Vehicle abnormal driving behavior recognition method and device and electronic equipment
Through the classification and clustering of vehicle motion state characteristics, combined with the hidden Dirichlet distribution model, the abnormal driving behavior of vehicles is identified, which solves the problem of lack of in-depth exploration in the existing technology and improves the decision-making ability of intelligent connected vehicles.
Patent Information
- Application Number
- CN202510023896.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-07
- Publication Date
- 2025-04-25
AI Technical Summary
The lack of in-depth and multi-perspective exploration of abnormal vehicle driving behavior in the prior art makes it difficult to establish an effective model of abnormal vehicle driving behavior, limiting the decision-making ability of intelligent connected vehicles.
Vehicle motion state feature classification and clustering are used to construct vehicle motion state words, use the implicit Dirichrey distribution model for training, identify vehicle driving behavior, and feature extraction and recognition are performed through bag-of-word model and labeled implicit Dirichrey distribution model.
It realizes accurate identification of abnormal driving behavior of vehicles, provides management solutions for intelligent traffic management systems, and improves the decision-making capabilities of intelligent connected vehicles.
Smart Images
Figure CN119884829A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent transportation and automobile management technology, and in particular to a method, device and electronic equipment for identifying abnormal driving behavior of a vehicle. Background Art
[0002] The development of autonomous driving technology has brought people a new way of travel, but the performance limitations of on-board perception sensors have led to problems such as missing and inaccurate perception information of intelligent connected vehicles, making it difficult to analyze and predict the motion status of surrounding vehicles, which may still cause problems when intelligent connected vehicles make decisions, thereby limiting the implementation of intelligent connected vehicles. The vehicle-road collaborative environment can provide vehicles with broad and multi-dimensional perception capabilities, make up for the shortcomings of on-board perception, and provide new ideas for the implementation of autonomous driving. However, it also has problems such as difficult spatiotemporal synchronization of traffic scene data, data heterogeneity, and high complexity, making it difficult to analyze and predict the vehicle motion status and provide support for vehicle decision-making. Deep learning and artificial intelligence technologies can effectively extract the motion status characteristics of surrounding vehicles and identify vehicle driving behaviors, realize vehicle motion trajectory prediction, and assist in the decision-making of intelligent connected vehicles.
[0003] Reliable accident prediction and proactive prevention are undoubtedly very beneficial and necessary. The occurrence of accidents is a complex mechanism with many contributing factors. In general, driver-centric factors can be found in most collision accidents, and vehicle driving behavior assessment is an important aspect of improving safety and reducing collisions. The classification and identification detection of abnormal driving behaviors is a long-term exploration. A series of features have been developed to describe unsafe behaviors, such as speeding, sudden braking or jerkiness, rear-end collisions, frequent and drastic lane changes, yaw, etc. Typically, statistical distribution maps of motion-related variables are used as behavioral features. In order to conduct reliable driving evaluation, in-depth and multi-perspective mining of features is required, especially predictive feature mining. Summary of the invention
[0004] In view of this, it is necessary to provide a method, device and electronic device for identifying abnormal driving behavior of a vehicle to solve the technical problem that the prior art does not conduct in-depth and multi-perspective mining of features to establish a model of abnormal driving behavior of a vehicle.
[0005] In order to solve the above problems, the present invention provides a method for identifying abnormal driving behavior of a vehicle, comprising: Classify and cluster vehicle motion state features to generate vehicle motion state terms; Determine the feature range of the vehicle motion state aggregation sequence according to the classification standard of the vehicle motion state feature, and construct the vehicle motion state aggregation sequence according to the vehicle motion state words and the feature range; Taking the aggregate sequence of vehicle motion states as input, training the initial latent Dirichlet distribution model to obtain a fully trained latent Dirichlet distribution model; The acquired real-time vehicle motion state words are input into a well-trained implicit Dirichlet distribution model to obtain the real-time vehicle motion state, and the vehicle driving behavior is determined according to the real-time vehicle motion state.
[0006] In a possible implementation, the vehicle motion state features include speed, acceleration, heading angle, lane deviation and obstacle conditions; and classifying and clustering the vehicle motion state features to generate vehicle motion state terms includes: The vehicle motion state features are encoded based on the preset Ikebukuro model to obtain vehicle motion state terms corresponding to the vehicle motion state features.
