A driver assistance system that identifies and warns of dangerous and fatigued driving behavior

Through the fuzzy inference system combined with sensor data, accurate identification and warning of driver hazards and fatigue driving behaviors is achieved, and the problems of high cost and low recognition accuracy in the prior art are solved, and the universality and safety of the system are improved.

CN114852088BActive Publication Date: 2024-11-19NANJING UNIV OF POSTS & TELECOMM
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210607635.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-05-31
Publication Date
2024-11-19
Estimated Expiration
2042-05-31

AI Technical Summary

Technical Problem

When detecting drivers' dangerous and fatigued driving behavior, the prior art has high costs, limited recognition accuracy, and cannot effectively distinguish short-term and long-term data, resulting in one-sided judgment results and limited applicable scenarios, making it difficult to timely warning and detect fatigued driving behavior.

Method used

A driver assistance system based on fuzzy reasoning is adopted to collect vehicle behavior, driver status and road conditions information through the sensor layer to form low-dimensional feature vectors, and a fuzzy reasoning system is used to make short-term and long-term judgments, combining high-frequency sound, high-speed vibration and picture prompts. Make a warning and adjust the threshold through the driving control layer to adjust the false alarms to adaptively.

Benefits of technology

It realizes low-cost and accurate identification of dangerous and fatigue driving behaviors, reduces detection costs, takes into account the detection of short-term dangerous and long-term fatigue behaviors, improves the universality and safety of the driver assistance system, and reduces the occurrence of traffic accidents.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114852088B_ABST
    Figure CN114852088B_ABST
Patent Text Reader

Abstract

A driver assistance system and method for identifying and warning of dangerous and fatigue driving behaviors, which does not require the collection of normal driving characteristics and known dangerous driving characteristics, nor does it require the training of detection models, thereby reducing implementation costs and being more universal; combining three factors, namely, road conditions, driver status, and vehicle driving characteristics, and having an adaptive adjustment mechanism, which fits the driving styles of different drivers, making the judgment of driving behavior more comprehensive and more valuable for reference, and making the detection more accurate and effective; taking into account both the timely early warning discovery of dangerous behaviors and the detection of hard-to-detect fatigue driving behaviors, which can effectively assist the driver in driving safely; causing little interference to the driver, effectively avoiding traffic accidents caused by dangerous and fatigue driving, helping the driver to discover slowly changing driving behaviors, and making the driver aware of the danger in a timely manner when the driver's driving level and decision-making level are reduced, thereby effectively assisting the driver in driving safely.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of vehicle driving assistance, and in particular relates to a driver assistance system and method for identifying and warning of dangerous and fatigue driving behaviors. Background Art

[0002] With the growing size of the automobile market, the pressure on the carrying load of urban roads and highways is also increasing, resulting in increasingly serious traffic congestion and traffic accidents, causing serious casualties and economic losses. But at the same time, the development of smart cities, vehicle networking, and autonomous driving has also attracted a lot of attention, and driver assistance as the second level of smart vehicles has been included in the research. By statistically analyzing the vehicle information, driver information or road condition information obtained by the sensor, a logical judgment is made on the driver's behavior, and finally the driver's behavior is supervised and assisted to achieve safer driving. In reality, many dangerous behaviors are carried out in a short period of time and threaten the safety of other road users, but most of these violations cannot be tracked at present, and many of them can only be corrected after a traffic accident occurs.

[0003] Taking dangerous driving as the object of investigation, many existing technologies (patent numbers CN106611169A, CN110171426B, CN112043280A, CN113762213A, CN110991353A, CN113033261A) mainly use machine learning methods to detect dangerous behaviors of drivers. In the above technologies, in order to ensure the effectiveness of the detection model, a large amount of dangerous driving data needs to be collected, and real-time collection and comparison are required. Whether it is image or data sequence comparison, the cost is high and the implementation cost is high. At the same time, since most of them only consider a few dangerous driving situations, they are greatly affected by recognition accuracy, driver's physiological habits, clothing, lighting, etc., and there are certain constraints on effectiveness. In addition, the collected data is generally limited to vehicle data or driver data, and no comprehensive investigation is conducted, which makes the judgment result one-sided and the applicable scenarios limited. Finally, most of the existing research focuses on short-term data identification, while long-term data is not taken into consideration. Most studies that include fatigue driving only regard fatigue driving as a type of dangerous driving, making its detection not targeted. Fatigue driving, as a driving behavior that is difficult to detect, can only be reflected in long-term driving data. Summary of the invention

