CN113635905B

CN113635905B - Real-time risk prompting system and device for driving behavior insurance

Info

Publication number: CN113635905B
Application number: CN202110973278.7A
Authority: CN
Inventors: 饶雪峰; 张余明; 曲会晨; 郭振军; 朱昌洪; 林奕森; 刘洪林
Original assignee: Guilin University of Aerospace Technology
Current assignee: Guilin University of Aerospace Technology
Priority date: 2021-08-24
Filing date: 2021-08-24
Publication date: 2023-04-18
Anticipated expiration: 2041-08-24
Also published as: CN113635905A

Abstract

The invention discloses a risk real-time prompting system and device facing driving behavior insurance, and relates to the technical field of an intelligent terminal of a vehicle network, namely, image data in front of a vehicle is obtained through a visual information processing unit, and the image data in front of the vehicle is used for identifying the condition of a traffic participation entity and the condition of line pressing of the vehicle; acquiring vehicle positioning data and navigation data through a map navigation unit; acquiring real-time driver operation data through a driving behavior acquisition unit, and identifying the operation intention of a driver according to the real-time driver operation data; the risk driving behavior detection and prompt unit generates a combined safe driving suggestion according to the condition of a traffic participating entity, the line pressing condition of a vehicle, vehicle positioning data, navigation data and the operation intention of a driver, and the driver is prompted through the combined safe driving suggestion, so that the deviation of the conventional device on driving risk evaluation caused by the fact that the driving risk evaluation is not combined with the external driving environment of the vehicle or the operation behavior intention of the driver is overcome.

Description

Risk real-time prompt system and device for driving behavior insurance

Technical Field

The invention belongs to the technical field of safety, and particularly relates to a real-time risk prompting system and device for driving behavior insurance.

Background

In recent years, with the increasing scale of the automobile market, the insurance scale of the vehicles reaches the trillion level. In the face of a huge car insurance market, competition among large insurance companies is also becoming fierce. The existing vehicle insurance pricing model based on the vehicle basic information lacks identification and evaluation of driver behaviors and safety consciousness, cannot accurately price specific customers, has the condition of cooking in a big pot, and does not meet the requirements of differential service innovation and market competition of insurance companies. Therefore, through innovative technical means, the risk characteristics of the car insurance customers are effectively identified, differentiated prices and service strategies are formulated for the car insurance customers, and the method is a powerful guarantee for an insurance company to optimize business modes, improve operation efficiency and improve market competitive advantages.

The technical means of the traditional UBI terminal equipment is as follows: the vehicle track is obtained through a GPS positioning technology, events such as 'four emergencies' (rapid acceleration, rapid deceleration, rapid braking and rapid turning) and the like with high risk correlation on driving behaviors are identified by combining vehicle communication bus data and vehicle body dynamic acquisition data, the events are uploaded to a cloud server through a wireless data communication means, and after data are analyzed by the cloud server, a risk assessment result is obtained, and insurance cost is calculated.

There is room for improvement in the technical approach of conventional UBI terminals. First, for the judgment of the data of the emergency and the like, the judgment system is set based on the statistical average, and the current actual traffic situation cannot be combined. Secondly, the device mainly takes collected and returned data as a main function, and can generate a related risk result only after data analysis is carried out by the cloud, and a user needs to check the risk result by a mobile phone or a web mode, so that the risk evaluation result is known to have hysteresis. Preferably, the real-time risk prompt can be carried out on the driving behavior of the driver in the driving process, so that the driving behavior of the driver is improved, the risk is reduced, and the win-win situation of customers, insurance service providers and even the society is achieved.

Disclosure of Invention

The invention aims to provide a driving behavior insurance-oriented risk real-time prompting system and device, so that the deviation of the conventional device for driving risk assessment caused by the fact that the driving risk assessment is not combined with the external driving environment of a vehicle or the operation behavior intention of a driver is overcome.

In order to achieve the above object, the present invention provides a real-time risk prompting system for driving behavior insurance, comprising:

the visual information processing unit is used for acquiring image data in front of a vehicle and identifying the condition of a traffic participation entity and the line pressing condition of the vehicle for the image data in front of the vehicle;

the map navigation unit is used for acquiring vehicle positioning data and navigation data;

the driving behavior acquisition unit is used for acquiring real-time driver operation data and identifying the operation intention of the driver according to the real-time driver operation data; and

and the risk driving behavior detection and prompt unit is used for generating a combined safe driving suggestion according to the condition of the traffic participation entity, the line pressing condition of the vehicle, the vehicle positioning data, the navigation data and the operation intention of the driver, and prompting the driver through the combined safe driving suggestion.

Furthermore, the risk driving behavior detection and prompt unit is further configured to compare the merged safety driving advice with real-time driver operation data, determine whether the driver executes according to the merged safety driving advice, determine that a risk driving behavior exists if the driver does not execute according to the merged safety driving advice, and send the event to the cloud uploading unit.

The system further comprises a vehicle body dynamic abnormity detection unit, wherein the vehicle body dynamic abnormity detection unit is used for acquiring vehicle driving state information, detecting whether a vehicle body dynamic abnormity event exists according to the vehicle driving state information, and sending a judgment result to the risk driving behavior detection and prompt unit.

Further, the dangerous driving behavior detecting and prompting unit is configured to generate a merged safe driving suggestion according to the condition of the traffic participation entity, the vehicle line pressing condition, the vehicle positioning data and the navigation data, and the operation intention of the driver, and specifically includes:

judging the rear-end collision situation, the pedestrian collision risk and the red light running situation of the vehicle according to the situation of the traffic participation entity identified by the visual information processing unit and the line pressing situation of the vehicle to generate a risk avoidance driving suggestion;

generating a navigation safe driving suggestion according to the vehicle positioning data and the navigation data;

and summarizing the risk avoidance driving advice and the navigation safety driving advice, and removing repeated safety driving advice to obtain combined safety driving advice.

Further, the merged safety driving advice is to continuously perform the detection action periodically, and the latest merged safety driving will override the previous merged safety driving.

