Motorcycle riding dangerous behavior identification method and system based on feature classification

Abstract

The invention discloses a motorcycle riding dangerous behavior identification method and system based on feature classification, and the method comprises the steps: obtaining the historical riding data of a user from a database and a label value indicating whether the riding is dangerous or not, and cleaning the historical riding data of the user; aligning time steps of the dominant riding characteristic data and the implicit riding characteristic data; constructing and training a motorcycle dangerous riding behavior recognition model; and carrying out cleaning and time step alignment processing on the riding behavior data of the user to be identified, and inputting the data into the trained motorcycle dangerous riding behavior identification model to obtain an identification result of whether the riding behavior is a dangerous riding behavior or not. According to the method, potential information contained in various characteristics of the riding behavior of the motorcycle user is fully mined, the time dependence of the riding behavior of the motorcycle user is captured in combination with deep learning, more reliable analysis and prediction are provided for a complex driving scene, potential dangerous driving behaviors of the motorcycle can be effectively identified, and the driving experience of the motorcycle is improved. And the riding safety of a motorcycle user is ensured.

Description

Technical Field

[0001] The present invention relates to the field of motorcycle user riding behavior recognition, and in particular to a method and system for identifying dangerous motorcycle riding behaviors based on feature classification. Background Art

[0002] The rapid development of intelligent transportation and intelligent driving technologies has provided important technical support for improving user driving safety and driving experience at multiple levels. As a facility that combines multiple functions such as commuting and leisure and entertainment, motorcycles are becoming more and more popular among consumers. However, due to the complexity of its operation and the dangers in specific driving environments, motorcycle driving accidents occur frequently. Applying intelligent driving technology to analyze user driving behavior, identify user driving behavior patterns and make predictions to respond and ensure user driving safety has become a research hotspot in the motorcycle industry.

[0003] However, the application of intelligent driving technology in the motorcycle field started late, mostly relying on simple statistical analysis or traditional machine learning methods, which makes it difficult to capture the dynamic changes in the user's riding behavior pattern and its time dependency during riding. In addition, the current analysis of user riding behavior in the motorcycle field is mainly based on relatively simple riding features, such as riding speed, lane change, and following a car, without considering the unique complexity of motorcycle riding operations, resulting in the failure to fully tap the potential information contained in the user's riding feature data, resulting in poor final technology application results. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention proposes a method and system for identifying dangerous motorcycle riding behaviors based on feature classification.

[0005] The specific technical solutions are as follows:

[0006] A method for identifying dangerous motorcycle riding behaviors based on feature classification comprises the following steps:

[0007] S1: Obtain the user's historical riding data and the label value of whether each riding segment is dangerous riding from the database, and clean the user's historical riding data; the user's historical riding data includes explicit riding feature data and implicit riding feature data. The explicit riding feature refers to the feature that reflects the user's real-time riding status through the data at a single time point, and the implicit riding feature refers to the feature whose meaning cannot be directly observed through the data at a single time point, and it is necessary to combine the time series to mine the potential information;

[0008] S2: The time step of aligning the explicit riding feature data and the implicit riding feature data obtained in S1;

[0009] S3: Use the riding feature data obtained in S2 to build and train a motorcycle dangerous riding behavior recognition model, the motorcycle dangerous riding behavior recognition model includes: an explicit riding feature analysis module based on a fully connected neural network, an implicit riding feature analysis module based on an LSTM network, and a riding feature fusion and classification module; the explicit riding feature analysis module is used to model and analyze the explicit riding features and output a feature vector representing the explicit riding features, the implicit riding feature analysis module is used to model and analyze the implicit riding features and output a feature vector representing the implicit riding features, the riding feature fusion and classification module is used to splice and fuse the feature vector representing the explicit riding features with the feature vector representing the implicit riding features to obtain a feature vector representing the user's overall riding behavior, and classify the user's riding behavior as a dangerous riding behavior based on the feature vector of the user's overall riding behavior;

[0010] S4: Process the riding behavior data of the user to be identified in S1 and S2, and input it into the trained motorcycle dangerous riding behavior identification model to obtain the identification result of whether it is a dangerous riding behavior.