[0004] To make up for the shortcomings of the prior art, the present invention provides a method for identifying and warning drivers' dangerous driving and fatigue driving behaviors based on fuzzy reasoning. The method has low detection cost and can distinguish between dangerous driving behaviors in the short term and fatigue driving behaviors that can only be discovered through long-term judgments. It takes into account both the timely warning discovery of dangerous behaviors and the detection of difficult-to-detect fatigue driving behaviors.

[0005] A driver assistance system for identifying and warning of dangerous and fatigued driving behaviors, comprising a sensor layer, a perception layer, a cognitive layer, a warning execution layer and a drive control layer connected in sequence, wherein the drive control layer is connected to the cognitive layer;

[0006] The sensor layer collects information about vehicle behavior, driver status, and road conditions through sensor perception;

[0007] The perception layer processes the initial information obtained by the sensor to form a low-dimensional feature vector corresponding to the vehicle;

[0008] The cognitive layer is equipped with a fuzzy reasoning system to make a short-term real-time judgment on whether the vehicle characteristic information is dangerous driving. If there is no dangerous driving behavior in the short term, the short-term driving data will be used to make a long-term behavior judgment, including the long-term following distance change and the driver's own condition. If it meets the fatigue driving characteristics, it will be judged as fatigue driving.

[0009] The warning execution layer, based on the judgment of the cognitive layer, issues high-frequency sound prompts to surrounding vehicles and pedestrians if it is emergency and dangerous driving, and issues high-frequency sound, high-speed vibration, and visual prompts to the target vehicle. If it is emergency and dangerous driving, the driver control layer is fed back to adjust the threshold. If it is ordinary and dangerous driving, the driver of the target vehicle is warned. If it is fatigue driving, visual and voice prompts are issued to remind the driver to take a rest and concentrate on driving.

[0010] The driving control layer makes a choice between the system's false positive and false negative judgments based on the driver's judgment of his own driving characteristics and the occurrence of emergency situations, thereby achieving adaptive adjustment of the dangerous and fatigue driving judgment thresholds.

[0011] Furthermore, the sensor layer includes a steering wheel angle sensor, a Beidou satellite navigation system, a vehicle inertial measurement unit, a wearable accelerometer and heart rate meter, a radar and a computing device.

[0012] A driver assistance method for identifying and warning of dangerous and fatigue driving behaviors comprises the following steps:

[0013] Step 1: Use sensors to sense vehicle behavior, driver status, and road conditions;

[0014] Step 2: Process the initial information obtained by the sensor to form a low-dimensional feature vector corresponding to the vehicle;

[0015] Step 3: Make a short-term real-time judgment on the vehicle characteristic information. If there is an emergency dangerous driving behavior that meets the characteristics in the short term, the driver, surrounding vehicles, and pedestrians on the road will be warned immediately by high-frequency sound, high-speed vibration, and image.

[0016] Step 4: Use the fuzzy inference system to make a short-term real-time judgment on the vehicle feature information. If there is a common dangerous driving behavior that meets the characteristics, the driver will be warned immediately by high-frequency sound, high-speed vibration and screen images;

[0017] Step 5: If there is no dangerous driving behavior in the short term, the short-term driving data is used to judge the long-term behavior, mainly based on the long-term following distance and the driver's own condition. If it meets the characteristics of fatigue driving, the driver will be reminded to take a rest through vibration and screen;

[0018] Step 6: If a false alarm confirmed by the driver occurs, an adaptive adjustment is made within a small range that does not affect safety. If an emergency dangerous situation occurs, the threshold is reset. The driver can also reset the threshold to adapt to the driving style of different drivers.

[0019] Furthermore, in step 1, steering wheel angle change information, the horizontal and vertical coordinates of the vehicle relative to the road, vehicle acceleration change information, driver's motion swing and heart rate status, front vehicle distance information and road condition information are obtained.