The utility model provides a real-time suggestion device of risk towards driving action insurance, real-time suggestion device of risk is applied to real-time reminder system of risk, includes:

the AI vision processing module is used for acquiring image data in front of a vehicle and identifying the condition of a traffic participation entity and the line pressing condition of the vehicle for the image data in front of the vehicle;

the vehicle positioning module is used for acquiring positioning information of a vehicle;

the vehicle bus data acquisition module is used for acquiring real-time driver operation data;

the main control module is respectively and electrically connected with the AI vision processing module, the vehicle positioning module and the vehicle bus data acquisition module, and is used for generating a merged safe driving suggestion according to the vehicle information transmitted by the vehicle positioning module and the vehicle bus data acquisition module; and

and the human-computer interaction module is connected with the main control module and is used for displaying and prompting the driver according to the merged safe driving.

The vehicle driving state information acquisition module is used for detecting whether a vehicle body dynamic abnormal event exists or not according to the vehicle driving state information and sending a detection result to the main control module.

Furthermore, the system also comprises a remote communication module which is electrically connected with the main control module.

Compared with the prior art, the invention has the following beneficial effects:

according to the driving behavior insurance-oriented risk real-time prompting system and device, the risk real-time prompting system acquires data information through the device, namely acquires image data in front of a vehicle through the visual information processing unit, and identifies the conditions of traffic participation entities and the line pressing conditions of the vehicle according to the image data in front of the vehicle; acquiring vehicle positioning data and navigation data through a map navigation unit; acquiring real-time driver operation data through a driving behavior acquisition unit, and identifying the operation intention of a driver according to the real-time driver operation data; the risk driving behavior detection and prompt unit generates a merged safe driving suggestion according to the condition of a traffic participation entity, the line pressing condition of a vehicle, vehicle positioning data, navigation data and the operation intention of a driver, and the driver is prompted through the merged safe driving suggestion, namely the merged safe driving suggestion and the prompt are obtained by acquiring the external driving environment of the vehicle through the visual information processing unit and the map navigation unit and acquiring the operation behavior intention of the driver through the driving behavior acquisition unit, so that the deviation of the conventional device for driving risk assessment caused by the fact that the driving risk assessment is not combined with the external driving environment of the vehicle or the operation behavior intention of the driver is overcome.

Drawings

In order to more clearly illustrate the technical solution of the present invention, the drawings required to be used in the description of the embodiments are briefly introduced below, it is obvious that the drawings in the following description are only one embodiment of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.

FIG. 1 is a flow chart of a real-time risk alert system for driving behavior insurance of the present invention;

FIG. 2 is a schematic diagram of the YOLO v3 network structure of the present invention;

FIG. 3 is a schematic view of the pinhole imaging of the present invention;

FIG. 4 is a schematic structural diagram of a real-time risk prompting device for driving behavior insurance according to an embodiment of the invention;

FIG. 5 is a schematic structural diagram of a risk real-time prompting device for driving behavior insurance according to an embodiment of the present invention;

wherein: 1. a main control module; 2. a vehicle positioning module; 3. a vehicle bus data acquisition module; 4. a vehicle pose acquisition module; 5. a human-computer interaction module; 6. a remote communication module; 7. an AI vision processing module; 8. a camera is provided.

Detailed Description

The technical solutions in the present invention are clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without inventive efforts based on the embodiments of the present invention, shall fall within the scope of protection of the present invention.

Fig. 1 shows that a real-time risk prompting system for driving behavior insurance, provided by one embodiment of the invention, includes: the system comprises a visual information processing unit, a map navigation unit, a driving behavior acquisition unit and a risk driving behavior detection and prompt unit.

The visual information processing unit is used for acquiring image data in front of the vehicle, identifying the conditions of the traffic participation entity and the line pressing conditions of the vehicle from the image data in front of the vehicle, and transmitting the identified conditions of the traffic participation entity and the line pressing conditions of the vehicle to the risky driving behavior detection and prompt unit.

The traffic participation entity condition is identified, and the YOLO v3 deep neural network is adopted for target detection, and compared with other algorithms, the YOLO v3 deep neural network algorithm meets the requirement of the real-time performance of the target detection algorithm in the industrial boundary. Yolo v3 uses a 53-layer network, called Darknet-53.

Specifically, the case of identifying the traffic participant entity from the image data in front of the vehicle includes the steps of:

s1011, acquiring training data;

specifically, frame marking is carried out on the traffic condition in the image data in front of the vehicle manually, and the traffic condition is recorded into a protocol text meeting the deep learning framework requirement specification to obtain a self-owned traffic target detection data set, so that training data are obtained; the traffic participant entities in the data set that need to be marked include: besides collecting data by itself, the data models of various motor vehicles, non-motor vehicles, pedestrians, traffic lights and the like can also comprise typical data sets disclosed on a network, such as a COCO data set, a UA-DETRAC data set and the like. In the embodiment, the deep learning framework requires a standardized data set protocol text to adopt VOC2007;

s1012, enhancing training data;

specifically, the operations such as picture contrast adjustment, random cropping, shifting, rotation and the like are performed on sample data in the traffic data set, so that a plurality of similar sub-samples are derived from a single sample, the number of the samples is increased, the YOLO v3 model is helped to learn the essential characteristics of the sample more easily, and the generalization capability of the AI model is enhanced;

s1013, the traffic data set obtained in the step S1012 is divided into the following parts according to a certain proportion (for example, 7: the method comprises the following steps of (1) a training set, a verification set and a test set, wherein only the training set and the verification set participate in the automatic iterative optimization training process of a model, and the test set is only used for evaluating the generalization effect of the model after the iteration of the model training is terminated;

the invention adopts a YOLO v3 deep neural network YOLO v3 model to carry out target detection, compared with other algorithms, the algorithm of the YOLO v3 deep neural network is more in line with the requirement of the real-time property of the target detection algorithm in the industrial boundary, the YOLO v3 uses a 53-layer network called Darknet-53, and the network structure is shown in figure 2;

