CN106915354B

CN106915354B - Vehicle-mounted device for identifying driver and identification method

Info

Publication number: CN106915354B
Application number: CN201710085385.XA
Authority: CN
Inventors: 高文
Original assignee: Dalian Yiwulian Information Technology Co ltd
Current assignee: Dalian Yiwulian Information Technology Co ltd
Priority date: 2017-02-17
Filing date: 2017-02-17
Publication date: 2019-03-29
Anticipated expiration: 2037-02-17
Also published as: CN106915354A

Abstract

The invention relates to vehicle-mounted equipment for identifying a driver and an identification method, belonging to the field of vehicle-mounted intelligent equipment. The technical scheme is as follows: the vehicle bus data monitoring and reading system comprises a vehicle bus data monitoring and reading unit, an MCU (microprogrammed control unit) computing unit, a local storage unit, a network communication unit and a data characteristic storage and analysis cloud platform, wherein one end of the vehicle bus data monitoring and reading unit is connected with a vehicle, the other end of the vehicle bus data monitoring and reading unit is connected with the MCU computing unit, the MCU computing unit is respectively connected with the local storage unit and the network communication unit, and the network communication unit is connected with the data characteristic storage and analysis cloud platform. The beneficial effects are that: according to the vehicle-mounted device management system, the vehicle-mounted device is accessed into the bus to perform data analysis, a certain calculation method is applied to obtain driving characteristics, the driving characteristics are transmitted to the cloud platform, and the purpose of accurately distinguishing different drivers is achieved by combining continuous self-learning updating and matching optimization of big data stored by the cloud platform, so that the vehicle and the drivers are managed more intelligently and efficiently.

Description

A kind of mobile unit and identification method for recognizing driver

Technical field

The present invention relates to a kind of car-mounted terminal monitoring device more particularly to a kind of mobile unit for recognizing driver and distinguish Verifying method.

Background technique

There is the terminal monitoring equipment of a variety of access vehicle bus on current market, this kind of equipment can be such that people obtain Vehicle real-time running data, to check Vehicular behavior and fault condition.It can be by total but there are no one kind in the market Knot analysis driver reaches the equipment for distinguishing driver to the related data of vehicle operating, and such hardware device will It is widely used in vehicle management, settlement of insurance claim, disposition of breaking rules and regulations, automatic Pilot, the fields such as automobile leasing, shared trip, the present invention Device and method can by extract driver driving characteristics and application class algorithm obtain distinguish driver mesh 's.

Summary of the invention

In order to more intelligence and efficiently vehicle and driver are managed, the present invention provides a kind of identification driver Mobile unit and identification method, the device and method can by analysis and summary driver to the related data of vehicle operating from And achieve the purpose that distinguish driver, to more efficiently be managed intelligently and to vehicle and driver, Ke Yiguang It is general to be applied to vehicle management, settlement of insurance claim, disposition of breaking rules and regulations, automatic Pilot, the fields such as automobile leasing, shared trip.The technology Scheme is as follows:

It is a kind of recognize driver mobile unit, including vehicle bus data decryptor reading unit, MCU computing unit, Local storage unit, network communication unit and data characteristic storage and analysis cloud platform, the vehicle bus data decryptor are read Unit one end is attached with vehicle, and the other end is connect with the MCU computing unit, the MCU computing unit respectively with local Storage unit is connected with network communication unit, and the network communication unit and the data characteristics store and analyze cloud platform and connect It connects.

Further, the vehicle bus data decryptor reading unit and vehicle bus or vehicle standard diagnosis interface connect It connects.

Further, the vehicle bus data decryptor reading unit is connect by CAN bus or Ethernet with vehicle.

The invention also includes a kind of method for recognizing driver, using the mobile unit of above-mentioned identification driver, Execute following steps:

S1, vehicle bus data decryptor reading unit obtain the speed of vehicle, engine by accessing automobile bus in real time Revolving speed, throttle, brake, gear and steering wheel angle information；

S2, using median filtering, filter noise data；

Whether the data in S3, real-time judge newest several seconds have direction information, if not turning to message, return to S1, If there is turning to message, into S4；

S4,5 dimensional characteristics vectors are extracted, 5 dimensional feature vector includes surpassing turn most value, surpass and turn the percentage for accounting for whole cycle Than, return mean location, lack turn most value, scarce turn accounts for the percentage of whole cycle；

S5, data characteristics storage and analysis cloud platform generate classifier using ANN algorithm, and the feature vector that equipment uploads is logical It crosses and is compared with property data base, judge the identity of driver.