[0020] Furthermore, the step 2 specifically includes the following steps:

[0021] Step 2-1, use the angle sequence of the steering wheel angle change information to obtain the vehicle lateral sway characteristics; log the steering wheel angle sequence {θ n}, where n represents the time-related stream data sequence number, and performs fast Fourier transform to obtain its power spectrum density:

[0022] F(x) = FFT{θ n}

[0023] Then sum the amplitudes to get the corresponding swing value S:

[0024]

[0025] S is a vehicle lateral sway characteristic that reflects the amplitude and frequency of the vehicle's lateral movement. The larger its value is, the more frequent and larger the amplitude of the vehicle's lateral sway is, that is, the more unstable the driving state is;

[0026] Step 2-2, using the horizontal and vertical coordinates of the vehicle relative to the road to solve the longitudinal speed information parallel to the road and the lateral speed information perpendicular to the road, and averaging them within the same time length as step 2-1;

[0027] Step 2-3, divide the acceleration of the vehicle into a certain range, divide the acceleration range into four right-angle sectors, front, back, left, and right, and determine three acceleration ranges to form 16 acceleration sectors, which are numbered with 4 bits to determine the maximum safe acceleration value a in the four directions of front, back, left, and right. fm 、a bm 、a lm 、a rm , and divide the value into four acceleration sectors according to the step length of one third of this value, forming a series of numerical sequences during the vehicle driving process. The acceleration thresholds of the front and rear are different, and the acceleration thresholds of the left and right are the same, that is, a fm >a bm >a lm =a rm ;

[0028] Step 2-4, determining the percentage difference between the current heart rate and the normal heart rate according to the driver's calm heart rate, and obtaining the driver's heart rate deviation as the driver's swing feature;

[0029] Step 2-5, averaging the obtained following distances, and saving the average following distance every minute;

[0030] Step 2-6, obtain traffic information, determine whether the road is congested or not, determine the wetness of the road surface based on weather conditions, and infer the friction conditions. The more congested the traffic, the lower the road friction, and the higher the danger level, the higher this value will be.

[0031] Furthermore, in step 2,<F1,F2,F3,F4,F5,F6,F7,F8> , that is, the order of the corresponding values ​​of each feature constitutes the driving state feature to be detected, and the next step of judgment is made; among them, F1 is the vehicle lateral swing feature; F2 is the vehicle absolute speed information; F3 is the vehicle lateral speed information; F4 is the vehicle acceleration block; F5 is the driver's swing feature; F6 is the driver's heart rate deviation; F7 is the vehicle's short-term forward following distance; F8 is the road condition coefficient.

[0032] Furthermore, the step 3 specifically includes the following steps:

[0033] Step 3-1: Real-time tracking and judging the following vehicle distance. Different speed ranges have different warning thresholds. The standard is divided into three levels. When the relative speed is 15-20m / s, the collision threshold is 3.5s, when it is 10-15m / s, it is 3s, and when it is less than 10m / s, it corresponds to 2.5s. If it is within the threshold warning range, it is directly judged as emergency dangerous driving;

[0034] Step 3-2: Analyze the vehicle acceleration block. If the zone is rearward and the acceleration value is greater than the maximum rearward safety acceleration value a bm , that is, if it is in the rear four zones, the driver is judged to have taken emergency braking behavior, if it is in the left and right four zones, it is a sharp turn, and if it is in the front four zones, it is a sudden acceleration, all of which are judged as emergency and dangerous driving;

[0035] Step 3-3, except for the above exceptions,<F1,F2,F4,F5,F6,F7> All of them are included in the fuzzy reasoning framework to judge dangerous driving behaviors with unclear characteristics. For these six characteristics, the average within a few seconds is taken, and then weighted according to the F8 road congestion, weather conditions, and road slipperiness. Then, they are normalized according to the maximum value of the corresponding characteristic attributes of the surrounding vehicles;

[0036] Step 3-4, judging whether there is dangerous driving behavior according to the dangerous driving index, if it exceeds the threshold value, it is judged as ordinary dangerous driving;

[0037] Step 3-5, collect the average following distance and vehicle lateral sway characteristics in units of 1 minute. If the average following distance continues to decrease and the vehicle lateral sway continues to increase within ten sets of data, it is determined that the driver has fatigue driving behavior;

[0038] Step 3-6, collect the dangerous driving index within 10 minutes. If the average value is greater than a certain threshold or all values ​​are less than the dangerous driving threshold, it is determined to be fatigue driving.

