Driving pattern representation and identity recognition method based on natural driving data

Abstract

The present invention discloses a method for characterizing and identifying driving patterns based on natural driving data, comprising the following steps: 1) obtaining driving parameter data of a vehicle; 2) preprocessing the data in step 1); the preprocessing includes data interpolation, data screening, and data normalization processing; 3) constructing a sample set for the data processed in step 2); by setting a time window length L w , using sliding windows to construct samples and construct sample sets; 4) according to the sample set of driving data, obtaining the distribution of various types of data in the driver's driving data as the driver's driving fingerprint; 5) driver identity recognition. The present invention proposes a driving fingerprint representation and identity recognition method based on natural driving data. By analyzing and processing natural driving data, the driver's driving fingerprints at different times are extracted as the basis for identifying the driver's identity.

Description

Technical Field

[0001] The present invention relates to intelligent driving assistance technology, and in particular to a driving pattern representation and identity recognition method based on natural driving data. Background Art

[0002] With the continuous development of the automobile industry and the advancement of informatization, more and more sensors and controllers are installed in cars. These sensors (accelerometers, steering wheel angle sensors) provide solid data support for researchers to study the driving characteristics of drivers. At present, most of the research on driving characteristics analysis of drivers based on vehicle data divides drivers into several categories based on a certain characteristic, or requires the installation of additional sensor equipment to further analyze the driving characteristics of individual drivers.

[0003] Just like every person has a unique fingerprint, every driver has his or her own unique driving characteristics, also called driving fingerprints or driving patterns. Driver identity verification has very important application prospects in the special vehicle and insurance industries, so it is necessary to design a driver's driving pattern representation and identity recognition method based on existing sensor data. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a driving pattern characterization and identity recognition method based on natural driving data in view of the defects in the prior art.

[0005] The technical solution adopted by the present invention to solve the technical problem is: a driving pattern characterization and identity recognition method based on natural driving data, comprising the following steps:

[0006] 1) Obtain the vehicle's driving parameter data, including vehicle speed v x , vehicle speed acceleration a x 、Steering wheel angle θ x 、Steering wheel rotation angular velocity ω x ; The driving parameter data is time series data corresponding to time;

[0007] 2) Preprocessing the data in step 1); the preprocessing includes data interpolation, data screening, and data normalization;

[0008] 3) Construct a sample set for the data processed in step 2); by setting the time window length L w ,Use sliding window to construct samples and construct sample sets;

[0009] The sample set includes: driver number, vehicle speed, vehicle speed acceleration, steering wheel angle, and steering wheel rotation angular velocity data;

[0010] 4) Based on the sample set of driving data, the distribution of various types of data in the driver's driving data is obtained as the driver's driving fingerprint; by collecting driving data sample sets of different drivers, a driving fingerprint library is formed.

[0011] 5) Driver identification

[0012] By comparing the similarity between the driving fingerprint of the driving data to be identified and the driving fingerprint in the driving fingerprint library, the driver in the driving fingerprint library corresponding to the driving data to be identified is identified; the driving data to be identified is input into the convolutional neural network, and the input of the convolutional neural network includes vehicle speed v, vehicle speed acceleration a, steering wheel angle θ, and steering wheel rotation angular velocity ω, the size of which is L w ×4 two-dimensional data matrix, the network output is a driver number in the matching driving fingerprint library.

[0013] According to the above scheme, the data preprocessing in step 2) is specifically as follows:

[0014] 2.1) First, interpolate various types of data and obtain the interpolation curve v of various types of data through the spline curve interpolation method (t) 、a (t) ,θ (t) ,ω (t) , according to the set sampling rate F s Interpolate various types of data to obtain a series of data;

[0015] 2.2) Screen the new data set obtained by sampling, select the data with vehicle speed between 80km / h and 120km / h, and retain the corresponding vehicle speed, vehicle speed acceleration, steering wheel angle, and steering wheel rotation angular velocity data;

[0016] 2.3) Normalize the data set.

[0017] According to the above scheme, the structure of the convolutional neural network in step 5) is as follows:

[0018] The structure of the convolutional neural network includes an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer and a fully connected layer connected in sequence, wherein the pooling method is maximum pooling and the activation function is a ReLU function.

[0019] The confusion matrix of the classification results output by the convolutional neural network is used to obtain the matching driver number.

