the vehicle, and increases the risk of traffic accidents. According to the statistical analysis of road traffic accidents, 90% of traffic accidents are caused by drivers, and driver distraction is an important cause of traffic accidents. However, existing collision prediction studies mostly focus on trajectory or speed prediction and do not fully consider driver information (such as driver distraction status). In addition, few studies use traffic conflicts to assist in predicting collisions at the driving level to improve the efficiency and accuracy of prediction. These have all led to the problem of low accuracy in collision prediction. Summary of the Invention

[0004] In order to solve the above problems of the prior art, the present invention provides a method, system, device and storage medium for predicting serious conflicts and collisions considering driver distraction, and realizes accurate prediction of serious traffic conflicts and collisions.

[0005] The present invention is realized through the following technical solutions: In a first aspect, the present invention provides a method for predicting serious conflicts and collisions considering driver distraction, including: Obtaining driving data of a vehicle; Inputting the driving data into a driver distraction recognition model to identify the driver distraction state characteristics; Inputting the driver distraction state characteristics and the driving data into a serious conflict and collision prediction model to predict serious conflict events and collision events, and outputting a prediction result; wherein, the class labels in the serious conflict and collision prediction model include non-interference passing events, minor conflict events, general conflict events, serious conflict events, and collision events; Among them, the driving distraction recognition model is constructed based on a support vector machine, and the construction method includes: Construct different driving scenarios through a driving simulator, and obtain the driving performance data of the driver in different scenarios; Using the driving performance data as samples, calculate the mean and standard deviation as time-domain features through statistical methods, extract the low-frequency energy value of the driving performance variable as the frequency-domain feature through fast Fourier transform, and screen the time-domain and frequency-domain features through non-parametric test methods; Standardize the selected features and use them as the features of the samples to jointly form the input of the driving distraction recognition model. Based on the support vector machine, build the driving distraction recognition model and train the driving distraction recognition model in the way of K-fold cross-validation.

[0006] Preferably, the driving data includes: the speed of the host vehicle, the speed of the leading vehicle, the speed of the adjacent vehicle, the distance between the host vehicle and the leading vehicle, the longitudinal distance between the host vehicle and the adjacent vehicle, the lateral distance between the host vehicle and the adjacent vehicle, and the steering wheel angle in the driving cockpit.

[0007] Furthermore, the selected features include: the mean value of the lane departure amount, the standard deviation of the steering wheel angle, the standard deviation of the lane departure amount, the low-frequency energy value of the steering wheel angle, and the low-frequency energy value of the lane departure amount.

[0008] Furthermore, the construction method of the severe conflict and collision prediction model includes: Input the driving performance data into the driving distraction recognition model to identify the driving distraction state features; According to the driving performance data, identify the interaction mode between the surrounding vehicles and the host vehicle through the lane subordination relationship, initially screen the conflict events in different interaction modes based on the traffic conflict recognition index, and then combine the hazard avoidance behavior index to identify the longitudinal traffic conflict and lateral traffic conflict of the initially screened conflict events to obtain the conflict events; According to the traffic conflict severity index, perform clustering analysis on the conflict events based on the K-means++ clustering algorithm to obtain the clustering results; among them, the categories of the clustering results include non-interference passing events, minor conflict events, general conflict events, severe conflict events, and collision events; Build a severe conflict and collision prediction model based on LSTM. Combine the conflict events, the corresponding clustering results, and the driving distraction state features to form a sample data set, and use the probability values of severe conflict events and collision events as the prediction variables to train the severe conflict and collision prediction model.

[0009] Furthermore, the traffic conflict recognition index is the time to collision; the traffic conflict severity index includes the exposed time to collision, the cumulative critical time to collision, and the deceleration for collision avoidance.