Further, steps are as follows for extraction feature vector in S4:

P1, the axle center coordinate that (Xr, Yr) and (Xf, Yf) is respectively vehicle rear axle and front axle is divided into inertial coodinate system OXY, Φ is the yaw angle of car body, and Φ f is front wheel slip angle, and Vr is vehicle rear axle central speed, and Vf is automobile front-axle central speed, and L is Vehicle wheelbase, R are rear-axle steering radius, and P is that vehicle rotates the center of circle, and M is vehicle rear axle axle center, and N is front axle axle center；

P2, speed at rear axle traveling axle center (Xr, Yr) is obtained using following equatioies:

Vr=Xr ' cos (Φ)+Yr ' sin (Φ),

Xr and Yr is coordinate, and Xr ' and Yr ' are speed of the rear-wheel under relative coordinate system；

P3, the kinematical constraint according to the antero posterior axis of automobile

Xf ' sin (Φ+Φ f)-Yf ' cos (Φ+Φ f)=0,

Xr ' sin (Φ)-Yr ' cos (Φ)=0；

It pushes away:

Xr '=Vrcos (Φ),

Yr '=Vrsin (Φ)；

P4, it is obtained according to the geometrical relationship of front and back wheel:

Xf=Xr+Lcos (Φ),

Yf=Yf+Lsin (Φ)；

And then derive the yaw velocity of automobile are as follows:

W=Vr/L*tan (Φ f)；

P5, turning radius R and front wheel slip angle Φ f are obtained according to yaw velocity W and vehicle velocity V r

R=Vr/W,

Φ f=arctan (L/R)；

P6, the kinematics model for obtaining vehicle:

Φ '=tan (Φ f)/L*Vr；

P7, according to sideway angle formula, five features of turn are extracted using the following formula,

T be from yaw velocity be 0 to yaw velocity maximum again to 0 period, t_rFor the time point in 0-T,

Feature1 be it is super turn most value, feature2 be it is super turn the percentage for accounting for whole cycle, feature3 is to return Value position, feature4 be lack turn most value, feature5 is scarce to turn the percentage for accounting for whole cycle.

Further, judged according to the result obtained in S5 using statistics, i.e., is sought to multiple result the mathematics phase It hopes, obtains final result.

Further, the classifier is obtained by big data learning method, and this method is using extracting more people's early period Multiple characteristic values carry out data training using the BP neural network in artificial neural network, to obtain one classification Device.

Further, the BP neural network divides three-layer network, and input layer is 5 nodes, and middle layer is 30 nodes, defeated Layer is 2 nodes out.

The beneficial effects of the present invention are:

The mobile unit and identification method of identification driver of the present invention utilizes automobile self-sensor device data, leads to Data parsing acquirement driving characteristics are carried out after crossing mobile unit access bus, and apply certain calculation method, extract 5 dimensional characteristics Vector, the continuous self study update of big data for combining cloud to save in addition and matching optimization, accurately distinguish different driving to reach The purpose of personnel can be widely applied to vehicle management to more efficiently be managed intelligently and to vehicle and driver, Settlement of insurance claim, disposition of breaking rules and regulations, automatic Pilot, the fields such as automobile leasing, shared trip.

Detailed description of the invention

Fig. 1 is present device composition schematic diagram；

Fig. 2 is work flow diagram of the present invention；

Fig. 3 is motor turning motion model figure of the present invention；

Specific embodiment

Embodiment 1:

Further, the vehicle bus data decryptor reading unit and vehicle bus or vehicle standard diagnose interface (OBD) it connects.