[0039] Further, in step 3-3, a fuzzy reasoning framework based on Mamdani is used, including a fuzzy generator, a fuzzy reasoning engine, a fuzzy eliminater and a fuzzy rule base connected to the fuzzy reasoning engine in sequence;

[0040] In the fuzzy reasoning framework, appropriate fuzzy linguistic variables and corresponding membership functions are assigned to the corresponding attributes of F1, F2, F4, F5, F6, and F7. Each attribute corresponds to two high and low membership functions. Fuzzy logic rules are then written for the relationship between the six elements. The vehicle has more lateral sway, faster absolute speed, larger acceleration value, larger driver swing, more heart rate deviation, and closer following car, which are all positive contributions to the increase of dangerous driving index. The dangerous driving probability obtained by fuzzy reasoning is high, that is, the judgment result is that the driver has a high probability of dangerous driving behavior. On the contrary, the reduction of the six attributes has a negative contribution to the increase of dangerous driving index. It is judged that the driver's driving is relatively stable and safe, thus determining 2 6 That is, 64 fuzzy reasoning rules correspond to the dangerous driving index, which is divided into three parts of membership functions: high, medium and low;

[0041] Fuzzy reasoning and judgment are performed according to fuzzy rules. In fuzzy logic, the intersection operation of μ(u) for elements on any domain is:

[0042] μ A∩B (u) = min{μ A (u),μ B (u)}

[0043] Where u represents the parameter in the domain, A and B represent different element types of fuzzy reasoning input, and similarly, the union operation and complement operation are:

[0044] μ A∪B (u) = max{μ A (u),μ B (u)}

[0045] μ A (u) = 1-μ A (u)

[0046] Using the commonly used fuzzy relation synthesis operation, let X, Y, Z be the domain, x, y, z be the parameters in the corresponding domain, R be the fuzzy relation on X×Y, S be the fuzzy relation on Y×Z, T be the synthesis from R to S, denoted as Its membership function is defined as follows:

[0047]

[0048] Here ∪ represents the maximum value for all y, and × is the binomial product operator;

[0049] Then use the centroid method to defuzzify and finally get the output between [0,1];<F1,F2,F4,F5,F6,F7> As input, the output is a dangerous driving index between [0,1]. The closer the index is to 1, the higher the degree of danger.

[0050] Furthermore, the step 5 specifically includes the following steps:

[0051] Step 5-1: After the system makes a dangerous driving judgment, if the driver believes that his driving behavior is relatively safe, he chooses to judge it as a false positive of the system, and a false alarm occurs. The driver manually increases the dangerous driving threshold;

[0052] Step 5-2: After the system makes a fatigue driving judgment, if the driver believes that he is not fatigued, it is judged as a false positive by the system, and a false alarm occurs. The driver manually increases the fatigue driving threshold;

[0053] Step 5-3: If the driver thinks the threshold is too high, the driver can manually choose to reset the driving threshold of the drive control layer to the initial value of 0.5 to supervise himself to drive safely;

[0054] Step 5-4: If any of the reminders in step 3-1, step 3-2, and step 3-5 appear, that is, emergency dangerous driving behavior or obvious fatigue driving behavior, the system will consider that a false negative judgment has occurred and will also feedback to the drive control layer to automatically reset the driving threshold.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] 1) There is no need to collect a large number of normal driving characteristics and known dangerous driving characteristics, nor is there a need to train detection models, which ensures the running speed and does not require expensive equipment such as video input devices, thus reducing the implementation cost and making it more versatile;

[0057] 2) The method combines road conditions, driver status, and vehicle driving characteristics to analyze the lateral movement and longitudinal following distance of the vehicle, and has an adaptive adjustment mechanism to suit the driving styles of different drivers, making the judgment of driving behavior more comprehensive and more valuable, making the detection more accurate and effective;

[0058] 3) It can distinguish short-term dangerous driving behaviors from long-term fatigue driving behaviors. Based on the characteristics of actual dangerous situations and taking into account the identification of various emergency situations, it can also take into account the timely warning and detection of dangerous behaviors and the detection of hard-to-detect fatigue driving behaviors. It can effectively assist drivers in driving safely and has important practical value.