[0020] The beneficial effects produced by the present invention are:

[0021] The present invention proposes a driving fingerprint characterization and identity recognition method based on natural driving data. By analyzing and processing the natural driving data, the driving fingerprints of the driver at different times are extracted as the basis for identifying the driver's identity. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0023] Figure 1 is a schematic structural diagram of an embodiment of the present invention;

[0024] Figure 2 is a schematic diagram of the data structure of a driving data sample according to an embodiment of the present invention;

[0025] Figure 3 Schematic diagram of two driving fingerprints of the same driver in an embodiment of the present invention;

[0026] Figure 4 It is a structural diagram of a convolutional neural network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0028] like Figure 1 As shown, a driving pattern representation and identity recognition method based on natural driving data includes the following steps:

[0029] Step 1: Obtain the vehicle's driving parameter data, including the vehicle speed v x , vehicle speed acceleration a x 、Steering wheel angle θ x 、Steering wheel rotation angular velocity ω x ; The driving parameter data is time series data corresponding to time;

[0030] In step 1, the CAN bus is connected through the vehicle OBD interface to obtain the vehicle's driving parameters, including but not limited to the vehicle speed v x , vehicle speed acceleration a x 、Steering wheel angle θ x 、Steering wheel rotation angular velocity ω x The series of data collected are:

[0031] Speed ​​V 1 {v 1 ,v 2 ,v 3 ,…,v n1}、

[0032] Vehicle speed acceleration A 1 {a 1 ,a 2 ,a 3 ,…,a n2}、

[0033] Steering wheel angle Θ 1 {θ 1 ,θ 2 ,θ 3 ,…,θ n3}、

[0034] Steering wheel rotation angular velocity Ω 1 {ω 1 ,ω 2 ,ω 2 ,…,ω n4}.

[0035] where n 1 、n 2 、n 3 、n 4 Not necessarily equal.

[0036] The vehicle OBD interface is a physical interface on the vehicle that can obtain CAN bus data in real time. The CAN bus refers to a network bus on the vehicle that connects various sensors and components of the vehicle, through which the real-time data of various sensors of the vehicle can be obtained;

[0037] Vehicle speed refers to the speed of the vehicle driven by the driver in the direction of travel;

[0038] The vehicle speed acceleration is the acceleration of the vehicle driven by the driver in the driving direction. It is positive if it is in the same direction as the driving direction and negative if it is opposite to the driving direction.

[0039] The steering wheel angle is the rotation angle of the steering wheel of the vehicle driven by the driver, with counterclockwise rotation being positive and clockwise rotation being negative;

[0040] The steering wheel rotation angular velocity is the angular velocity of the steering wheel of the vehicle driven by the driver, and the acceleration direction is counterclockwise for positive and clockwise for negative.

[0041] Step 2: preprocess the data in step 1. First, interpolate the data and obtain the interpolation curve v of each data by using the spline curve interpolation method. (t) 、a (t) ,θ (t) ,ω (t) According to the sampling rate F s Interpolate various data to obtain a series of data (the sampling rate F in this embodiment is s =15Hz):

[0042] Speed ​​V 2 {v (0) ,v (Δt) ,v (2Δt) ,…,v (nΔt)}、

[0043] Vehicle speed acceleration A 2 { a(0) ,a (Δt) ,a (2Δt) ,…,a (nΔt)}、

[0044] Steering wheel angle Θ 2 {θ (0) ,θ (Δt) ,θ (2Δt) ,…,θ (nΔt)}、

[0045] Steering wheel rotation angular velocity Ω 2 { ω(0) ,ω (Δt) ,ω (Δt) ,…,ω (nΔt)}.

[0046] Where, Δt = 1 / F s .

[0047] Then, the new data set obtained by interpolation is screened. If 80 ≥ v iΔt (0≤i≤n) or 120≤v iΔt (0≤i≤n), then delete the data v (iΔt) 、a (iΔt) ,θ (iΔt) ,ω (iΔt) If the condition is not met, the data v is retained. (iΔt) 、a (iΔt) ,θ (iΔt) ,ω (iΔt) After this step, we get the data set:

[0048] Speed ​​V 3 { v(0) ,v (Δt) ,v (2Δt) ,…,v (mΔt)}、

[0049] Vehicle speed acceleration A 3 { a(0) ,a (Δt) ,a (2Δt) ,…,a (mΔt)}、

[0050] Steering wheel angle Θ 3 {θ (0) ,θ(Δt) ,θ (2Δt) ,…,θ (mΔt)}、

[0051] Steering wheel rotation angular velocity Ω 3 {ω (0) ,ω (Δt) ,ω (2Δt) ,…,ω (mΔt)}.