[0010] Second aspect, the present invention provides a serious conflict and collision prediction system considering driving distraction, including: A data acquisition module for acquiring driving data of a vehicle; A driving distraction recognition module for inputting the driving data into a driving distraction recognition model to identify driving distraction state features; A prediction module for inputting the driving distraction state features and the driving data into a serious conflict and collision prediction model to predict serious conflicts and collisions and output a prediction result; wherein, the class labels in the serious conflict and collision prediction model include non-disturbed passing events, minor conflict events, general conflict events, serious conflict events, and collision events; Among them, the driving distraction recognition model is constructed based on a support vector machine, and the construction method includes: Constructing different driving scenarios through a driving simulator to obtain the driving performance data of the driver under different scenarios; Taking the driving performance data as samples, calculating the mean and standard deviation as time-domain features through statistical methods, extracting the low-frequency energy value of the driving performance variable as a frequency-domain feature through fast Fourier transform, and screening the time-domain features and frequency-domain features through non-parametric test methods; Normalizing the selected features and using them as the features of the samples to jointly constitute the input of the driving distraction recognition model. Based on the support vector machine, build the driving distraction recognition model and train the driving distraction recognition model in a K-fold cross-validation manner.

[0011] Preferably, the driving data includes: the speed of the host vehicle, the speed of the leading vehicle, the speed of the adjacent vehicle, the distance between the host vehicle and the leading vehicle, the longitudinal distance between the host vehicle and the adjacent vehicle, the lateral distance between the host vehicle and the adjacent vehicle, and the steering wheel angle in the driving cockpit.

[0012] Further, the selected features include five features: the mean value of the lane departure amount, the standard deviation of the steering wheel angle, the standard deviation of the lane departure amount, the low-frequency energy value of the steering wheel angle, and the low-frequency energy value of the lane departure amount.

[0013] Third aspect, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned serious conflict and collision prediction method considering driving distraction when executing the computer program.

[0014] Fourth aspect, the present invention provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, wherein the computer program implements the above-mentioned serious conflict and collision prediction method considering driving distraction when executed by a processor.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The method for predicting severe conflicts and collisions considering driving distraction provided by the present invention incorporates the conflict type into the consideration scope of traffic conflict recognition, and builds a driving distraction recognition model based on driving performance data and machine learning. The driving data is input into the driving distraction recognition model to determine whether there is driving distraction. The driving data can more directly reflect the results of driving behaviors, such as lane departure, hard braking, etc., which are the direct consequences of distraction and have higher accuracy compared to the image method. Moreover, the driving data is usually directly obtained through the CAN bus with low processing delay, which is suitable for real-time warning. Taking the driving distraction state as the input of the severe conflict and collision prediction model and considering the driving distraction characteristics in the severe conflict and collision prediction significantly improves the prediction accuracy and sensitivity of the model. Using traffic conflicts to assist in the collision prediction at the driving level improves the efficiency and precision of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 It is a schematic flowchart of an embodiment for constructing a severe traffic conflict and collision model considering driving distraction provided by the present invention; Figure 2 It is a schematic diagram of the result of an embodiment showing the influence of driving distraction on the severity of traffic conflicts in an embodiment of the present invention; Figure 3 It is a schematic diagram of the result of an embodiment of the method for predicting severe traffic conflicts and collisions considering driving distraction provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, rather than all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0019] It should be understood that the schematic drawings are not drawn to actual scale. The flowcharts used in the present invention illustrate operations implemented according to some embodiments of the present invention. It should be understood that the operations in the flowchart may not be implemented in sequence, and steps without a logical context relationship may be reversed or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart or remove one or more operations from the flowchart under the guidance of the content of the present invention.

[0020] Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor systems and / or microcontroller systems.

[0021] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in connection with the embodiments can be included in at least one embodiment of the present invention. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0022] Before the description of the embodiments, the relevant terms are defined as follows: Cluster analysis: Cluster analysis is a commonly used method in the field of data mining. It divides samples into different clusters, and requires that the samples within the same cluster have small differences and the samples between different clusters have large differences, so as to mine the deep information inside the data. K-means is a classic unsupervised clustering algorithm, that is, it does not need to label the samples in advance, and the algorithm will automatically learn the internal rules of the data. Because of its fast running speed and strong applicability, it is widely used. However, the K-means clustering algorithm randomly selects the clustering centers for the first time, which may make the clustering centers too close, resulting in the algorithm possibly converging to different local minima. K-means++ is an improved version of the K-means clustering algorithm. The principle is to make the initial clustering centers as scattered as possible through constraint conditions, which can effectively overcome the above problems.