[0007] In a possible implementation, determining the feature range of the vehicle motion state aggregation sequence according to the classification standard of the vehicle motion state feature includes: The characteristic range of the vehicle motion state aggregation sequence is determined according to the quantitative index corresponding to each vehicle motion state feature.
[0008] In a possible implementation, before training the initial implicit Dirichlet distribution model, the method further includes: defining a vehicle driving behavior, wherein the vehicle driving behavior includes abnormal speed, abnormal acceleration, abnormal lane change, and no abnormality; The association between the vehicle motion state aggregation sequence and the vehicle driving behavior is constructed based on the label projection matrix.
[0009] In a possible implementation, the initial implicit Dirichlet distribution model is trained using the vehicle motion state aggregation sequence as input to obtain a fully trained implicit Dirichlet distribution model, including: Construct an initial labeled latent Dirichlet distribution model; According to the correlation between the aggregated sequence of vehicle motion states and vehicle driving behavior, the initial latent Dirichlet distribution model is trained to obtain a fully trained latent Dirichlet distribution model.
[0010] In a possible implementation, the training of the initial implicit Dirichlet distribution model using the aggregated sequence of vehicle motion states as input further includes: The preset Gibbs sampling is used to train and sample the training data of the vehicle motion state aggregation sequence.
[0011] In a possible implementation, determining the vehicle driving behavior according to the real-time vehicle motion state includes: Based on the correlation between the vehicle motion state aggregation sequence and the vehicle driving behavior, the vehicle driving behavior is determined according to the real-time vehicle motion state.
[0012] In a second aspect, the present invention further provides a vehicle abnormal driving behavior recognition device, comprising: The first generation module is used to classify and cluster the vehicle motion state features to generate vehicle motion state terms; An aggregate sequence construction module, used to determine the feature range of the vehicle motion state aggregate sequence according to the classification standard of the vehicle motion state feature, and construct the vehicle motion state aggregate sequence according to the vehicle motion state words and the feature range; A training module, used for training the initial implicit Dirichlet distribution model with the aggregate sequence of vehicle motion states as input to obtain a fully trained implicit Dirichlet distribution model; The driving behavior determination module is used to input the acquired real-time vehicle motion state words into a well-trained implicit Dirichlet distribution model to obtain the real-time vehicle motion state, and determine the vehicle driving behavior according to the real-time vehicle motion state.
[0013] In a third aspect, the present invention further provides an electronic device, comprising: a processor and a memory; The memory stores a computer-readable program executable by the processor; When the processor executes the computer-readable program, the steps in the vehicle abnormal driving behavior identification method as described above are implemented.
[0014] In a fourth aspect, the present invention also provides a computer-readable storage medium, which stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the vehicle abnormal driving behavior identification method as described above.
[0015] The beneficial effects of the present invention are as follows: first, a bag-of-words model of vehicle motion data is constructed to mine features and identify vehicle driving behaviors in vehicle motion state sequences. The driving behaviors are calibrated with prior labels based on the historical event database, and the characteristic characterization parameters of the driving behaviors are extracted; according to the continuous information of all elements of the driving scene, the current driving behaviors of the vehicles are encoded through the bag-of-words model, and a vehicle driving behavior recognition model based on a labeled latent Dirichlet distribution is constructed to recognize abnormal vehicle driving behaviors such as abnormal speed, abnormal acceleration and deceleration, and abnormal lane change. The labeled latent Dirichlet distribution model is used to identify the subject and determine which class label the current driving behavior belongs to. This is to identify abnormal vehicle driving behaviors on urban roads, identify vehicles for intelligent traffic management systems, and provide management solutions. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 A method flow chart of an embodiment of a method for identifying abnormal driving behavior of a vehicle provided by the present invention; Figure 2 A schematic diagram of a vehicle motion state data aggregation model based on a bag-of-words model in the vehicle abnormal driving behavior identification method provided by the present invention; Figure 3 A schematic diagram of a labeled latent Dirichlet allocation (LLDA) model in the vehicle abnormal driving behavior identification method provided by the present invention; Figure 4 is a schematic diagram of an embodiment of a vehicle abnormal driving behavior identification device provided by the present invention; Figure 5 It is a schematic diagram of the operating environment of an electronic device provided by the present invention. DETAILED DESCRIPTION
[0017] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.