s104, training and evaluating a YOLO v3 model by utilizing a training set and a verification set, evaluating the recognition effect of the YOLO v3 model obtained by training through a test set, adjusting the training iteration times, processing the size batch _ size in batch, training the selected iterative optimization algorithm and the like, and continuously improving the recognition effect of the model;

s105, identifying the current states of the vehicle, the pedestrian and the traffic light through the YOLO v3 model trained in the step S104, namely detecting a Bounding Box (Bounding Box) of each traffic participant entity and corresponding signal state information (no light, red light, yellow light and green light);

s106, estimating the distance between the vehicle and the pedestrian in the boundary frame of the entity detected in the step S105 and the camera according to the pinhole imaging model, and outputting and selecting the vehicle with the closest distance and the pedestrian with the closest distance as detection results to obtain the traffic participation entity condition in front of the vehicle when a plurality of vehicles or a plurality of pedestrians exist;

fig. 3 shows an aperture imaging model, where d is the distance from the measured object to the camera lens, f is the focal length of the lens, w is the actual width of the measured object, and w' is the width of the object on the imaging plane (photosensitive element), obtained according to the principle of similar triangle:

f/d＝w′/w (1)

by equation (1), the variant can be obtained:

d＝f×w/w′ (2)

formula (2), the focal length f is known from the lens parameters; the actual width w of the object to be measured can be selected to be a proper empirical value according to the recognized type of the target entity; the width w' of the object on the imaging plane (photosensitive element) is a target entity frame output from the AI recognition result, and is calculated according to the formula (3):

w′＝w_obj_pixels×w _unit (3)

in the formula (3), w _ obj _ pixels is the pixel occupied by the object width, w _unit For a single pixel across the width of the photosensitive element, w _unit Can be calculated from equation (4):

w _unit ＝w _sensor /sensor_w_pixels (4)

in the formula (4), w _sensor The actual width of the photosensitive element of the camera; sensor _ w _ pixels is the total number of pixels in the width direction of the photosensitive element;

integrating the formulas (2), (3) and (4) to obtain the required distance d between the vehicle and the target entity in front of the vehicle:

d＝f×w/(w_obj_pixels×w _sensor /sensor_w_pixels) (5)

calculation examples: a camera with the focal length of 35mm and the width of a photosensitive element of 12.7mm is used for collecting images, and the width of the collected images is 1920 pixels; recognizing that a vehicle target entity exists in front through an AI recognition module, wherein the width of a boundary frame of the vehicle target entity is 100 pixels, and the vehicle distance between the vehicle and the target entity in front of the vehicle is required to be estimated; the calculation was made as the physical width of the identified vehicle 1.8m (1800 mm).

The vehicle distance d to be obtained and various parameters are shown in table 1,

table 1:

according to equation (5), the distance d =95.2 meters is obtained.

If the pedestrian is a pedestrian, the boundary frame of the pedestrian can be set to be 500mm according to experience, and the calculation is carried out by substituting the boundary frame into the formula (5).

If there are a plurality of vehicles and pedestrians, only the vehicle and the pedestrian closest to each other are selected to output.

Specifically, the step of identifying the line pressing condition of the vehicle according to the image data in front of the vehicle comprises the following steps:

s1021, extracting lane lines from RGB image data in front of the collected vehicle;

specifically, the method comprises the following steps:

s10211, cutting the image, intercepting the image in advance based on the prior knowledge that half of the image is sky and background, reducing the calculation amount of the subsequent steps, and reducing unnecessary interference on lane line identification caused by the sky and the background to a certain extent;

s10212, positioning the position of the lane line based on the gradient and the color feature of the image.

S10213, perspective transformation, namely converting the image generated in the S10212 into a bird' S-eye view;

the perspective of the image can be transformed into a bird's eye view using the getPerspectiveTransform and warp perspective functions provided by OpenCV.

S10214, precisely locating the position of the lane line, and counting the number of pixels in each column along the X-axis (from left to right) for the image generated in S10213, and representing the number by a histogram. Wherein the X-coordinate of the peak position corresponds to the lane lines on the left and right sides. In this way, the lane line position information is acquired.

S10215, curve fitting to obtain lane lines on the left side and the right side. Performing curve fitting by using peak points on the left side and the right side of the histogram statistics as initial point coordinates of the left lane line and the right lane line, and respectively fitting pixel points of the left lane line and the right lane line by using a quadratic polynomial, wherein for the pixel points with higher noise, filtering processing is performed or curve fitting is performed by using a random sampling consistency algorithm, so that the lane lines on the left side and the right side are obtained;

s1022, judging whether the lane line is a solid line, if so, entering the next step S1023, and if not, stopping the processing;

specifically, determining whether the lane line is a solid line includes: and respectively detecting the virtual solid lines of the detected lane lines on the left side and the right side, wherein the maximum difference between the solid lines and the virtual lines is whether the pixel points of the lane lines are continuous enough. Counting the values of the binary image of the lane line along the slope direction of the lane line; recording a binary image by using a uint8, wherein only two values are 255 and 0; if the proportion of the binary image of the lane line with the value of 0 to the total lane line pixel proportion is more than 1/3, judging as a dotted line; in addition, if only a single lane line can be recognized in the image, only a single side is detected;

s1023, calculating the center offset of the vehicle lane line according to the condition of the lane line extracted in the step S1021;

specifically, in the image obtained in the calculation step S1022, a coordinate system is established for the image, with the vertex at the upper left corner of the image as an origin (central point), the right side as the positive direction of the x axis, the lower side as the positive direction of the y axis, and the coordinate deviation ratio between the intersection point of the extension lines of the lane lines on the left and right sides and the top of the image and the central point of the image; if only one-side lane line is detected in the image, only the coordinate deviation proportion of the intersection point of the extension line of the one-side lane line and the top of the image and the center of the image is judged;

s1024, judging whether the vehicle presses the line or not according to the central offset, and transmitting a judgment result to a risk driving behavior detection and prompt unit; when the center offset exceeds the threshold value, and the lane line on the line pressing side is marked as a solid line in step S1022, it is determined that the vehicle has a solid line pressing behavior, and otherwise, the line is not pressed or is pressed as a dotted line.