[0011] Furthermore, in S1, the explicit riding characteristics include: x-axis angle, y-axis angle, z-axis angle, brake system pressure, and riding speed; the implicit riding characteristics include: x-axis linear acceleration, axis linear acceleration, z-axis linear acceleration, throttle opening, and torque.

[0012] Furthermore, the step of cleaning the acquired user historical riding data in S1 includes: data denoising, missing value processing, and outlier processing.

[0013] Furthermore, S2 is specifically implemented by the following sub-steps:

[0014] S2.1: Determine the target time step according to the collection frequency of explicit riding characteristic data and implicit riding characteristic data;

[0015] S2.2: Downsample the cycling feature data whose data collection time step is smaller than the target time step, expand the data through interpolation for the cycling feature data whose data collection time step is larger than the target time step, and unify the time step of explicit cycling feature data and implicit cycling feature data.

[0016] Furthermore, in S2.2, the riding characteristic data whose data collection time step is longer than the target time step is expanded by interpolation, and the interpolation method adopts linear interpolation, and the expression is as follows:

[0017]

[0018] Where t is the target interpolation time point, X(t) is the interpolation target, that is, fill in the data value at time point t; t1 and t2 are two known adjacent time points; is the riding characteristic data value at time point t1, is the riding characteristic data value at time point t2.

[0019] Furthermore, in S2.2, the riding feature data whose data collection time step is smaller than the target time step is downsampled, and the specific operation is as follows:

[0020] (1) Create a window with a window length and a moving step length equal to the target time step, and segment the riding feature data whose data collection time step is smaller than the target time step;

[0021] (2) Calculate the average value of the riding characteristic data in each section, arrange the obtained average values ​​in order, and obtain the downsampled riding characteristic data.

[0022] Furthermore, in S3, the explicit cycling feature analysis module is implemented based on a fully connected neural network, which aims to perform vector modeling on the explicit cycling features and capture the potential interactive relationship between the explicit cycling behaviors, and finally output a feature vector representing the explicit cycling features; the explicit cycling feature analysis module includes, in order: an input layer, a linear layer, an activation layer, and an output layer;

[0023] The implicit riding feature analysis module is implemented based on the LSTM network, and captures the user's riding trend and riding mode changes by the dynamic information of the user's riding operation in the time series, and finally outputs a feature vector representing the implicit riding feature; the implicit riding feature analysis module includes, in order: an input layer, a convolutional layer, a first LSTM layer, and a second LSTM layer;

[0024] The riding feature fusion and classification module includes, in order: a fusion layer, a fully connected linear layer, and a fully connected nonlinear layer; the fully connected linear layer is used to reduce the dimension of the feature vector of the user's riding behavior and capture the potential relationship between the explicit riding behavior and the implicit riding behavior; the fully connected nonlinear layer is used to perform binary classification on the user's riding behavior and finally output whether the user's riding behavior is a dangerous riding behavior.

[0025] A motorcycle riding dangerous behavior identification system based on feature classification, used to implement the motorcycle riding dangerous behavior identification method based on feature classification, including: a data acquisition module, a data processing module, a dangerous riding behavior detection module, and a dangerous riding behavior response module;

[0026] The data acquisition module is used to acquire user riding behavior data in real time;

[0027] The data processing module is used to clean the acquired user riding behavior data and align the time steps of the explicit riding feature data and the implicit riding feature data therein; the processed data is input into the dangerous riding behavior detection module;

[0028] The dangerous riding behavior detection module: performs real-time analysis and dangerous riding behavior detection on the user's riding behavior data by loading the trained motorcycle dangerous riding behavior recognition model, and sends a signal to the dangerous riding behavior response module when it is recognized that the user may have dangerous riding behavior;

[0029] The dangerous riding behavior response module: after receiving the signal sent by the dangerous riding behavior detection module, the dangerous riding behavior response module will issue a sound warning to the motorcycle user on the vehicle side to remind the user to pay attention to riding safety.

[0030] The beneficial effects of the present invention are:

[0031] The present invention proposes a method and system for identifying dangerous motorcycle riding behaviors based on feature classification, which classifies and processes the riding features of motorcycle users and fully explores the potential information contained in various feature data. Combined with the dynamic modeling capabilities of deep learning algorithms, it can accurately capture the time dependency of motorcycle users' riding behaviors and provide more reliable analysis and prediction for complex driving scenarios. In addition, when applied, the present invention can perform real-time analysis on user riding behavior data and issue early warnings in a timely manner, effectively ensuring user driving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a flow chart of a method for identifying dangerous motorcycle riding behaviors based on feature classification in an embodiment of the present invention.