[0018] A specific embodiment of the present invention discloses a method for identifying abnormal driving behavior of a vehicle. Figure 1 ,include: S101, classifying and clustering vehicle motion state features to generate vehicle motion state terms; S102, determining a feature range of a vehicle motion state aggregation sequence according to a classification standard of vehicle motion state features, and constructing a vehicle motion state aggregation sequence according to the vehicle motion state words and the feature range; S103, using the vehicle motion state aggregation sequence as input, training the initial implicit Dirichlet distribution model to obtain a fully trained implicit Dirichlet distribution model; S104: input the acquired real-time vehicle motion state words into a well-trained implicit Dirichlet distribution model to obtain the real-time vehicle motion state, and determine the vehicle driving behavior according to the real-time vehicle motion state.
[0019] In this embodiment, a bag-of-words model of vehicle motion data is first constructed to mine features and identify vehicle driving behaviors in vehicle motion state sequences. The driving behaviors are calibrated with prior labels based on the historical event database, and the characteristic characterization parameters of the driving behaviors are extracted. According to the continuous information of all elements of the driving scene, the current driving behaviors of the vehicles are encoded through the bag-of-words model, and a vehicle driving behavior recognition model based on a labeled latent Dirichlet distribution is constructed to identify abnormal vehicle driving behaviors such as abnormal speed, abnormal acceleration and deceleration, and abnormal lane change. The labeled latent Dirichlet distribution model is used to identify the subject and determine which class label the current driving behavior belongs to. This is to identify abnormal vehicle driving behaviors on urban roads, identify vehicles for the intelligent traffic management system, and provide management solutions.
[0020] In one embodiment, the vehicle motion state features include speed, acceleration, heading angle, lane deviation and obstacle conditions; the vehicle motion state features are classified and clustered to generate vehicle motion state terms, including: The vehicle motion state features are encoded based on the preset Ikebukuro model to obtain vehicle motion state terms corresponding to the vehicle motion state features.
[0021] In this embodiment, the vehicle motion state features including the information about the vehicle and the surrounding environment are analyzed as words, the vehicle motion state sequence within a period of time is taken as a document, and the corresponding vehicle driving behavior is taken as the subject. After encoding the vehicle motion state data using the bag-of-words model, its vehicle motion state terms are obtained. Based on this setting, the vehicle motion state terms can be extracted from the original driving data and the driving environment scene feature data, and the corresponding vehicle driving behavior can be identified therefrom, thereby completing the identification of the vehicle driving behavior. However, the original data on the vehicle motion state is often complex and high-dimensional, so a vehicle motion and driving environment data aggregation model based on the bag-of-words model is proposed to reduce the dimensionality of the original data, please refer to Figure 2 .
[0022] In a specific embodiment, Figure 2 As shown, the characteristics related to the vehicle's motion state during six types of vehicle driving behaviors are taken into account, and the driving environment characteristics of lane line deviation and obstacles that affect the vehicle's motion state are also taken into account. In this embodiment, the bag-of-words model defines six characteristics related to vehicle motion and driving environment, including speed, acceleration, heading angle, lane line deviation, and surrounding obstacles. Although speed and acceleration features are generally displayed as continuous data during the acquisition process, for the process of continuously changing vehicle motion states over a long period of time, it is difficult to identify vehicle driving behaviors based on the vehicle motion state data at one moment through rule-based word analysis. In this article, Each feature in the vehicle motion state data is set to a range as a cluster for representation. Figure 2 As shown in: The six characteristics indicated by the light blue color; Yellow means that the feature is divided into two groups, such as acceleration and deceleration. Light green represents a range of clusters. Taking acceleration as an example, it is divided into very low, low, medium, high and very high; speed is divided into very low speed, low speed, medium speed, high speed and very high speed; According to the heading angle, it is divided into left deviation, no deviation and right deviation; Lane deviation is divided into three types: left, center and right; lane information is divided into 1 lane, 2 lanes, 3 lanes, etc.; for obstacle types, pedestrians, vehicles and non-motor vehicles around the vehicle are all regarded as obstacles and encoded, divided into near, medium and far distances according to distance, and divided into eight types according to eight directions relative to the vehicle's position. The type of obstacle in a certain direction can be represented by a binary state of existence or non-existence, so that when there are multiple targets, only existence can be used to represent it.