[0023] SVM: SVM (support vector machine) is a support vector machine, which is a machine learning method that can be trained. It is a supervised learning model and related learning algorithms for analyzing data in classification and regression analysis. Given a set of training instances, each training instance is labeled as belonging to one of two classes, and the SVM training algorithm creates a model that assigns new instances to one of the two classes, making it a non-probabilistic binary linear classifier.

[0024] LSTM: Long Short-Term Memory (LSTM) is a deep learning model commonly used to process sequential data. Compared with traditional RNN (Recurrent Neural Network), LSTM introduces three gates (input gate, forget gate, output gate) and a cell state, which enable LSTM to better handle long-term dependencies in sequences.

[0025] An embodiment of the present invention provides a method for predicting severe traffic conflicts and collisions considering driver distraction, which will be described in detail below.

[0026] As Figure 1 shown, it is a schematic flowchart of an embodiment of the process for constructing a severe traffic conflict and collision prediction model considering driver distraction provided by the present invention, including: Step 1, obtain the driving performance data of the driver as the driving performance data set.

[0027] Specifically, different driving scenarios are constructed through a driving simulator, and the driving performance data of the driver under different scenarios is obtained as the driving performance data set.

[0028] The driving performance data set includes the design of driving distraction subtasks and the design of traffic risk events; specifically, the driving performance data includes: data such as the speed of each vehicle, the acceleration of each vehicle, the centroid position coordinates of each vehicle, and the steering wheel angle in the driving cockpit.

[0029] Step 2, propose a traffic conflict recognition method considering conflict types, divide the severity of conflicts based on cluster analysis, construct an index system for the severity of traffic conflicts, and explore the impact of driver distraction on traffic conflicts.

[0030] Step 2.1, propose a traffic conflict recognition method considering conflict types, and the recognition method proves the necessity of dividing the severity of traffic conflicts.

[0031] Specifically, combining the evasive behavior and the spatio-temporal proximity degree, propose a traffic conflict recognition method considering conflict types. The recognition method first identifies the interaction patterns between the surrounding vehicles and the host vehicle through the lane affiliation relationship according to the driving performance data set, then preliminarily screens the conflict events in different interaction patterns based on the traffic conflict recognition index, and finally combines the evasive behavior index (braking deceleration) to complete the recognition of longitudinal traffic conflicts and lateral traffic conflicts, obtaining the final conflict events.

[0032] In this embodiment of the present invention, TTC (Time to Collision) is selected as the traffic conflict recognition index:

[0033] Among them, is the longitudinal distance between the leading vehicle and the host vehicle, is the speed difference between the leading vehicle and the host vehicle, and are the longitudinal position coordinates of the leading vehicle and the host vehicle respectively, is the vehicle length parameter, and are the vehicle speeds of the leading vehicle and the host vehicle respectively.

[0034] Step 2.2: Select the traffic conflict severity index, divide the traffic conflict severity based on the K-means++ clustering algorithm, and construct the traffic conflict severity index system.

[0035] Specifically, select the traffic conflict severity index from the spatial distance index, time distance index, speed index, deceleration index, energy index, etc. Divide the traffic conflict severity based on the selected indexes through the K-means++ clustering algorithm to explore the impact of driving distraction on traffic conflicts.

[0036] In this embodiment of the present invention, the traffic conflict severity index specifically includes the Time Exposed TTC (TET), Time Integrated TTC (TIT), and Deceleration Rate to Avoid the Crash (DRAC), which are calculated based on the driving performance data.

[0037] The embodiment of the present invention reasonably sets the number of clustering clusters, and uses the K-means++ clustering algorithm to divide the traffic events in the driving performance data set obtained in Step 1 into non-interference passing events, minor conflict events, general conflict events, severe conflict events, and collision events, and uses them as the prediction label basis for Step 4.