[0033] Figure 2 4 is a structural diagram of a motorcycle riding dangerous behavior recognition model in an embodiment of the present invention.

[0034] Figure 3 1 is a flow chart of a motorcycle riding dangerous behavior identification system based on feature classification in an embodiment of the present invention. DETAILED DESCRIPTION

[0035] The present invention will be described in detail below according to the accompanying drawings and preferred embodiments, and the purpose and effect of the present invention will become more clear. The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0036] like Figure 1 As shown, a method for identifying dangerous motorcycle riding behaviors based on feature classification specifically includes the following steps:

[0037] S1: Collection and cleaning of user riding behavior characteristic data.

[0038] Obtain the user's historical riding data from the database, as well as the label value of whether each riding segment is dangerous riding. The user's historical riding data includes explicit riding feature data and implicit riding feature data; the label value includes 0 and 1, where 0 means "not dangerous riding" and 1 means "dangerous riding". Clean the acquired historical explicit riding feature data and implicit riding feature data.

[0039] Furthermore, explicit riding characteristics refer to characteristics that can reflect the user's real-time riding status through data at a single point in time, including: x-axis angle, y-axis angle, z-axis angle, brake system pressure, riding speed, etc.

[0040] Implicit riding features refer to features whose significance cannot be directly observed through data at a single time point. They need to be mined by combining time series with deep learning methods to extract potential information, including: linear acceleration in the x-axis direction, linear acceleration in the y-axis direction, linear acceleration in the z-axis direction, throttle opening, torque, etc.

[0041] Specifically, the steps of cleaning the explicit and implicit riding feature data include: data denoising, missing value processing, and outlier processing.

[0042] S2: The explicit riding feature data obtained in S1 is aligned with the implicit riding feature data in time step.

[0043] The user's historical riding data in the database is collected by vehicle-side sensors, gyroscopes and other devices. Considering that the sampling frequencies of explicit riding features and implicit riding features may be different, it is necessary to align the time steps of the feature data points to ensure that the explicit riding features and implicit riding features correspond to the same actual driving state at the same time step. The specific operations are as follows:

[0044] The target time step is determined according to the data collection frequency of explicit cycling characteristics and implicit cycling characteristics. The feature data with higher collection frequency (that is, the corresponding time step is smaller than the target time step) is downsampled, and the feature data with lower collection frequency (that is, the corresponding time step is larger than the target time step) is expanded through interpolation, so as to unify the time step of explicit cycling feature data and implicit cycling feature data.

[0045] In this embodiment, as shown in Tables 1 and 2 below, the acquisition frequency of explicit riding characteristic data is 2 Hz (i.e., 2 times / second), and the corresponding time step is 0.5 s; the acquisition frequency of implicit riding characteristic data is 10 Hz (i.e., 10 times / second), and the corresponding time step is 0.1 s.

[0046] Table 1 Schematic diagram of explicit cycling characteristics within 5 seconds

[0047] Data collection time point (t) 0.5s 1s 1.5s ...... 3.5s 4s 5s Data(X) <![CDATA[X1]]> <![CDATA[X2]]> <![CDATA[X3]]> ...... <![CDATA[X8]]> <![CDATA[X9]]> <![CDATA[X 10 ]]>

[0048] Table 2 Implicit cycling characteristics within 5 seconds

[0049] Data collection time point (T) 0.1s 0.2s 0.3s 0.4s ...... 4.8s 4.9s 5s Data(Y) <![CDATA[Y1]]> <![CDATA[Y2]]> <![CDATA[Y3]]> <![CDATA[Y4]]> ...... <![CDATA[Y 48 ]]> <![CDATA[Y 49 ]]> <![CDATA[X 50 ]]>

[0050] The target time step is set to 0.2s. Since the time step of the explicit riding feature data collection is larger than the target time step, it is necessary to perform interpolation to fill the data volume and obtain the explicit riding feature data after the time step is aligned. The data is output in the form of a sequence. The interpolation method can be linear interpolation, polynomial interpolation and other methods. This embodiment uses linear interpolation because the explicit riding feature data usually changes more smoothly, the amount of calculation is small, and the data processing is more efficient. The expression of linear interpolation is as follows:

[0051]

[0052] Where t is the target interpolation time point, X (t) is the interpolation target, that is, fill in the data value at time point t; t1 and t2 are two adjacent and known time points; is the explicit riding characteristic data value at time point t1, is the explicit riding characteristic data value at time point t2.