[0023] After encoding the vehicle motion state feature data with a bag-of-words model based on the above classification, the original feature sequence of the vehicle driving data is processed with dimensionality reduction to obtain the vehicle motion state words that have been analyzed and judged. The vehicle motion state words are combined into a sequence to represent the vehicle's lane deviation, obstacle conditions, acceleration, speed, and heading angle. For example, "centering in front; slow deceleration; low speed; no deviation; lane 1" is used to further represent the vehicle motion state.
[0024] In a specific embodiment, vehicle motion state characteristics based on the bag-of-words model are shown in Table 1 and Table 2.
[0025] Table 1: Example of original feature sequence of vehicle driving data
[0026] Table 2: Example of vehicle motion state word sequence
[0027] In some embodiments, determining the feature range of the vehicle motion state aggregation sequence according to the classification standard of the vehicle motion state feature includes: The characteristic range of the vehicle motion state aggregation sequence is determined according to the quantitative index corresponding to each vehicle motion state feature.
[0028] In this embodiment, as shown in Table 3, Table 3 gives the range of each feature representation, so as to achieve clustering and dimensionality reduction of vehicle driving behavior. Since the time window is used to capture the vehicle motion state, the offset, signal and obstacle conditions are all used within the time window, the speed is analyzed by the average quantity, and the acceleration is calculated by the following formula.
[0029]
[0030] Table 3: Cluster range of vehicle motion state features
[0031] Based on the above clustering range, the vehicle motion state is extracted from the vehicle motion data and driving environment data in the vehicle-road cooperative perception environment, and a bag-of-words model based on vehicle motion feature-driving environment feature elements is constructed as the data analysis basis for the subsequent latent Dirichlet distribution model.
[0032] In some embodiments, before training the initial latent Dirichlet distribution model, the method further includes: defining a vehicle driving behavior, wherein the vehicle driving behavior includes abnormal speed, abnormal acceleration, abnormal lane change, and no abnormality; The association between the vehicle motion state aggregation sequence and the vehicle driving behavior is constructed based on the label projection matrix.
[0033] In this embodiment, in order to obtain the vehicle driving behavior corresponding to the vehicle motion state sequence, it is necessary to train the model to calculate the multinomial distribution of each variable. The embodiment of the present invention uses Gibbs sampling to train and sample the training set data, and calculates the expected value of their multinomial distribution. First, it is necessary to calculate the prior distribution relationship between the vehicle motion state sequence and their labels, and use the Bernoulli coin tossing method to sample the vehicle driving behavior k and the prior distribution Φk of the label. The label projection matrix L(m) corresponding to each vehicle motion state sequence m is a matrix of size M × K. This matrix is defined as follows: the element i∈{1, . . ., M} of each row refers to the vehicle driving behavior distribution constraint corresponding to the vehicle motion state sequence corresponding to each vehicle driving behavior, and the element j∈{1, . .., K} of each column represents the vehicle motion state sequence constraint corresponding to the vehicle driving behavior. The matrix is assigned values by the following formula:
[0034] Where λ is the value sampled from one of the elements, and the Dirichlet prior distribution α= (α1. . . αK) T of vehicle driving behavior is projected from the label matrix L(m) to a lower-dimensional vector α(d), which is obtained by the following formula:
[0035] In this way, the corresponding vehicle driving behaviors k are constrained by their corresponding labels, giving a label vector Λ(m). The prior distribution Φk of the label is separated from the rest of the model and the distribution is obtained separately.
[0036] Among them, based on the proposed extraction model of vehicle motion state words, the vehicle driving behavior can be well defined from the vehicle motion state sequence, and the relationship between its vehicle driving behavior and each vehicle motion state can be mined.
[0037] The implicit Dirichlet distribution model can regard the vehicle driving behavior as an implicit subject. Each vehicle motion state sequence has a corresponding vehicle driving behavior, and the vehicle motion state is the basic element that constitutes the vehicle motion state sequence.