[0059] 4) It causes little interference to the driver, improves the accuracy of identifying the driver's dangerous behavior and the detection rate of fatigue driving, effectively avoids traffic accidents caused by dangerous and fatigue driving, helps the driver to discover slowly changing driving behaviors, and makes the driver aware of the danger in time when the driver's driving level and decision-making level are reduced. Therefore, it can effectively assist the driver to drive safely and has important practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 4 is an overall architecture diagram of the driver assistance system described in an embodiment of the present invention.

[0061] Figure 2 Schematic diagram of the vehicle acceleration sector division described in an embodiment of the present invention.

[0062] Figure 3 It is a schematic diagram of a fuzzy reasoning system model based on the Mamdani fuzzy system modeling method described in an embodiment of the present invention.

[0063] Figure 4 The present invention is a detailed flowchart of a method for identifying and warning a driver's dangerous driving and fatigue driving behaviors based on fuzzy reasoning in an embodiment of the present invention. DETAILED DESCRIPTION

[0064] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings.

[0065] The present invention provides a driver assistance system and method for identifying and warning of dangerous and fatigue driving behaviors. The method comprises the following steps: (1) sensing vehicle behavior, driver status, and road condition information through sensors; (2) processing the initial information obtained by the sensors to form a low-dimensional feature vector corresponding to the vehicle; (3) making a short-term real-time judgment on the vehicle feature information. If there is an emergency dangerous driving behavior that meets the characteristics in the short term, the driver, surrounding vehicles, and pedestrians on the road are immediately warned by high-frequency sound, high-speed vibration, and picture images; (4) using a fuzzy inference system to make a short-term real-time judgment on the vehicle feature information. If there is an ordinary dangerous driving behavior that meets the characteristics, the driver is immediately warned by high-frequency sound, high-speed vibration, and picture images; (5) if there is no dangerous driving behavior in the short term, the long-term behavior is judged on the short-term driving data, mainly based on the long-term following distance and the driver's own status. If it meets the fatigue driving characteristics, the driver is reminded to rest through vibration and picture images; (6) if a false alarm confirmed by the driver is generated, a small range adaptive adjustment that does not affect safety is made. If an emergency dangerous situation occurs, the threshold is reset. The driver can also reset the threshold to adapt to the driving style of different drivers.

[0066] A driver assistance system that recognizes and warns of dangerous and fatigued driving behavior, such as Figure 1 The hierarchy shown includes:

[0067] Sensor layer: collects information about vehicle behavior, driver status, and road conditions through sensor perception;

[0068] Perception layer: processes the initial information obtained by the sensor to form a low-dimensional feature vector corresponding to the vehicle;

[0069] Cognitive layer: Use fuzzy reasoning system to make short-term real-time judgments on vehicle characteristic information. If there is dangerous driving behavior that meets the characteristics in the short term, it is judged as dangerous driving. If there is no dangerous driving behavior in the short term, the short-term driving data is used to make long-term behavior judgments, mainly focusing on the long-term changes in the following distance and the driver's own condition. If it meets the characteristics of fatigue driving, it is judged as fatigue driving;

[0070] Warning execution layer: Based on the judgment of the cognitive layer, if it is judged as emergency and dangerous driving, a high-frequency sound prompt will be issued to surrounding vehicles and pedestrians, and a high-frequency sound, high-speed vibration, and screen prompt warning will be issued to the target vehicle. If it is emergency and dangerous driving, the drive control layer will be fed back to adjust the threshold. If it is ordinary dangerous driving, the target vehicle driver will be given the above corresponding warning. If it is judged as fatigue driving, a screen and voice prompt will be issued to remind the driver to take a rest and concentrate on driving;

[0071] Driving control layer: According to the driver's judgment of his own driving characteristics and the occurrence of emergency situations, the system's false positive and false negative judgments are selected to achieve adaptive adjustment of the thresholds for dangerous and fatigue driving judgments;

[0072] The present invention adopts the following method steps to achieve the above-mentioned invention object:

[0073] (1) Collect information about the target vehicle.