[0053] Combined with the above explicit riding characteristics data table, if you want to insert the data value X of t = 0.8s within 0.5s-1s (t) , then the inserted data value can be calculated by the following formula:

[0054]

[0055] Since the time step of implicit riding feature data collection is smaller than the target time step, downsampling is required. The specific operations are as follows:

[0056] The implicit riding feature data is segmented with a window length of 0.2s and a moving step length equal to the window length, that is, also 0.2s, to obtain [Y1,Y2], [Y3,Y4], [Y5,Y6],…, and then the average value of the data in each window is calculated. The average value series obtained The downsampled implicit riding feature data will be used in subsequent steps.

[0057] S3: Construction and training of motorcycle dangerous riding behavior recognition model based on deep learning.

[0058] The motorcycle dangerous riding behavior recognition model is based on deep learning algorithms, such as Figure 2 As shown in the figure, the model is divided into three parts: explicit riding feature analysis module, implicit riding feature analysis module, and riding feature fusion and classification module. The explicit riding feature analysis module is used to model and analyze explicit riding features and output feature vectors representing explicit riding features. The implicit riding feature analysis module is used to model and analyze implicit riding features and output feature vectors representing implicit riding features. The riding feature fusion and classification module is used to splice and fuse the two feature vectors obtained by the above two modules to obtain a feature vector representing the user's overall riding behavior, and classify the user's riding behavior as dangerous riding based on the feature vector of the user's overall riding behavior. The following is a detailed description of each module.

[0059] (1) Explicit cycling feature analysis module: This module is implemented based on a fully connected neural network and is used to perform vector modeling of explicit cycling features and capture potential interactions between explicit cycling behaviors. For example, a change in the x-axis angle at a high speed may indicate a dangerous turning behavior. Specifically, this module includes the following structures in sequence:

[0060] Input layer: accepts the user's explicit riding feature sequence after alignment of time steps as input.

[0061] Linear layer: Captures the interaction between features through linear calculations of weight parameters and input data.

[0062] Activation layer: Through nonlinear calculations, further learn the potential and more complex relationships between features.

[0063] Output layer: Outputs the feature vector representing the user's explicit riding characteristics.

[0064] (2) The implicit riding feature analysis module is implemented based on the LSTM model. The implicit riding feature data at a single time point is not sufficient to reflect the user's riding status. For example, the user's riding acceleration at a certain time point is 2m / s 2 , if the user's acceleration at the last time point is 0.5m / s 2 , it means that he may be accelerating rapidly at this moment; if the user's acceleration at the previous time point is 3m / s 2 , it means that he may be decelerating suddenly at this moment. Therefore, the LSTM algorithm is introduced into the implicit riding feature analysis module to capture the user's riding trend and riding mode changes through the dynamic information of the user's riding data in the time series. Specifically, the module includes the following structures in sequence:

[0065] Input layer: accepts the user's implicit riding feature sequence after alignment of time steps as input.

[0066] Convolutional layer: In particular, the present invention adds a convolutional layer to the traditional LSTM model to capture the short-term change trend of the user's implicit riding behavior through convolution operations. The introduction of the convolutional layer can help the implicit riding behavior analysis module to efficiently extract low-level features (such as acceleration peaks) in the time series, reduce the noise that needs to be processed by the subsequent LSTM layer, and optimize the model operation efficiency.

[0067] The first LSTM layer is used to capture the basic changing trends in the user's implicit riding feature data.

[0068] The second LSTM layer is used to further extract higher-level behavior patterns from the user's implicit riding feature data.

[0069] Output layer: Outputs the feature vector representing the user's implicit riding characteristics.