[0038] The typical Latent Dirichlet Allocation (LDA) model is an unsupervised text processing model without prior knowledge. The Labeled Latent Dirichlet Allocation (LLDA) model is a typical supervised Latent Dirichlet Allocation model. It contains the one-to-one connection between words and implicit topics in the traditional LDA model, and links implicit topics with their corresponding prior labels. The original LLDA model successfully discovered restricted topics from labeled documents. Unlike the unmodified LDA model, the implicit topics are learned from supervised labels, so they can be learned from the rich information of the labels. Table 4 explains the meaning of the mathematical symbols used in the LLDA model. In this model, according to the possible vehicle driving behaviors in the data set, the vehicle driving behaviors to be identified are divided into four categories, namely "speed abnormality", "acceleration abnormality", "lane change abnormality", and "no abnormality". In this embodiment, the vehicle driving behavior is restricted by the vehicle driving behavior corresponding to the prior related label set obtained based on experience and then learned. Figure 3As shown in the figure, each vehicle driving behavior m is represented as a vehicle motion state index list w(m)= (w1, …, wNm) and a corresponding binary topic indicator Λ(m)= (l1, …, lk), where wi ∈ {1, …, V} and lk ∈ {0, 1}. For the generation process, unlike the traditional LDA model, this model can obtain a multinomial distribution for the label Λ(m) for the vehicle driving behavior, which is used to restrict the distribution of each vehicle motion state sequence for the corresponding vehicle driving behavior.
[0039] Table 4 Meaning of mathematical symbols used in the vehicle driving behavior recognition model based on LLDA
[0040] Furthermore, based on the relationship between the distribution variables of the supervised latent Dirichlet distribution model and introducing the constraint relationship between the label and the vehicle driving behavior, the embodiment of the present invention uses Gibbs sampling to train the model and gives a probability sampling formula for the vehicle driving behavior distribution z of the vehicle motion state sequence m in the LLDA model, such as the formula:
[0041] The symbol definitions in the formula are given in Table 4, which will be used for Gibbs sampling in this study and finally obtain the distribution of vehicle driving behavior on the original vehicle motion data and the given labels, where represents the number of vehicle motion states wi in vehicle driving behavior j, but does not include the vehicle motion state currently assigned to vehicle driving behavior z. This allows the parameters of each distribution and statistical variable to be initialized, and after multiple rounds of iterative Gibbs sampling, the corresponding vehicle driving behaviors and their vehicle motion states in each vehicle state sequence are finally obtained, and the vehicle driving behavior corresponding to the vehicle motion state sequence can be identified later.
[0042] Based on the above-mentioned vehicle abnormal driving behavior recognition method, the embodiment of the present invention also provides a vehicle abnormal driving behavior recognition device, please refer to Figure 4 ,include: The first generating module 410 is used to classify and cluster the vehicle motion state features to generate vehicle motion state terms; The aggregate sequence construction module 420 is used to determine the feature range of the vehicle motion state aggregate sequence according to the classification standard of the vehicle motion state feature, and construct the vehicle motion state aggregate sequence according to the vehicle motion state words and the feature range; A training module 430 is used to train the initial implicit Dirichlet distribution model using the vehicle motion state aggregation sequence as input to obtain a fully trained implicit Dirichlet distribution model; The driving behavior determination module 440 is used to input the acquired real-time vehicle motion state words into the well-trained implicit Dirichlet distribution model to obtain the real-time vehicle motion state, and determine the vehicle driving behavior according to the real-time vehicle motion state.
[0043] like Figure 5 As shown, based on the above-mentioned vehicle abnormal driving behavior recognition method, the present invention also provides an electronic device, which can be a computing electronic device such as a mobile terminal, a desktop computer, a notebook, a palm computer, and a server. The electronic device includes a processor 510, a memory 520, and a display 530. Figure 5 Only some components of the electronic device are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.
[0044] In some embodiments, the memory 520 may be an internal storage unit of the electronic device, such as a hard disk or memory of the electronic device. In other embodiments, the memory 520 may also be an external storage electronic device of the electronic device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Further, the memory 520 may also include both an internal storage unit of the electronic device and an external storage electronic device. The memory 520 is used to store application software and various types of data installed in the electronic device, such as program codes for installing the electronic device. The memory 520 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a vehicle abnormal driving behavior recognition program 540 is stored on the memory 520, and the vehicle abnormal driving behavior recognition program 540 can be executed by the processor 510, thereby realizing the vehicle abnormal driving behavior recognition method of each embodiment of the present application.
[0045] In some embodiments, the processor 510 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run program codes or process data stored in the memory 520, such as executing a method for identifying abnormal driving behavior of a vehicle.