The map navigation unit is used for acquiring vehicle positioning data and navigation data and transmitting the vehicle positioning data and the navigation data to the risk driving behavior detection and prompt unit; specifically, the map navigation unit receives destination input from a user, acquires a map and navigation information, and outputs the map and navigation information to the man-machine interaction module for display, so that a map navigation instruction is provided for a driver; and when the vehicle is in a navigation state, acquiring navigation instructions and road condition information provided by a map navigation cloud service provider from a third party, and outputting the navigation instructions and the road condition information to a risk driving behavior detection and prompt unit after comprehensive processing.

The driving behavior acquisition unit is used for acquiring real-time driver operation data and identifying the operation intention of the driver according to the real-time driver operation data; the method specifically comprises the following steps:

s31, connecting and communicating with an OBD interface of the automobile, periodically collecting driving data according to the standard speed of the OBD communication interface, and storing the driving data in a buffer queue;

the maximum length of the buffer queue is L, which means that L groups of driving data can be stored at most, and the driving data respectively correspond to data of a plurality of historical moments T, T-delta T, T-2 delta T … T- (L-1) delta T which are stored in the queue most recently; delta T is the sampling time interval of the vehicle bus acquisition unit on the vehicle driving data;

each set of driving data includes: instantaneous vehicle speed (Km/Hour), instantaneous brake force (0-100%), instantaneous accelerator opening (0-100%), instantaneous steering wheel rotation amplitude (-100% to + 100%), turn signal state (off, left turn, right turn) and the like;

s32, reading out the driving behavior data from the buffer queue, and calculating and outputting an operation intention of the driver, wherein the operation intention comprises: the method comprises the following steps of (1) accelerating intention, decelerating intention and turning intention, wherein the value of the accelerating intention is {0, 100% }, wherein 0 represents no accelerating intention; the value range of the deceleration intention is {0, 100% }, wherein 0 represents no deceleration intention; the turning intention value range is-100% and 100%, wherein-100 represents the maximum left-turning and 100 represents the maximum right-turning;

specifically, the operation intention calculation includes:

taking out N different time data from T time to T- (N-1) delta T time from the buffer queue for analysis and calculation;

intention to accelerate = Σ (instantaneous throttle opening)/N;

deceleration intent = Σ (instantaneous brake effort)/N;

turn intention = Σ (instantaneous steering wheel rotation amplitude)/N;

wherein, N is the length of the historical data to be read from the buffer queue, the value of N is determined according to the actual requirement, and the data recording time corresponding to N groups of data is usually 0.2-1.5 seconds.

The risk driving behavior detection and prompt unit is used for generating a merged safe driving suggestion according to the condition of the traffic participation entity, the line pressing condition of the vehicle, the vehicle positioning data, the navigation data and the operation intention of the driver, and prompting the driver through the merged safe driving suggestion. The information types to be acquired include: the distance of the nearest vehicle on the front road (unit: meter), the distance of the nearest pedestrian on the front road (unit: meter), the state of the traffic light on the front (red light, yellow light, green light, undetected) and whether the current vehicle presses the solid line (yes, no);

specifically, the generation of the merged safe driving advice includes the steps of:

s41, judging the rear-end collision condition, the pedestrian collision risk and the red light running condition of the vehicle according to the condition of the traffic participation entity identified by the visual information processing unit and the line pressing condition of the vehicle to generate a risk avoidance driving suggestion;

s42, generating a navigation safe driving suggestion according to the vehicle positioning data and the navigation data;

s43, summarizing the risk avoidance driving suggestions and the navigation safety driving suggestions, and removing the repeated safety driving suggestions to obtain the combined safety driving suggestions.

The step S41 of judging the rear-end collision risk condition and generating the risk avoiding driving advice comprises the following steps:

s411, acquiring a current vehicle following distance CurDestance from the visual information processing unit; if the step S411 is executed for the first time, the CurDeistance is stored in the history vehicle following distance Lastdistance, and the step S411 is skipped to the initial position for executing again; if the step S411 is not executed for the first time (i.e. there is a save record in the LastDistance), the subsequent steps are continuously executed;

s412, calculating the following distance variable according to the current following distance CurDestance and the following distance LastDistance recorded at the previous time:

DeltaDistance＝CurDistance-LastDistance；

if the Delta distance is less than 0, further calculating the time of the rear-end collision under the current state:

TimeToCollision＝CurDistance/abs(DeltaDistance)；

if the TimeToCollision is smaller than the threshold TimeCollision threshold (generally, the selected value is 3-5 seconds), the rear-end collision risk can be considered to exist, a risk evasion driving suggestion is generated and sent to the man-machine interaction module, and the driver is prompted in a voice-picture mode.

The step S41 of judging the collision risk condition of the pedestrian and generating the risk avoiding driving advice comprises the following steps of:

s421, acquiring the current distance CurDeistanceWalker between the pedestrian and the visual information processing unit; if the step S411 is executed for the first time, the CurDeistanceWalker is stored in the LastDistantEnker of the historical pedestrian distance, and the step S421 is skipped to the starting position to be executed again; if the step S421 is not executed for the first time (i.e. the lastDistanceWalker has a save record), the subsequent steps are executed continuously;

s422, calculating the distance variable according to the current distance CurDeistanceWalker from the pedestrian and the previous recorded historical distance LastDistanceWaker from the pedestrian:

DeltaDistanceWalker＝CurDistanceWalker-LastDistanceWalker；

if the DeltaDistanceWalker is less than 0, further calculating the time for keeping the collision in the current state:

TimeToCollisionWalker＝CurDistanceWalker/abs(DeltaDistanceWalker)

if the TimeCollision Walker is smaller than the threshold TimeCollision threshold Walker (the value is usually selected to be 3-5 seconds), the pedestrian collision risk can be considered to exist, a risk avoiding driving suggestion is generated and sent to the man-machine interaction module, and the driver is prompted in a voice-picture mode.