[0070] (3) The riding feature fusion and classification module will first concatenate the feature vector representing the user's explicit riding features with the feature vector representing the user's implicit riding features to obtain a feature vector representing the user's overall riding behavior. Then, the feature vector representing the user's overall riding behavior is input into the fully connected linear layer, which is used to reduce the dimension of the feature vector of the user's overall riding behavior and capture the potential relationship between explicit riding behavior and implicit riding behavior; the output of the fully connected linear layer is used as the input of the fully connected nonlinear layer, which is used to perform binary classification of the user's riding behavior and finally output whether the user's riding behavior is dangerous riding behavior (0 is not dangerous, 1 is dangerous).

[0071] The user's historical riding behavior data processed by S2 is divided into training set, validation set and test set in a ratio of 7.5:1.5:1.5, and input into the above motorcycle dangerous riding behavior recognition model for training. The model adjusts the parameters by learning the error between the output value and the true label value, so that the next output is closer to the true label value. In the training stage, it is optimized through continuous iterations to achieve the effect of identifying dangerous riding behaviors based on user riding data.

[0072] S4: Process the riding behavior data of the user to be identified in S1 and S2, and input it into the trained motorcycle dangerous riding behavior identification model to obtain the identification result of whether it is a dangerous behavior.

[0073] like Figure 3 As shown, in order to realize the above-mentioned method for identifying dangerous motorcycle riding behaviors based on feature classification, an embodiment of the present invention proposes a system for identifying dangerous motorcycle riding behaviors based on feature classification, including: a data acquisition module, a data processing module, a dangerous riding behavior detection module, and a dangerous riding behavior response module.

[0074] Data acquisition module: used to obtain user riding behavior data in real time from vehicle-side sensors, gyroscopes and other devices.

[0075] Data processing module: used to clean the acquired user riding behavior data and align the time steps of the explicit riding feature data and the implicit riding feature data. The processed data is input into the dangerous riding behavior detection module.

[0076] Dangerous riding behavior detection module: By loading the trained motorcycle dangerous riding behavior recognition model, the user's riding behavior is analyzed in real time and dangerous riding behavior is detected. When it is identified that the user may have dangerous riding behavior, a signal will be sent to the dangerous riding behavior response module.

[0077] Dangerous riding behavior response module: After receiving the signal from the dangerous riding behavior detection module, the dangerous riding behavior response module will give a sound warning to the motorcycle user on the vehicle side to remind the user to pay attention to riding safety.

[0078] Those skilled in the art can understand that the above are only preferred examples of the invention and are not intended to limit the invention. Although the invention is described in detail with reference to the above examples, those skilled in the art can still modify the technical solutions recorded in the above examples or replace some of the technical features therein with equivalents. Any modification, equivalent replacement, etc. made within the spirit and principle of the invention shall be included in the protection scope of the invention.

Claims

1. A method for identifying dangerous motorcycle riding behaviors based on feature classification, characterized in that: The following steps are involved: S1: Obtain the user's historical riding data and the label value of whether each riding segment is dangerous riding from the database, and clean the user's historical riding data; the user's historical riding data includes explicit riding feature data and implicit riding feature data. The explicit riding feature refers to the feature that reflects the user's real-time riding status through the data at a single time point, and the implicit riding feature refers to the feature whose meaning cannot be directly observed through the data at a single time point, and it is necessary to combine the time series to mine the potential information; S2: The time step of aligning the explicit riding feature data and the implicit riding feature data obtained in S1; S3: Use the riding feature data obtained in S2 to build and train a motorcycle dangerous riding behavior recognition model, the motorcycle dangerous riding behavior recognition model includes: an explicit riding feature analysis module based on a fully connected neural network, an implicit riding feature analysis module based on an LSTM network, and a riding feature fusion and classification module; the explicit riding feature analysis module is used to model and analyze the explicit riding features and output a feature vector representing the explicit riding features, the implicit riding feature analysis module is used to model and analyze the implicit riding features and output a feature vector representing the implicit riding features, the riding feature fusion and classification module is used to splice and fuse the feature vector representing the explicit riding features with the feature vector representing the implicit riding features to obtain a feature vector representing the user's overall riding behavior, and classify the user's riding behavior as a dangerous riding behavior based on the feature vector of the user's overall riding behavior; S4: Process the riding behavior data of the user to be identified in S1 and S2, and input it into the trained motorcycle dangerous riding behavior identification model to obtain the identification result of whether it is a dangerous riding behavior.