[0046] In some embodiments, the display 530 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 530 is used to display information in the vehicle abnormal driving behavior recognition electronic device and to display a visual user interface. The components 510-530 of the electronic device communicate with each other via a system bus.
[0047] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.
[0048] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.
Claims
1. A method for identifying abnormal driving behavior of a vehicle, characterized in that: include: Classify and cluster vehicle motion state features to generate vehicle motion state terms; Determine the feature range of the vehicle motion state aggregation sequence according to the classification standard of the vehicle motion state feature, and construct the vehicle motion state aggregation sequence according to the vehicle motion state words and the feature range; Taking the aggregate sequence of vehicle motion states as input, training the initial latent Dirichlet distribution model to obtain a fully trained latent Dirichlet distribution model; The acquired real-time vehicle motion state words are input into a well-trained implicit Dirichlet distribution model to obtain the real-time vehicle motion state, and the vehicle driving behavior is determined according to the real-time vehicle motion state.
2. The method for identifying abnormal vehicle driving behavior according to claim 1, characterized in that: The vehicle motion state features include speed, acceleration, heading angle, lane deviation and obstacle conditions; the vehicle motion state features are classified and clustered to generate vehicle motion state terms, including: The vehicle motion state features are encoded based on the preset Ikebukuro model to obtain vehicle motion state terms corresponding to the vehicle motion state features.
3. The method for identifying abnormal vehicle driving behavior according to claim 1, characterized in that: The step of determining the feature range of the vehicle motion state aggregation sequence according to the classification standard of the vehicle motion state feature includes: The characteristic range of the vehicle motion state aggregation sequence is determined according to the quantitative index corresponding to each vehicle motion state feature.
4. The method for identifying abnormal vehicle driving behavior according to claim 1, characterized in that: Before training the initial latent Dirichlet distribution model, the method further includes: defining a vehicle driving behavior, wherein the vehicle driving behavior includes abnormal speed, abnormal acceleration, abnormal lane change, and no abnormality; The association between the vehicle motion state aggregation sequence and the vehicle driving behavior is constructed based on the label projection matrix.
5. The method for identifying abnormal vehicle driving behavior according to claim 4, characterized in that: The initial implicit Dirichlet distribution model is trained by taking the vehicle motion state aggregation sequence as input to obtain a fully trained implicit Dirichlet distribution model, including: Construct an initial labeled latent Dirichlet distribution model; According to the correlation between the aggregated sequence of vehicle motion states and vehicle driving behavior, the initial latent Dirichlet distribution model is trained to obtain a fully trained latent Dirichlet distribution model.
6. The method for identifying abnormal vehicle driving behavior according to claim 4, characterized in that: The method of training the initial implicit Dirichlet distribution model using the vehicle motion state aggregation sequence as input also includes: The preset Gibbs sampling is used to train and sample the training data of the vehicle motion state aggregation sequence.
7. The method for identifying abnormal vehicle driving behavior according to claim 4, characterized in that: Determining the vehicle driving behavior according to the real-time vehicle motion state includes: Based on the correlation between the vehicle motion state aggregation sequence and the vehicle driving behavior, the vehicle driving behavior is determined according to the real-time vehicle motion state.
8. A vehicle abnormal driving behavior recognition device, characterized in that: include: The first generation module is used to classify and cluster the vehicle motion state features to generate vehicle motion state terms; An aggregate sequence construction module, used to determine the feature range of the vehicle motion state aggregate sequence according to the classification standard of the vehicle motion state feature, and construct the vehicle motion state aggregate sequence according to the vehicle motion state words and the feature range; A training module, used for training the initial implicit Dirichlet distribution model with the aggregate sequence of vehicle motion states as input to obtain a fully trained implicit Dirichlet distribution model; The driving behavior determination module is used to input the acquired real-time vehicle motion state words into a well-trained implicit Dirichlet distribution model to obtain the real-time vehicle motion state, and determine the vehicle driving behavior according to the real-time vehicle motion state.
9. An electronic device, characterized in that: include: Processor and memory; The memory stores a computer-readable program executable by the processor; When the processor executes the computer-readable program, the steps in the vehicle abnormal driving behavior identification method as described in any one of claims 1-7 are implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the vehicle abnormal driving behavior identification method as described in any one of claims 1-7.