[0038] Step 2.3: Explore the impact of driving distraction on traffic conflicts. Divide the driving performance data set obtained in Step 1 into a non-distracted sub-task group and a driving distraction group. Figure 2 The figure shows the comparison chart of traffic conflict situations between the non-distracted sub-task group and the driving distraction group in Step 2.3 of the present invention. Compared with the non-distracted sub-task group, in the driving distraction state, the number of general conflict events, severe conflict events, and collision events increases; at the same time, the number of non-interference passing events and minor conflict events decreases. Through comparative analysis, it shows that in the same driving environment, when dealing with the same risk events (including severe conflicts and collisions), it is crucial to consider the driving distraction state when predicting severe conflicts and collisions, and the prerequisite for considering the driving distraction state is the accurate identification of driving distraction.

[0039] Step 3: Establish a driving distraction recognition model based on machine learning. Calculate time-domain features through statistical methods, extract the low-frequency energy values of driving performance variables as frequency-domain features through the Fast Fourier Transform (FFT), and screen the time-domain features and frequency-domain features based on non-parametric test methods; subsequently, combine the screened time-domain features and frequency-domain features to build a driving distraction recognition model based on the Support Vector Machine (SVM).

[0040] Step 3.1: Select driving performance variables, analyze the time-domain and frequency-domain features of the selected driving performance variables, and perform feature screening through non-parametric tests.

[0041] Specifically, use the driving performance data obtained in Step 1 as samples, calculate the mean and standard deviation as time-domain features through statistical methods, extract the low-frequency energy values of driving performance variables as frequency-domain features through the Fast Fourier Transform, and finally screen the time-domain features and frequency-domain features through non-parametric test methods.

[0042] In this embodiment of the present invention, the driving performance variables include steering wheel angle, lane departure amount, lateral speed, vehicle speed, and following distance.

[0043] Through the analysis of the driving performance variables by non-parametric test methods, five features, namely the mean value of the lane departure amount, the standard deviation of the steering wheel angle, the standard deviation of the lane departure amount, the low-frequency energy value of the steering wheel angle, and the low-frequency energy value of the lane departure amount, are finally selected for the driving distraction recognition model.

[0044] Step 3.2: Build a driving distraction recognition model based on machine learning algorithms.

[0045] Specifically, standardize the features screened in Step 3.1 and use them as the features of the samples to jointly form the input of the driving distraction recognition model. Build driving distraction recognition models based on the machine learning algorithms of Support Vector Machine (SVM), Random Forest (RF), and XGBoost respectively, and train the three driving distraction recognition models respectively by means of K-fold cross-validation.

[0046] It should be noted that SVM, RF, and XGBoost are all suitable for classification problems and regression problems. The classification principle of SVM is to place the data set in an n-dimensional space that matches its feature dimension. The algorithm needs to find a hyperplane to separate data of different classes, and the objective function tries to maximize the sum of the distances from each data point to the hyperplane to find the best classification basis, which is suitable for small-scale data sets. RF is an ensemble learning algorithm based on decision trees. By constructing multiple decision trees and combining the prediction results of all trees, the performance of the model can be improved, and it is not easy to overfit. XGBoost combines multiple weak classifiers, iteratively adjusts the parameters by means of gradient descent, and at the same time introduces a regularization term in the loss function to limit the complexity of the model and reduce the overfitting phenomenon.

[0047] It should be noted that cross-validation is a commonly used model selection and validation method in the field of machine learning. Its basic idea is to divide the entire data set into a data set and a test set. K-fold cross-validation evenly divides the entire data set into K groups. Each time, the data of K - 1 groups are used to train the model, and the data of the remaining one group are used to test the classification effect of the model (repeated K times). The value of K is generally 5 or 10, and the average performance of the model in K times is used as the final score of the classification model. K-fold cross-validation can more comprehensively display the classification ability of the model and make the classification effect more scientific and persuasive.

[0048] Step 3.3, evaluating the driving distraction recognition model.

[0049] Specifically, the present invention uses precision to compare the driving distraction recognition effects of the driving distraction recognition models constructed based on three algorithms of SVM, RF, and XGBoost. The results show that the precision of the driving distraction recognition model constructed based on SVM is the best. Therefore, in the embodiments of the present invention, the driving distraction recognition result of the driving distraction recognition model constructed based on SVM is used as the driving distraction state feature for input to the severe conflict and collision prediction model.