[0074] Using steering wheel angle sensor, Beidou satellite navigation system, vehicle inertial measurement unit (accelerometer, gyroscope, magnetometer), wearable accelerometer and heart rate monitor (such as smart bracelet), electronic scanning radar or millimeter wave radar, on-board computer or industrial computer (road condition acquisition), the specific process is as follows.

[0075] (1-1) Use a steering wheel angle sensor to obtain steering wheel angle change information;

[0076] (1-2) Use the Beidou satellite navigation system to obtain the horizontal and vertical coordinates of the vehicle relative to the road;

[0077] (1-3) Use the vehicle inertial measurement unit to obtain accurate vehicle acceleration change information;

[0078] (1-4) Obtaining the driver's motion swing and heart rate status through a wearable accelerometer and heart rate meter;

[0079] (1-5) Obtaining the distance information of the vehicle ahead through millimeter wave radar;

[0080] (1-6) Obtain road condition information through the on-board computer.

[0081] (2) Process the initial information to form a low-dimensional feature vector corresponding to the vehicle.

[0082] The step (2) specifically comprises the following steps:

[0083] (2-1) Use the steering wheel angle sequence to obtain the vehicle's lateral swing characteristics.

[0084] The steering wheel angle sequence {θ n}, where n represents the time-related stream data sequence number, and performs fast Fourier transform to obtain its power spectrum density:

[0085] F(x) = FFT{θ n}

[0086] Then sum the amplitudes to get the corresponding swing value S:

[0087]

[0088] S is the lateral sway characteristic of the vehicle, which can reflect the amplitude and frequency of the vehicle's lateral movement. The larger its value is, the more frequent the lateral swing of the vehicle is and the larger the amplitude is, that is, the more unstable the driving state is.

[0089] (2-2) Use the horizontal and vertical coordinates of the vehicle relative to the road to solve the longitudinal speed information parallel to the road and the lateral speed information perpendicular to the road, and take the average over the same time length as (2-1).

[0090] (2-3) Divide the acceleration into a certain range, divide the acceleration range into four right-angle sectors, front, back, left, and right, and determine three acceleration ranges to form 16 acceleration sectors. Use 4 bits to number them and determine the maximum safe acceleration value a in the four directions of front, back, left, and right. fm 、a bm 、a lm 、a rm , and divide the value into four acceleration sectors according to the step length of one third of this value, forming a series of numerical sequences during the vehicle driving process, which are divided as follows: Figure 2 As shown, the middle rectangle represents the vehicle, the arrow indicates the direction of the vehicle, the front and rear acceleration thresholds are different, and the left and right acceleration thresholds are the same, that is, a fm >a bm >a lm =a rm , so the size range of the divided sectors is roughly as shown in the schematic diagram.

[0091] (2-4) The driver's swing amplitude is processed in the same way as (2-1) as the driver's swing feature. The percentage difference between the current heart rate and the normal heart rate is measured based on the driver's calm heart rate to obtain the driver's heart rate deviation.

[0092] (2-5) Average the obtained following distances and save the average following distance every minute. At least ten sets of data need to be saved.

[0093] (2-6) Obtain basic road condition information, determine whether the road is congested or not, determine the wetness of the road surface according to the weather conditions, and infer the friction conditions. The lower the road friction, the higher the danger level, and the higher this value. According to existing studies, the average number of accidents is snowy days > rainy days > sunny days, foggy days > cloudy days > windy days, and the adhesion coefficient of the road surface is icy road surface < snowy road surface < dry road surface after snowmelt < damp road surface < wet road surface < high temperature road surface < dry road surface. Therefore, we determine the weather coefficient based on the weather conditions and the wetness of the road surface at that time. The coefficient of the windy dry road surface is 0.9. As the number of accidents corresponding to the weather increases and the adhesion coefficient of the road surface decreases, this coefficient gradually increases until the highest coefficient of the snowy and icy road surface is 1.2. At the same time, according to the vehicle density around the target vehicle, determine the density coefficient from low to high. The coefficient increases with the increase of density, and the range is also [0.9,1.2]. Finally, multiply the two coefficients to get the final road condition coefficient. According to the above process, the minimum coefficient is not less than 0.81 and the maximum does not exceed 1.44.