Specifically, the step S41 of determining the rear-end collision risk condition to generate the risk avoidance driving recommendation includes: and judging whether the vehicle runs the red light according to the image data in front of the vehicle, and if the judgment result is that the vehicle runs the red light or the judgment result in the step S1 is that the vehicle presses the solid line, generating a risk avoiding driving suggestion with the risk of violating the traffic rules.

S42, acquiring vehicle positioning data and navigation data, and generating a navigation safe driving suggestion according to the vehicle positioning data and the navigation data;

the navigation information comprises current road speed limit information, navigation action information and the like, and the navigation action information comprises: vehicle turning information, passing intersection information, rotary island information and the like.

Specifically, the generating of the navigation safe driving suggestion according to the navigation information includes: and generating safe driving suggestions such as deceleration, ahead turning on a turn signal, lane changing and the like according to the road speed limit and navigation action information in the map navigation unit and by comparing the road speed limit and navigation action information with the current driving data of the vehicle of the driving behavior acquisition unit.

For example: if the speed limit of the current road is 60Km/h and the current vehicle speed is 80Km/h (which is higher than the speed limit) obtained from the navigation information, a safe driving behavior suggestion for prompting the driver to decelerate is generated.

Wherein, step S43 specifically is: and summarizing the risk avoiding driving advice with the rear-end collision risk, the risk avoiding driving advice with the pedestrian collision risk, the risk avoiding driving advice with the risk of violating the traffic rules and the navigation safety driving advice obtained in the step S41, removing repeated safety driving advice to obtain combined safety driving advice, sending the combined safety driving advice to a man-machine interaction module, prompting a driver in a sound-picture mode, and simultaneously recording the generation time of the safety driving advice by the system for each safety driving advice in the combined safety driving advice. The real-time risk prompting system facing the driving behavior insurance is used for continuously and periodically executing detection actions, and the latest safe driving suggestion can cover the previous suggestion result. Therefore, if the risk condition disappears, the corresponding driving advice is cancelled, the human-computer interaction module stops displaying the corresponding risk driving information, and the generation time record of the corresponding safe driving advice is cleared.

The merged safe driving advice from which the duplicated safe driving advice is removed is shown in table 1.

Table 1 merging safety driving advice checklists

Risky behavior	Driving advice content	Correcting time limit (according to actual conditions)
			Overspeed driving	Brake/release valve deceleration	T overspeed
Vehicle rear-end collision	Brake/release valve deceleration	T-shaped rear-end collision
			Pedestrian collision	Brake/release valve deceleration	T pedestrian
Run red light	Brake/release valve deceleration	T red light
			Left/right solid line	Steering wheel for steering in reverse direction	T compaction line

In one embodiment, the risky driving behavior detection and prompt unit is further configured to compare the merged safety driving advice with real-time driver operation data, determine whether the driver executes the merged safety driving advice, such as braking deceleration and turning a turn signal, if not, determine that the risky driving behavior exists, and send the event to the cloud uploading unit. Specifically, the method comprises the following steps:

s51, vehicle bus data and real-time driver operation data are obtained through a driving behavior acquisition unit;

s52, checking whether the user executes behaviors in the combined safe driving advice item one by one, and clearing the safe driving advice item if the judgment result of a certain safe driving advice item is execution; otherwise, the subsequent step S53 is continuously performed.

S53, checking all safety driving suggestions which are still not executed, and comparing the current time with the generation time of the combined safety driving suggestion to obtain the duration T _ ignore of the safety driving suggestion which is not executed;

s54, if the continuous non-execution time T _ align of a certain safe driving suggestion is greater than the overtime threshold (usually 3-5 seconds) of the combined safe driving suggestion, the dangerous driving event is considered to occur, the overtime dangerous driving event is sent to a cloud uploading unit, and meanwhile, the generation time record of the safe driving suggestion is emptied;

and S55, repeating S51-S54 until all the merged safe driving suggestions are timed out or cleared.

Step S5 will be described by taking the detected risk behavior as overspeed driving as an example:

a51, reading brake and speed information of the vehicle;

a52, checking whether the user executes the merged safe driving suggestion obtained in the S4, namely judging that the opening of the accelerator is =0 or the braking force is >0; if the driver executes the merged safe driving suggestion, the comparison task is ended; otherwise, continuing to execute the subsequent step A53;

a53, reading the latest safe driving suggestion, and continuing to execute the task when the safe driving suggestion of A52 still exists; if not, ending the comparison task;

and A54, if the driver does not execute the merged safe driving suggestion in the S4, the final comparison task (safe driving suggestion comparison in the A53) is overtime, the system identifies a dangerous driving event with overspeed and sends the dangerous driving event to the cloud server.

In one embodiment, in the existing driving behavior analysis system, the vehicle body dynamic abnormity is inferred to be caused by the violent driving behavior, and the violent driving behavior is further considered to cause higher driving risk, so that the risk probability is increased.

The method for deducing the driving behavior by simply adopting the abnormity is a common conventional method at present, has high efficiency and rationality, and can bring about a certain degree of misjudgment. The reason for the erroneous determination is that, in addition to the driving behavior of the driver, the response dynamic characteristics of the vehicle itself and the objective conditions of the road also have a great influence on the final vehicle body dynamic characteristics, including: the nonlinearity of the gain of the vehicle power along with the stepping depth of an accelerator, the nonlinearity of the gain of the brake strength along with the stepping depth, the nonlinearity of the gain of the steering along with the speed along with the angle, the bad flatness condition of the running road and the like.

Considering the huge base number of potential markets, even if there is a very small proportion of misjudgments, the absolute number is not negligible, causing unnecessary trouble to users whose user behavior is insured. Therefore, in the process of judging the risk driving behavior by the system, enough evidence and information should be collected as far as possible, and false alarm is reduced to the minimum.