[0050] Step 4, under the condition of considering the driving distraction state feature, build a severe conflict and collision prediction model to achieve accurate prediction of severe conflicts and collisions.

[0051] Step 4.1, analyze the trigger time of severe conflict events and construct a sample data set for predicting severe conflict and collision events.

[0052] Specifically, based on the clustering results of the severity of traffic conflicts in step 2.2, severe conflict events and collision events are used as positive class samples, with the label "label" recorded as 1, and non-interference passing events, minor conflict events, and general conflict events are used as negative class samples, with the label "label" recorded as 0. The positive class samples, negative class samples, and their driving distraction state features are combined to form a sample data set for severe conflict and collision prediction.

[0053] In the embodiment of the present invention, each frame of data in the sample data set includes the driving distraction state ( ), the speed of the host vehicle ( ), the speed of the leading vehicle ( ), the speed of the adjacent vehicle ( ), the distance between the host vehicle and the leading vehicle ( ), the longitudinal distance between the host vehicle and the adjacent vehicle ( ), and the lateral distance between the host vehicle and the adjacent vehicle ( ), a total of 7 features. Among them, the driving distraction state comes from the recognition result of the driving distraction recognition model constructed based on SVM in step 3.3 (driving distraction is recorded as 1, and the driver is not distracted is recorded as 0), and before the samples are used to train the model, each feature is standardized. The prediction variable is the probability value of severe conflict events and collision events, and the samples with probability values exceeding the threshold are predicted as positive class samples. And during the training process, the standardized sample data set is divided into a training set and a test set in a ratio of 7:3.

[0054] It should be noted that the present invention defines the time from the trigger moment of the risk event to the conflict critical point (when TTC is first lower than the threshold) as the conflict trigger time , to the most severe conflict moment (the collision event is the moment of collision) as the risk peak trigger time , the model prediction moment is , to the most severe conflict moment as the model prediction time , as shown in Figure 3 .

[0055] Step 4.2, build a severe conflict and collision prediction model based on LSTM, set the model hyperparameters, and implement it in Python based on the TensorFlow framework.

[0056] Specifically, after optimizing the grid parameters, the model hyperparameters are set. During the training process, since the prediction task is a binary classification problem, the loss function is set to the Binary Cross Entropy Loss Function.

[0057] The severe conflict and collision prediction model of the embodiment of the present invention inputs data in a time series form, obtains time series features after passing through the LSTM layer, and then obtains the final classification probability prediction value after passing through the Dropout and Dense layers. The prediction value is compared with a preset threshold. If it is greater than the threshold, it is determined as a positive class sample; otherwise, it is a negative class sample.

[0058] Step 4.3: Evaluate the effect of the severe conflict and collision prediction model considering the driving distraction state characteristics. By comparing the prediction effects of the two models, verify the important contribution of the driving distraction state characteristics to the prediction model.

[0059] It should be noted that the severe conflict and collision prediction model based on LSTM is trained through the TensorFlow deep learning framework. After training, the model converges, and no overfitting occurs during the training process. Subsequently, with the same model hyperparameters, the driving distraction state characteristics in the samples are removed, and a baseline model is trained for comparative analysis. Since the category weights are set in the loss function in step 4.2, the subsequent category segmentation threshold is set to 0.7. When the model prediction value is greater than or equal to 0.7, it is predicted as a positive class sample; when it is less than 0.7, it is predicted as a negative class sample.

[0060] The experimental results of the present invention show that for severe conflict events, severe traffic conflicts are predicted on average 5.05 s before the most severe moment of the conflict; while for collision events, collisions are predicted on average 4 s before the most severe moment of the conflict. Therefore, it can play a role in preventing the occurrence of severe traffic conflicts and collision events, proving the effectiveness of the method of the present invention.

[0061] In the embodiment of the present invention, the built model improves the prediction accuracy for various types of events. In particular, it is very sensitive to the prediction of severe conflict and collision events, meeting the requirements of active prevention and control of traffic safety. At the same time, the false alarm probability is also lower than that of the baseline model, and it will not affect the driver's mood due to excessive false identifications. This shows the important contribution of the driving distraction state characteristics to the model prediction effect.