[0094] In step (2), this embodiment uses the eight-tuple of <vehicle lateral swing characteristics, vehicle longitudinal speed information, vehicle lateral speed information, vehicle acceleration block, driver state swing characteristics, driver heart rate deviation, vehicle short-term forward following distance, road condition information> as the flow mark, extracts the flow data of the vehicle to be detected, and makes the next step of judgment based on the flow data.

[0095] Generate vehicle flow features. The vehicle flow features used include:

[0096] F1: Vehicle lateral sway characteristics;

[0097] F2: vehicle absolute speed information;

[0098] F3: Vehicle lateral speed information;

[0099] F4: Vehicle acceleration block;

[0100] F5: Driver swing characteristics;

[0101] F6: Driver heart rate deviation;

[0102] F7: short-term forward following distance of the vehicle;

[0103] F8: Road condition coefficient.

[0104] final,<F1,F2,F3,F4,F5,F6,F7,F8> , that is, the sequential arrangement of the corresponding values ​​of each feature constitutes the driving state feature to be detected.

[0105] (3) Use fuzzy reasoning system to judge short-term dangerous behavior and long-term fatigue driving behavior.

[0106] The step (3) specifically comprises the following steps:

[0107] (3-1) Real-time tracking and judging of following vehicle distance. Different speed ranges have different warning thresholds. The standard is divided into three levels. When the relative speed is 15-20m / s, the collision threshold is 3.5s, when it is 10-15m / s, it is 3s, and when it is less than 10m / s, it corresponds to 2.5s. If it is within the threshold warning range, it is directly judged as emergency dangerous driving.

[0108] (3-2) Analyze the vehicle acceleration block. If the partition is backward and the acceleration value is greater than the maximum backward safety acceleration value a bm That is, within the rear four zones, the driver is judged to have taken emergency braking action; if it is within the left and right four zones, it is a sharp turn; and if it is within the front four zones, it is a sudden acceleration, all of which are judged to be emergency and dangerous driving.

[0109] (3-3) Except for the above exceptions,<F1,F2,F4,F5,F6,F7> All of them are included in the fuzzy reasoning framework to judge dangerous driving behaviors with unclear characteristics. For these six characteristics, the average within a few seconds is taken (the acceleration is replaced by the sector median), and then weighted according to the F8 road congestion, weather conditions, and road slipperiness. Then, they are normalized according to the maximum value of the corresponding characteristic attributes of the surrounding vehicles. In this way, comprehensive road condition information can be taken into account.

[0110] The present invention uses Figure 3 The Mamdani fuzzy system modeling method shown in the figure is used for fuzzy reasoning. Appropriate fuzzy language variables and corresponding membership functions are assigned to the corresponding attributes of F1, F2, F4, F5, F6, and F7. Each attribute corresponds to two high and low membership functions. Fuzzy logic rules are then written for the relationship between the six elements. The vehicle has a lot of lateral sway, a fast absolute speed, a large acceleration value, a large driver swing, a lot of heart rate deviation, and is close to the car. These are all positive contributions to the increase in the dangerous driving index. The high probability of dangerous driving obtained by fuzzy reasoning can be judged as a high probability that the driver has dangerous driving behavior. On the contrary, the reduction of the six attributes has a negative contribution to the increase in the dangerous driving index. It can be judged that the driver's driving is relatively stable and safe. Therefore, it can be determined that 2 6 , that is, 64 fuzzy reasoning rules, corresponding to the dangerous driving index (divided into three parts of membership functions: high, medium and low). Among them, the membership function selects appropriate Gaussian, triangular, Z-type and S-type membership functions for fuzzy reasoning according to the characteristics of their respective attributes.