S2, using median filtering, filter noise data；

Whether the data in S3, real-time judge newest 10 seconds have direction information, if not turning to message, return to S1, such as Fruit has steering message, into S4；

Further, steps are as follows for extraction feature vector in S4:

P1, as shown in Fig. 3, dividing into (Xr, Yr) and (Xf, Yf) in inertial coodinate system OXY is respectively vehicle rear axle with before The axle center coordinate of axis, Φ are the yaw angle (course angle) of car body, and Φ f is front wheel slip angle, and Vr is vehicle rear axle central speed, and Vf is Automobile front-axle central speed, L are vehicle wheelbase, and R is rear-axle steering radius, and P is that vehicle rotates the center of circle, and M is vehicle rear axle axle center, N is front axle axle center；

Vr=Xr ' cos (Φ)+Yr ' sin (Φ),

Xr and Yr is coordinate, opposite to regard distance as, and Xr ' and Yr ' are speed of the rear-wheel under relative coordinate system；

Xf ' sin (Φ+Φ f)-Yf ' cos (Φ+Φ f)=0,

Xr ' sin (Φ)-Yr ' cos (Φ)=0；

It pushes away:

Xr '=Vrcos (Φ),

Yr '=Vrsin (Φ)；

Xf=Xr+Lcos (Φ),

Yf=Yf+Lsin (Φ)；

And then derive the yaw velocity of automobile are as follows:

W=Vr/L*tan (Φ f)；

R=Vr/W,

Φ f=arctan (L/R)；

P6, the kinematics model for obtaining vehicle:

Φ '=tan (Φ f)/L*Vr；

T is turn duration, i.e., from yaw velocity be 0 to yaw velocity maximum again to 0 period, t_rFor 0- Time point in T,

Further, the classifier is obtained by big data learning method, and this method is using extracting more people's early period Multiple characteristic values carry out data training using the BP neural network in artificial neural network (ANN), to show that this is one Classifier.

Embodiment 2:

As a kind of individual embodiment or to the supplement of embodiment 1, explanation of nouns in S1:

Speed: refer to automobile true velocity (unit km/h), obtained from vehicle bus；

Revolving speed: refer to rotating speed of automobile engine (unit rad/min), obtained from vehicle bus；

Throttle: refer to the half point ratio (dimensionless) that accelerator pedal of automobile is stepped on, range 0%~100%, 0% indicates do not have It steps on the gas, 100% expression throttle is floored, and is equally also obtained from vehicle bus；

Brake: refer to the half point ratio (dimensionless) that brake-pedal of automobile is stepped on, range 0%~100%, 0% indicates do not have Brake, 100% indicates that brake is floored, and equally also obtains from vehicle bus.

Feature extraction quantifies the characteristic value that everyone drives out by algorithm, needs to consider everyone to control errors Mode.In daily driving, the influence due to environmental factor to changes in vehicle speed feature be account for it is most of, so cannot Extraction as absolute feature.Actual conditions, in the case where vehicle turning, when environment road be it is the same, everyone drive Maximum difference is exactly the mode of control turning, and somebody can at the uniform velocity turn round, and somebody, which understands, to turn big curved early period, and the later period turns small curved again Etc..Therefore the present invention is aided with speed, revolving speed, throttle, brake, gear feature mainly based on feature of turning.According to above The sideway angle formula of release, five features (Primary Stage Data normalization) for extracting turn are respectively as follows: super turn and are most worth, and super turn accounts for The percentage of whole cycle, returns mean location, and scarce turn most is worth, lacks and turn the percentage for accounting for whole cycle.(five features are also five A dimension)

The training of data, equipment extract multiple characteristic values of more people early period and are transmitted to data characteristics storage and analysis Cloud platform, cloud platform application artificial neural network (ANN) carries out data training, to obtain a classifier.It is obtaining later When the characteristic value that equipment uploads, so that it may carry out distinguishing different drivers using this classifier.

Present invention application ANN one of which network is called BP neural network, it is a kind of multilayer inversely propagated by error Feedforward neural network.Here I will divide three-layer network input layer to be 5 nodes (because a feature is five dimensions) middle layer For 30 nodes, output layer is two nodes (two classification)).

The mode of big data refers to that platform is being trained BP neural network after data accumulation to certain degree, from Main correction study, so that the accuracy rate for making platform distinguish driver steps up.

The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto, Anyone skilled in the art within the technical scope of the present disclosure, according to the technique and scheme of the present invention and its Inventive concept is subject to equivalent substitution or change, should be covered by the protection scope of the present invention.