[0062] Based on the model constructed by the present invention above, in practical applications, the method for predicting severe conflicts and collisions considering driving distraction of the present invention includes: Obtain the driving data of the vehicle; the driving data includes: the speed of the host vehicle, the speed of the leading vehicle, the speed of the adjacent vehicle, the distance between the host vehicle and the leading vehicle, the longitudinal distance between the host vehicle and the adjacent vehicle, the lateral distance between the host vehicle and the adjacent vehicle, and the steering wheel angle in the driving cockpit; Input the driving data into the driving distraction recognition model to obtain the driving distraction state characteristics; Input the driving distraction state characteristics and the driving data into the severe conflict and collision prediction model to predict severe conflict events and collision events, and output the prediction results; among them, the class labels in the severe conflict and collision prediction model include non-interference passing event, minor conflict event, general conflict event, severe conflict event, and collision event.

[0063] The following is an apparatus embodiment of the present invention, which can be used to execute the method embodiment of the present invention. For the details not disclosed in the apparatus embodiment, please refer to the method embodiment of the present invention.

[0064] In another embodiment of the present invention, a severe conflict and collision prediction system considering driving distraction is provided, which is characterized by including: A data acquisition module for obtaining the driving data of the vehicle; A driving distraction recognition module for inputting the driving data into the driving distraction recognition model to obtain the driving distraction state characteristics; A prediction module for inputting the driving distraction state characteristics and the driving data into the severe conflict and collision prediction model to perform severe conflict and collision prediction and output the prediction results; among them, the class labels in the severe conflict and collision prediction model include non-interference passing, minor conflict, general conflict, severe conflict, and collision.

[0065] In another embodiment of the present invention, a computer device is provided. The computer device includes a processor and a memory. The memory is used to store a computer program, and the computer program includes program instructions. The processor is used to execute the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, and is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function. The processor described in the embodiment of the present invention can be used for the operation of the method for predicting serious conflicts and collisions considering driving distraction.

[0066] In another embodiment of the present invention, a storage medium is also provided, specifically a computer-readable storage medium (Memory). The computer-readable storage medium is a memory device in the computer device and is used to store programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and, of course, the extended storage medium supported by the computer device. The computer-readable storage medium provides a storage space, and the operating system of the terminal is stored in this storage space. And, one or more instructions suitable for being loaded and executed by the processor are also stored in this storage space. These instructions may be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. One or more instructions stored in the computer-readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the method for predicting serious conflicts and collisions considering driving distraction in the above embodiments.

[0067] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0068] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0069] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing devices to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0070] These computer program instructions can also be loaded onto a computer or other programmable data processing devices, such that a series of operation steps are executed on the computer or other programmable devices to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable devices provide steps for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: still can modify the specific implementation manners of the present invention or make equivalent replacements, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the protection scope of the claims of the present invention.

Claims

1. A method for predicting severe conflicts and collisions considering driving distraction, characterized in that Including: Obtain the driving data of the vehicle; Input the driving data into the driving distraction recognition model to identify the driving distraction state characteristics; Input the driving distraction state characteristics and the driving data into the severe conflict and collision prediction model to predict severe conflict events and collision events, and output the prediction results; wherein, the class labels in the severe conflict and collision prediction model include non-interference passing events, minor conflict events, general conflict events, severe conflict events, and collision events; Among them, the driving distraction recognition model is constructed based on a support vector machine, and the construction method includes: Construct different driving scenarios through a driving simulator to obtain the driving performance data of the driver under different scenarios; Using the driving performance data as samples, calculate the mean and standard deviation as time-domain features through statistical methods, extract the low-frequency energy values of the driving performance variables through fast Fourier transform as frequency-domain features, and screen the time-domain features and frequency-domain features through non-parametric test methods; Standardize the selected features and use them as the features of the samples to jointly constitute the input of the driving distraction recognition model. Based on the support vector machine, build the driving distraction recognition model and train the driving distraction recognition model using the K-fold cross-validation method.