[0052] Secondly, normalize the data set and find V 3 , A 3 , Θ 3 ,Ω 3 The maximum and minimum values ​​in

[0053] v max 、v min 、a max 、a min ,θ max ,θ min ,ω max ,ω min . Get the normalized data set:

[0054] Speed ​​V 4 =V 3 / (v max -v min ),

[0055] Vehicle speed acceleration A 4 =A 3 / (a max -a min ),

[0056] Steering wheel angle Θ 4 =Θ 3 / (θ max -θ min ),

[0057] Steering wheel rotation angular velocity Ω 4 =Ω 3 / (ω max -ω min ).

[0058] Finally, the training sample set is constructed.

[0059] The data structure of a sample in the training sample set is as follows: Figure 2 As shown, the time window length L w In this embodiment, there are 120 data, that is, the time span of the time window is t w = 8s. Each sample includes 480 data in total.

[0060] In this embodiment, driving data of 15 drivers in the same experimental vehicle are collected, each driver is numbered, and the number of driving data samples is between 1300 and 2000.

[0061] Step 3, driving fingerprint refers to the driver's driving habits, which is reflected in the distribution of various types of data in the driving data. The sample set contains the distribution of various types of driving data, that is, each sample contains the driver's driving fingerprint, each driver has multiple samples, each sample is different but has certain similarities, and the driver's driving fingerprint can be extracted by capturing the similarities in each sample.

[0062] Driving fingerprint Figure 3 , Figure 3 The horizontal axis is the number of the data arranged in time order, and the vertical axis is the normalized vehicle speed v, vehicle speed acceleration a, steering wheel angle θ, and steering wheel rotation angular velocity ω. Figure 3 shows the driving fingerprints of the same driver at different times, and there is a certain similarity between the two.

[0063] Step 4: The driver's driving fingerprint can be extracted by capturing the similarity between the driver's driving data and each sample in the sample set, and the data sample is input into the convolutional neural network. The input of the convolutional neural network is a two-dimensional data matrix, and the network output is the matching driver number.

[0064] The structure of the convolutional neural network used in this embodiment is as follows: Figure 4 , where the pooling method is maximum pooling and the activation function is the ReLU function.

[0065] The confusion matrix of the classification results output by the network is obtained, and it is found that the network's identity authentication accuracy for 15 drivers reaches 99.4%.

[0066] The entire driving fingerprint characterization and identification method process is completed.

[0067] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all these improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. A method for characterizing driving patterns and identifying identity based on natural driving data, characterized in that: The following steps are involved: 1) Obtain the vehicle's driving parameter data, including speed , vehicle speed acceleration , steering wheel angle , steering wheel rotation angular velocity ; The driving parameter data is time series data corresponding to time; 2) Preprocessing the data in step 1); the preprocessing includes data interpolation, data screening, and data normalization; The details are as follows: 2.1) First, interpolate various types of data and obtain the interpolation curves of various types of data through the spline curve interpolation method. , according to the set sampling rate Interpolate various types of data to obtain a series of data; 2.2) Filter the new data set obtained by sampling, select the data with vehicle speed between 80km / h and 120km / h, and retain the corresponding vehicle speed, vehicle speed acceleration, steering wheel angle, and steering wheel rotation angular velocity data; 2.3) Normalize the data set; 3) Construct a sample set for the data processed in step 2); by setting the time window length ,Use sliding window to construct samples and construct sample sets; The sample set includes: driver number, vehicle speed, vehicle speed acceleration, steering wheel angle, and steering wheel rotation angular velocity data; 4) Based on the sample set of driving data, the distribution of various types of data in the driver's driving data is obtained as the driver's driving fingerprint; by collecting driving data sample sets of different drivers, a driving fingerprint library is formed; 5) Driver identification; By comparing the similarity between the driving fingerprint of the driving data to be identified and the driving fingerprint in the driving fingerprint library, the driver in the driving fingerprint library corresponding to the driving data to be identified is identified; the driving data to be identified is input into the convolutional neural network, and the input of the convolutional neural network is the vehicle speed information. , vehicle speed acceleration , steering wheel angle , steering wheel rotation angular velocity The size is The network output is a two-dimensional data matrix of a driver in the matching driving fingerprint library.

2. The method for characterizing driving patterns and identifying identity based on natural driving data according to claim 1, characterized in that: The structure of the convolutional neural network in step 5) is as follows: The structure of the convolutional neural network includes an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer and a fully connected layer connected in sequence, wherein the pooling method is maximum pooling and the activation function is a ReLU function; The confusion matrix of the classification results output by the convolutional neural network is used to obtain the matching driver number.

Images