Claims

1. an in-vehicle device for identifying a driver is characterized in that, comprising a vehicle bus data monitoring and reading unit, an MCU computing unit, a local storage unit, a network communication unit and a data feature storage and analysis cloud platform, the vehicle bus data monitoring One end of the reading unit is connected to the vehicle, the other end is connected to the MCU computing unit, the MCU computing unit is respectively connected to the local storage unit and the network communication unit, and the network communication unit is connected to the data feature storage and analysis cloud platform connect;

Perform the following steps when identifying the driver:

S1. The vehicle bus data monitoring and reading unit obtains the speed, engine speed, accelerator, brake, gear and steering wheel angle information of the vehicle in real time by accessing the vehicle bus;

S2. Use median filter to filter noise data;

S3, judge in real time whether the data in the latest several seconds has steering information, if there is no steering information, return to S1, if there is steering information, enter S4;

S4, extracting a 5-dimensional feature vector, the 5-dimensional feature vector includes the maximum value of overturning, the percentage of overturning in the entire cycle, the position of the regression mean, the maximum value of missing turning, and the percentage of missing turning in the entire cycle;

S5. Data feature storage and analysis The cloud platform uses the ANN algorithm to generate a classifier, and the feature vector uploaded by the device is compared with the feature database to determine the driver's identity.

2 . The vehicle-mounted device for identifying a driver according to claim 1 , wherein the vehicle bus data monitoring and reading unit is connected to a vehicle bus or a vehicle standard diagnostic interface. 3 .

3 . The vehicle-mounted device for identifying a driver according to claim 1 , wherein the vehicle bus data monitoring and reading unit is connected to the vehicle through a CAN bus or an Ethernet. 4 .

4. the in-vehicle equipment of identifying driver as claimed in claim 1, is characterized in that, in S4, extracting feature vector step is as follows:

p1. Under the inertial coordinate system OXY, (Xr, Yr) and (Xf, Yf) are the axis coordinates of the rear axle and the front axle of the vehicle respectively, Φ is the yaw angle of the vehicle body, Φf is the declination angle of the front wheel, Vr is the center speed of the rear axle of the vehicle, Vf is the center speed of the front axle of the vehicle, L is the wheelbase of the vehicle, R is the steering radius of the rear wheel, P is the center of rotation of the vehicle, M is the center of the rear axle of the vehicle, and N is the center of the front axle;

p2. Use the following equation to obtain the speed at the rear axle center (Xr, Yr):

Vr=Xr'cos(Φ)+Yr'sin(Φ),

Xr and Yr are the coordinates, and Xr' and Yr' are the speeds of the rear wheels in the relative coordinate system;

p3, according to the kinematic constraints of the front and rear axles of the car

Xf'sin(Φ+Φf)-Yf'cos(Φ+Φf)=0,

Xr'sin(Φ)-Yr'cos(Φ)=0;

Pushed:

Xr'=Vrcos(Φ),

Yr'=Vrsin(Φ);

p4, according to the geometric relationship of the front and rear wheels:

Xf=Xr+Lcos(Φ),

Yf=Yf+Lsin(Φ);

Then the yaw rate of the car is derived as:

W=Vr/L*tan(Φf);

p5. According to the yaw rate W and the vehicle speed Vr, the steering radius R and the front wheel deflection angle Φf are obtained

R=Vr/W,

Φf=arctan(L/R);

p6, get the kinematics model of the vehicle:

Φ'=tan(Φf)/L*Vr;

P7. According to the yaw angle formula, the following formulas are used to extract the five features of the turn:

T is the time period from the yaw angular velocity of 0 to the maximum yaw angular velocity to 0, t _r is the time point within 0-T, feature1=g(t ₁ )

feature4=g(t ₃ )

feature1 is the maximum value of overturning, feature2 is the percentage of overturning in the whole cycle, feature3 is the position of regression to the mean, feature4 is the maximum value of missing turning, and feature5 is the percentage of missing turning in the whole cycle.

5. The vehicle-mounted device for identifying a driver according to claim 1, characterized in that, according to the results obtained in S5, statistics are used to judge, that is, mathematical expectations are obtained for multiple results, and the final result is obtained.

6. The vehicle-mounted device for identifying a driver as claimed in claim 1, wherein the classifier is obtained by a big data learning method, and the big data learning method is to extract multiple eigenvalues of many people in the early stage of application, and apply The BP neural network in the artificial neural network performs data training to obtain the one described classifier.

7. The vehicle-mounted device for identifying a driver as claimed in claim 6, wherein the BP neural network is divided into three layers, the input layer is 5 nodes, the middle layer is 30 nodes, and the output layer is 2 a node.