2. The method for identifying dangerous motorcycle riding behaviors based on feature classification according to claim 1 is characterized in that: In S1, the explicit riding characteristics include: x-axis angle, y-axis angle, z-axis angle, brake system pressure, and riding speed; the implicit riding characteristics include: x-axis linear acceleration, axis linear acceleration, z-axis linear acceleration, throttle opening, and torque.

3. The method for identifying dangerous motorcycle riding behaviors based on feature classification according to claim 1 is characterized in that: The steps of cleaning the acquired user historical riding data in S1 include: data denoising, missing value processing, and outlier processing.

4. The method for identifying dangerous motorcycle riding behaviors based on feature classification according to claim 1 is characterized in that: The S2 is specifically implemented by the following sub-steps: S2.1: Determine the target time step according to the collection frequency of explicit riding characteristic data and implicit riding characteristic data; S2.2: Downsample the cycling feature data whose data collection time step is smaller than the target time step, expand the data through interpolation for the cycling feature data whose data collection time step is larger than the target time step, and unify the time step of explicit cycling feature data and implicit cycling feature data.

5. The method for identifying dangerous motorcycle riding behaviors based on feature classification according to claim 4 is characterized in that: In S2.2, the riding feature data with a data collection time step longer than the target time step is expanded by interpolation, and the interpolation method adopts linear interpolation, and the expression is as follows: Where t is the target interpolation time point, X (t) is the interpolation target, that is, fill in the data value at time point t; t1 and t2 are two known adjacent time points; is the riding characteristic data value at time point t1, is the riding characteristic data value at time point t2.

6. The method for identifying dangerous motorcycle riding behaviors based on feature classification according to claim 4 is characterized in that: In S2.2, the riding feature data whose data collection time step is smaller than the target time step is downsampled, and the specific operation is as follows: (1) Create a window with a window length and a moving step length equal to the target time step, and segment the riding feature data whose data collection time step is smaller than the target time step; (2) Calculate the average value of the riding characteristic data in each section, arrange the obtained average values ​​in order, and obtain the downsampled riding characteristic data.

7. The method for identifying dangerous motorcycle riding behaviors based on feature classification according to claim 1, characterized in that: In S3, the explicit cycling feature analysis module is implemented based on a fully connected neural network, which aims to perform vector modeling on the explicit cycling features and capture the potential interactive relationship between the explicit cycling behaviors, and finally output a feature vector representing the explicit cycling features; Dominant The cycling feature analysis module includes, in order: input layer, linear layer, activation layer, and output layer; The implicit riding feature analysis module is implemented based on the LSTM network, and captures the user's riding trend and riding mode changes by the dynamic information of the user's riding operation in the time series, and finally outputs a feature vector representing the implicit riding feature; the implicit riding feature analysis module includes, in order: an input layer, a convolutional layer, a first LSTM layer, and a second LSTM layer; The riding feature fusion and classification module includes, in order: a fusion layer, a fully connected linear layer, and a fully connected nonlinear layer; the fully connected linear layer is used to reduce the dimension of the feature vector of the user's riding behavior and capture the potential relationship between the explicit riding behavior and the implicit riding behavior; the fully connected nonlinear layer is used to perform binary classification on the user's riding behavior, and finally output whether the user's riding behavior is a dangerous riding behavior.

8. A motorcycle riding dangerous behavior identification system based on feature classification, used to implement the motorcycle riding dangerous behavior identification method based on feature classification according to any one of claims 1 to 7, characterized in that: include: Data acquisition module, data processing module, dangerous riding behavior detection module, and dangerous riding behavior response module; The data acquisition module is used to acquire user riding behavior data in real time; The data processing module is used to clean the acquired user riding behavior data and align the time steps of the explicit riding feature data and the implicit riding feature data therein; The processed data is input into the dangerous riding behavior detection module; The dangerous riding behavior detection module: performs real-time analysis and dangerous riding behavior detection on the user's riding behavior data by loading the trained motorcycle dangerous riding behavior recognition model, and sends a signal to the dangerous riding behavior response module when it is recognized that the user may have dangerous riding behavior; The dangerous riding behavior response module: after receiving the signal sent by the dangerous riding behavior detection module, the dangerous riding behavior response module will issue a sound warning to the motorcycle user on the vehicle side to remind the user to pay attention to riding safety.

Images