Therefore, the driving behavior insurance oriented risk real-time prompt vehicle body dynamic abnormity detection unit is used for acquiring vehicle driving state information, judging whether the vehicle has vehicle body dynamic abnormity according to the vehicle driving state information, and if the vehicle body dynamic abnormity exists, sending a judgment result to the risk driving behavior detection and prompt unit. Wherein, the vehicle body dynamic anomaly includes: rapid acceleration, rapid deceleration, and sharp turns. Specifically, the step of judging whether the vehicle has dynamic vehicle body abnormality according to the vehicle driving state information comprises the following steps:

s61, acquiring triaxial acceleration data acquired by a vehicle pose sensing module through a vehicle body dynamic detection unit, wherein the triaxial acceleration data comprise longitudinal (advancing direction) acceleration of a vehicle, longitudinal (advancing direction) deceleration of the vehicle and transverse (centrifugal force direction, direction perpendicular to the advancing direction of the vehicle and parallel to the ground) acceleration of the vehicle;

s62, filtering the triaxial acceleration data;

considering that the vehicle pose sensing module acquires data through a sensor, and the data acquired by the sensor has certain noise fluctuation, which easily causes misjudgment of a subsequent algorithm or frequent jitter of a result, and the data needs to be filtered; filtering systems include, but are not limited to: the mean filtering, band-pass filtering and specific implementation structure can be implemented by FIR and IIR filters, and if the statistical distribution of the signals is known, statistical filtering such as maximum posterior and maximum likelihood filtering can also be used. Filtering does not adopt an excessively complex system so as to avoid generating larger time delay;

s63, performing threshold value judgment on the filtered triaxial acceleration data to judge whether the vehicle is subjected to rapid acceleration, rapid deceleration and rapid turning, and judging that the vehicle body is in dynamic abnormity if one of the rapid acceleration, the rapid deceleration and the rapid turning exists;

wherein, the rapid acceleration is judged as follows: if the longitudinal (forward) acceleration of the vehicle exceeds a first threshold value LimitAcc and the duration time reaches a first threshold time TimeOverAccLimit, the vehicle is accelerated suddenly, otherwise, the vehicle does not have the sudden acceleration;

the rapid deceleration is judged as follows: if the longitudinal (forward) deceleration of the vehicle exceeds a second threshold value LimitDeAcc and the duration time reaches a second threshold time TimeOverDeAccLimit, the vehicle is suddenly decelerated, otherwise, the vehicle does not have the sudden deceleration;

the sharp turn is judged as follows: the vehicle turns sharply if the acceleration in the transverse direction (the direction of the centrifugal force, perpendicular to the direction of forward movement of the vehicle, parallel to the ground) exceeds the third threshold limit turnacct and the duration reaches the third threshold time TimeOverTurnAccLimit, otherwise the vehicle does not turn sharply;

the threshold value can be generally set according to actual conditions, such as:

LimtAcc＝0.5G，

LimitDeAcc＝0.5G，

LimitTurnAcc＝0.6G，

TimeOverAccuLimit =0.3 seconds

Timeoverlaverdeaclimit =0.6 seconds

TimeOverTurnAccLimit =1 second.

And S64, judging whether the dynamic abnormality of the vehicle body obtained in the step S63 is a risk avoidance behavior, if the risk avoidance behavior of the driver generates abnormal dynamic state due to risk avoidance, determining that the vehicle body is a risk driving behavior event, otherwise, determining that the risk driving behavior does not belong to the risk avoidance behavior event, and sending a judgment result to a cloud server.

Specifically, step S64 includes the following steps:

s6401, acquiring vehicle body dynamic data of a vehicle body dynamic abnormity detection unit, wherein the vehicle body dynamic data comprises four conditions: case 1, no anomaly; case 2, rapid acceleration; case 3, rapid deceleration; case 4, sharp turn;

s6402, judging the dynamic data of the vehicle body, and if the dynamic abnormal data of the vehicle body is the condition 1, performing no subsequent processing, so that the vehicle has no abnormal dynamic state; if the dynamic abnormal data of the vehicle body is the condition 2-4, subsequent filtering is required, and the step S6403 is carried out;

s6403, checking whether the driving behavior corresponding to the vehicle body dynamic data in the step S6401 belongs to the active behavior of the driver, if so, entering the step S6404 to filter again, otherwise, determining a vehicle body dynamic abnormal event;

the method specifically comprises the following steps: acquiring the current calculated driver intention according to a driving behavior acquisition unit, judging the driver intention, and if the corresponding intention value is greater than an intention threshold value, indicating that the driver has obvious violent driving intention and the dynamic abnormality of the vehicle body is caused, and further processing and filtering in subsequent steps, namely S74 filtering; otherwise, if the driver does not find obvious intention, the vehicle body dynamic abnormal event is not considered as the risk driving behavior event. Various intention thresholds need to be adjusted and set according to actual conditions, and even dynamic adjustment is carried out according to the vehicle speed;

taking sharp turning driving as an example, if the absolute value of the turning intention is greater than the turning intention threshold, it indicates that the driver has obvious turning intention, resulting in dynamic abnormality of the vehicle body, and the subsequent step S6404 needs to be performed for filtering;

s6404, judging whether the vehicle body dynamic abnormal event is a safe driving suggestion for executing merging, if so, judging that the vehicle body dynamic abnormal event is a normal driving behavior event, otherwise, judging that the vehicle body dynamic abnormal event is a risk driving behavior event, and reminding through a man-machine interaction module;

s6405, after the steps S6401-S6404, the subsequent vehicle body dynamic abnormal events (rapid acceleration, rapid deceleration and rapid turning) are determined as risk driving behavior events and are sent to the cloud uploading unit. Some abnormal dynamic of the vehicle body caused by the emergency operation can be understood, and the abnormal dynamic is related to the dynamic characteristics of the vehicle, and does not represent the behavior style of violent driving of the driver.

Fig. 4 shows a structure of a driving behavior insurance-oriented risk real-time prompting device according to an embodiment of the present invention, where the risk real-time prompting device is applied to a risk real-time prompting system, and for convenience of description, only the parts related to the embodiment of the present invention are shown.

A risk real-time prompting device facing driving behavior insurance comprises: the system comprises an AI vision processing module 7, a vehicle positioning module 2, a vehicle bus data acquisition module 3, a main control module 1 and a man-machine interaction module 5; the main control module 1 is respectively and electrically connected with the vehicle positioning module 2, the vehicle bus data acquisition module 3, the AI vision processing module 7, the vehicle pose acquisition module 4 and the human-computer interaction module 5. The main control module 1, the vehicle positioning module 2, the vehicle bus data acquisition module 3, the AI vision processing module 7 and the human-computer interaction module 5 are all arranged on the vehicle.