2. The method for predicting severe conflicts and collisions considering driving distraction according to claim 1, wherein The driving data includes: the speed of the host vehicle, the speed of the leading vehicle, the speed of the adjacent vehicle, the distance between the host vehicle and the leading vehicle, the longitudinal distance between the host vehicle and the adjacent vehicle, the lateral distance between the host vehicle and the adjacent vehicle, and the steering wheel angle in the driving cockpit.

3. The method for predicting severe conflicts and collisions considering driving distraction according to claim 1, wherein The selected features include five features: the mean value of the lane departure amount, the standard deviation of the steering wheel angle, the standard deviation of the lane departure amount, the low-frequency energy value of the steering wheel angle, and the low-frequency energy value of the lane departure amount.

4. The method for predicting severe conflicts and collisions considering driving distraction according to claim 1, wherein The construction method of the severe conflict and collision prediction model includes: Input the driving performance data into the driving distraction recognition model to identify the driving distraction state characteristics; According to the driving performance data, identify the interaction mode between the surrounding vehicles and the host vehicle through lane subordination relationship, initially screen conflict events in different interaction modes based on traffic conflict recognition indicators, and then combine the evasive behavior indicators to identify longitudinal traffic conflicts and lateral traffic conflicts of the initially screened conflict events to obtain conflict events; Based on the traffic conflict severity index, perform clustering analysis on the conflict events using the K-means++ clustering algorithm to obtain the clustering results; wherein, the categories of the clustering results include non-interference passing events, minor conflict events, general conflict events, severe conflict events, and collision events; Build a severe conflict and collision prediction model based on LSTM, form a sample data set with the conflict events, the corresponding clustering results, and the driving distraction state characteristics, and use the probability values of severe conflict events and collision events as prediction variables to train the severe conflict and collision prediction model.

5. The method for predicting severe conflicts and collisions considering driving distraction according to claim 4, characterized in that The traffic conflict recognition indicator is the time to collision; the traffic conflict severity indicators include the exposure time to collision, the cumulative critical time to collision, and the deceleration for collision avoidance.

6. A serious conflict and collision prediction system considering driving distraction, characterized in that, Including: A data acquisition module for obtaining the driving data of the vehicle; A driving distraction recognition module for inputting the driving data into the driving distraction recognition model to identify the driving distraction state characteristics; A prediction module, configured to input driving distraction state features and driving data into a severe conflict and collision prediction model to predict severe conflicts and collisions and output prediction results; wherein, the class labels in the severe conflict and collision prediction model include non-interference passing events, minor conflict events, general conflict events, severe conflict events, and collision events. Wherein, the driving distraction recognition model is constructed based on a support vector machine, and the construction method includes: Constructing different driving scenarios through a driving simulator to obtain the driving performance data of the driver under different scenarios; Using the driving performance data as samples, calculating the mean and standard deviation as time-domain features through statistical methods, extracting the low-frequency energy values of the driving performance variables through fast Fourier transform as frequency-domain features, and screening the time-domain features and frequency-domain features through non-parametric test methods; Standardizing the screened features and using them as the features of the samples to jointly form the input of the driving distraction recognition model. Based on the support vector machine, build a driving distraction recognition model and train the driving distraction recognition model in the way of K-fold cross-validation.

7. The serious conflict and collision prediction system considering driving distraction according to claim 6, characterized in that, The driving data includes: the speed of the host vehicle, the speed of the leading vehicle, the speed of the adjacent vehicle, the distance between the host vehicle and the leading vehicle, the longitudinal distance between the host vehicle and the adjacent vehicle, the lateral distance between the host vehicle and the adjacent vehicle, and the steering wheel angle in the driving cockpit.

8. The severe conflict and collision prediction system considering driving distraction according to claim 6, characterized in that, The screened features include five features: the mean value of the lane departure amount, the standard deviation of the steering wheel angle, the standard deviation of the lane departure amount, the low-frequency energy value of the steering wheel angle, and the low-frequency energy value of the lane departure amount.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the severe conflict and collision prediction method considering driving distraction according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the severe conflict and collision prediction method considering driving distraction according to any one of claims 1 to 7.

Images