The AI vision processing module 7 is configured to acquire image data in front of a vehicle, and identify a traffic participant entity condition and a vehicle line pressing condition for the image data in front of the vehicle. The AI vision processing module 7 is respectively connected with the camera 8 and the main control module for obtain the image data in front of the vehicle through the camera 8, concretely, the AI vision processing module 7 adopts Nvidia Jetson TX2, hectometrical EdgeBoard and the like, the camera 8 adopts the model as a QR-USB3MP01H high definition camera 8 module, and the integrated high dynamic sensitization chip AR0331 of this module is suitable for the backlight work. The camera 8 is installed in the middle of the windshield of the vehicle, the data acquired by the camera 8 are transmitted to the AI vision processing module 7, and the data acquired by the camera 8 are processed by the AI vision processing module 7, for example, the information of the vehicle, the pedestrian, the lane line, the traffic signal lamp and the like in front of the vehicle are extracted and taken out, and the setting can be performed as required.

The vehicle positioning module 2 is used for acquiring positioning information of a vehicle; the vehicle positioning module 2 adopts a positioning module with output data meeting NEMA-0183 standard, and main information provided by the MEMA-0183 standard comprises: recommending minimum positioning information GPRMC, ground speed information GPVTG, GPS positioning information GPGGA and current satellite information GPGSA. Specifically, the vehicle positioning module 2 adopts SKG12D or BN-280 (based on UBX-M8030-KT or a chinese micro AT6558R chip), and the vehicle positioning module 2 is connected with the main control module 1 through a UART serial communication interface, for example, the main control module 1 can obtain longitude and latitude information of a vehicle from a GPGGA, and obtain a heading angle (0 ° in the north-positive direction) of the vehicle and a movement rate of the vehicle from a GPVTG, that is, the vehicle positioning module 2 can obtain accurate positioning of the vehicle with low delay, and accurate and reliable positioning is a basis for realizing navigation of a system and obtaining road information in front of the vehicle.

The vehicle bus data acquisition module 3 is used for acquiring real-time driver operation data; specifically, the vehicle bus data acquisition module 3 adopts PIC18F25K80-ELM327, is connected with the vehicle bus through an OBD interface, acquires vehicle state information, sends data to the main control module 1 through a UART serial port, and then controls and saves and processes the data of the vehicle bus data acquisition module 3 through the main control module 1. The data collected by the vehicle bus data collection module 3 includes: steering wheel action, throttle action, brake action, and the status of the turn lights, vehicle speed, engine speed, and mileage information.

The main control module 1 is respectively and electrically connected with the AI vision processing module 7, the vehicle positioning module 2, the human-computer interaction module 5 and the vehicle bus data acquisition module 3, and the main control module 1 is used for generating a merged safe driving suggestion according to the vehicle information transmitted by the AI vision processing module 7, the vehicle positioning module 2 and the vehicle bus data acquisition module 3; specifically, main control module 1 adopts the kainit kylin 970 chip to support the CAN bus (CAN be used to connect the vehicle bus), MIPI and USB interface (camera 8), support wired network, wireless WiFi and GPS location, possess display interface and audio interface, support and give the user with image and audio mode with system information feedback, for vehicle orientation module 2, vehicle bus data acquisition module 3 and vehicle position appearance acquisition module 4, human-computer interaction module 5, telecommunication module 6, AI vision processing module 7, speaker module, LED lamp and camera 8 etc. provide the interface. The main control module 1 is responsible for communication connection with all other hardware modules, bears the execution of each unit function, and completes the realization of the whole function of the system. The main control module 1 supports the operation of a Linux/Android operating system, has a high-performance memory interface and a complete peripheral interface, and is an SoC on-chip embedded system based on an ARM instruction framework.

And the man-machine interaction module 5 is connected with the main control module 1 and is used for displaying and prompting a driver according to the merged safe driving. Specifically, the human-computer interaction module 5 may use various touch screen models with standard interfaces, is connected with the main control module 1 through interfaces such as HDMI/I2C/I2S, sets the main control module 1 to display a graphical interface as required, controls the main control module 1 through a touch mode, and the like.

The working principle of the driving behavior insurance-oriented risk real-time prompting device is that the main control module 1 generates a merged safe driving suggestion through data obtained by the AI vision processing module 7, the vehicle positioning module 2 and the vehicle bus data acquisition module 3 at the same time.

According to one embodiment, the driving behavior insurance-oriented risk real-time prompting device further comprises a vehicle pose acquisition module 4, the vehicle pose acquisition module 4 is electrically connected with the main control module 1, and the vehicle pose acquisition module 4 is used for detecting whether a vehicle body dynamic abnormal event exists according to the vehicle driving state information and sending a detection result to the main control module 1. Specifically, the vehicle pose acquisition module 4 adopts an attitude sensor, the specifically adopted model is LPMS-ME1, the vehicle pose acquisition module 4 is connected with the main control module 1 through an I2C or UART serial port, and the dynamic driving information comprises: linear acceleration and turning acceleration of the vehicle.

According to one embodiment, the driving behavior insurance-oriented risk real-time prompting device further comprises a remote communication module 6, wherein the remote communication module 6 is electrically connected with the main control module 1. Specifically, the remote communication module 6 adopts an ME909S-821 4G full-network wireless communication module, and is connected with the main control processing module through a PCIE interface, and the main control module 1 uploads driving behavior data to a server platform such as a business cloud server provided by an insurance service through the remote communication module 6, or obtains a navigation route prompt, a front road condition, intersection information, road speed limit information and the like from a cloud map/navigation service provider through the remote communication module 6.

According to one embodiment, the real-time risk prompting device for driving behavior insurance further comprises a loudspeaker module, the loudspeaker module is electrically connected with the main control module 1, the loudspeaker module adopts a loudspeaker, the main control module 1 is set, and voice prompt of risk conditions and the like is carried out through the loudspeaker.

In one embodiment, the real-time risk prompting device for driving behavior insurance further comprises an LED lamp, the LED lamp is electrically connected with the main control module 1, the main control module 1 is set, and risk prompting is performed through flickering of LEDs and the like.

The working principle of the real-time risk prompting device for driving behavior insurance of the invention is explained in detail, so that the technical personnel in the field can understand the invention more:

fig. 5 shows a preferred embodiment of the driving behavior insurance-oriented risk real-time prompting device provided by the present invention, and the driving behavior insurance-oriented risk real-time prompting device includes: the system comprises a main control module 1, a vehicle positioning module 2, a vehicle bus data acquisition module 3, a vehicle pose acquisition module 4, a human-computer interaction module 5, a remote communication module 6, an AI vision processing module 7 and a camera 8.

The vehicle positioning module 2 acquires positioning data of a vehicle and transmits the positioning data to the main control module 1; the vehicle bus data acquisition module 3 acquires vehicle state information through a vehicle bus, for example: the action of a steering wheel, the action of an accelerator, the action of a brake, the state of a steering lamp, the vehicle speed, the engine speed, the traveling mileage information and the like are transmitted to the main control module 1; the vehicle pose acquisition module 4 acquires vehicle dynamic driving information, for example: the linear acceleration, the turning acceleration and the like of the vehicle are transmitted to the main control module 1; the main control module 1 judges risks according to data of the vehicle positioning module 2, the vehicle bus data acquisition module 3 and the vehicle pose acquisition module 4, and carries out risk reminding through the human-computer interaction module 5; the AI vision processing module 7 is installed in the middle of the front windshield of the vehicle through collecting and processing data of the camera 8, extracts visual information related to safe driving, such as lane lines, traffic light states, front vehicles, pedestrians and the like, and transmits the information to the main control module 1, the main control module 1 judges according to the AI vision processing module 7 or displays or reminds and the like through the human-computer interaction module 5, and can remind a driver more conveniently through the loudspeaker module and the LED, and the attention of the driver is reduced.

In conclusion, the invention provides a risk real-time prompting system and device for driving behavior insurance, which can make the judgment of a main control module more timely by combining a collection module and a unit of vehicle state related information, and is more convenient for a driver to carry out reminding through a human-computer interaction module, a loudspeaker module and an LED lamp, thereby reducing the risk of driving.

The above disclosure is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of changes or modifications within the technical scope of the present invention, and shall be covered by the scope of the present invention.

Claims

1. A risk real-time prompt system for driving behavior insurance is characterized by comprising the following components:

the step of identifying the line pressing condition of the vehicle by the image data in front of the vehicle comprises the following steps: s1021, extracting lane lines from RGB image data in front of the collected vehicle; s1022, judging whether the lane line is a solid line, if so, entering the next step S1023, and if not, stopping the processing; s1023, calculating the center offset of the vehicle lane line according to the condition of the lane line extracted in the step S1021; s1024, judging whether the vehicle presses the line or not according to the central offset, and transmitting a judgment result to a risk driving behavior detection and prompt unit; when the central offset exceeds the threshold value and the lane line on the line pressing side is marked as a solid line in the step S1022, the vehicle is determined to have a solid line pressing behavior, otherwise the line is not pressed or the pressed line is a dotted line;

the system comprises a risk driving behavior detection and prompt unit, a traffic information acquisition unit and a safety information acquisition unit, wherein the risk driving behavior detection and prompt unit is used for generating a combined safety driving suggestion according to the condition of the traffic participation entity, the vehicle line pressing condition, vehicle positioning data, navigation data and the operation intention of a driver, and prompting the driver through the combined safety driving suggestion;

the dangerous driving behavior detection and prompt unit is used for generating a merged safe driving suggestion according to the condition of the traffic participation entity, the line pressing condition of the vehicle, the vehicle positioning data, the navigation data and the operation intention of the driver, and specifically comprises the following steps:

and summarizing the risk avoidance driving suggestions and the navigation safety driving suggestions, and removing repeated safety driving suggestions to obtain combined safety driving suggestions.

2. The driving behavior insurance-oriented risk real-time prompting system of claim 1, wherein the risky driving behavior detection and prompting unit is further configured to compare the merged safety driving advice with real-time driver operating data, determine whether the driver executes according to the merged safety driving advice, determine that risky driving behavior exists if the driver does not execute according to the merged safety driving advice, and send the event to the cloud uploading unit.

3. The driving behavior insurance-oriented risk real-time prompting system according to claim 1, further comprising a vehicle body dynamic anomaly detection unit, wherein the vehicle body dynamic anomaly detection unit is used for acquiring vehicle driving state information, detecting whether a vehicle body dynamic anomaly event exists according to the vehicle driving state information, and sending a judgment result to the risk driving behavior detection and prompting unit.

4. The driving behavior insurance oriented risk real-time prompting system according to claim 1, wherein the merged safety driving advice is to continuously and periodically perform a detection action, and the latest merged safety driving will override the previous merged safety driving.

5. A real-time risk prompting device for driving behavior insurance, which is applied to the real-time risk prompting system of claims 1-4, and comprises:

the main control module is respectively and electrically connected with the AI vision processing module, the vehicle positioning module and the vehicle bus data acquisition module, and is used for generating a merged safe driving suggestion according to the vehicle information transmitted by the AI vision processing module, the vehicle positioning module and the vehicle bus data acquisition module; and

and the human-computer interaction module is connected with the main control module and is used for displaying and prompting a driver according to the merged safe driving.

6. The driving behavior insurance-oriented risk real-time prompting device according to claim 5, further comprising a vehicle pose acquisition module, wherein the vehicle pose acquisition module is electrically connected with the main control module, and the vehicle pose acquisition module is used for detecting whether a vehicle body dynamic abnormal event exists according to the vehicle driving state information and sending a detection result to the main control module.

7. The real-time risk prompting device for driving behavior insurance recited in claim 5, further comprising a remote communication module electrically connected with the main control module.