[0111] Fuzzy reasoning and judgment are performed according to fuzzy rules. In fuzzy logic, the intersection operation of μ(u) for elements on any domain is:

[0112] μ A∩B (u) = min{μ A(u),μ B (u)}

[0113] Where u represents the parameter in the domain, A and B represent different element types of fuzzy reasoning input, and similarly, the union operation and complement operation are:

[0114] μ A∪B (u) = max{μ A (u),μ B (u)}

[0115] μ A (u) = 1-μ A (u)

[0116] Using the commonly used fuzzy relation synthesis operation, let X, Y, Z be the domain, x, y, z be the parameters in the corresponding domain, R be the fuzzy relation on X×Y, S be the fuzzy relation on Y×Z, T be the synthesis from R to S, denoted as Its membership function is defined as follows:

[0117]

[0118] Here, ∪ represents the maximum value for all y, and × is the binomial product operator, which can be defined as the minimum value or algebraic product. The most commonly used synthesis is the maximum-minimum synthesis, and its calculation formula is:

[0119]

[0120] Then use the centroid method to defuzzify and finally get the output between [0,1].

[0121] In the application environment of the present invention<F1,F2,F4,F5,F6,F7> As an element of the input domain, the output is a dangerous driving index between [0,1]. The closer the index is to 1, the higher the degree of danger. The initial threshold value is determined to be 0.5.

[0122] (3-4) Whether there is dangerous driving behavior is determined based on the dangerous driving index. If it exceeds a certain threshold value (the threshold value is between [0.5, 0.8], and the initial threshold is 0.5), it is determined as ordinary dangerous driving.

[0123] (3-5) The average following distance and vehicle lateral sway characteristics are collected in units of 1 minute. If the average following distance continues to decrease and the vehicle lateral sway continues to increase within ten sets of data, it is determined that the driver is driving fatigued.

[0124] (3-6) Collect the dangerous driving index within 10 minutes. If the average value is greater than a certain threshold (all values ​​are less than the dangerous driving threshold), it is judged as fatigue driving.

[0125] (4) Issue corresponding danger warnings to the target vehicle, surrounding vehicles, and pedestrians.

[0126] (4-1) The vehicle is judged as being in an emergency and dangerous driving situation, and a high-frequency sound warning is issued to surrounding vehicles and pedestrians, and a high-frequency sound, high-speed vibration, and visual warning is issued to the target vehicle.

[0127] (4-2) It is judged as ordinary dangerous driving, and the target vehicle is warned of dangerous behavior through sound and picture.

[0128] (4-3) The driver is judged to be fatigued and a screen and voice prompt are issued to remind the driver to take a rest and concentrate on driving.

[0129] (5) Adaptive adjustment of the threshold for judging dangerous and fatigue driving is achieved based on the driver's judgment of his own driving characteristics.

[0130] The step (5) specifically comprises the following steps:

[0131] (5-1) After the system makes a dangerous driving judgment, if the driver believes that his driving behavior is relatively safe, he can choose to judge it as a false positive of the system. If a false alarm occurs, the driver can manually increase the dangerous driving threshold (adjustment step size is 0.02). The threshold has an unchangeable maximum safety value.

[0132] (5-2) After the system makes a fatigue driving judgment, if the driver believes that he is not fatigued, he can choose to judge it as a false positive of the system. If a false alarm occurs, the driver can manually increase the fatigue driving threshold (adjustment step size is 0.05). The threshold has an unchangeable maximum safety value.

[0133] (5-3) If the driver thinks the threshold is too high, the driver can manually choose to reset the driving threshold of the drive control layer to the initial value of 0.5 to supervise himself to drive safely;

[0134] (5-4) If any of the reminders (3-1), (3-2), or (3-5) appears, i.e., emergency dangerous driving behavior or obvious fatigue driving behavior, the system will consider that a false negative judgment has occurred and will also feedback to the drive control layer to automatically reset the driving threshold.

[0135] The overall process of the technical solution is as follows Figure 4 shown.

[0136] The following provides six specific implementation examples based on specific usage scenarios.

[0137] Embodiment 1:

[0138] The vehicle's steering wheel is installed with an angle sensor, millimeter-wave radar, an on-board computer that can be connected to the Internet of Vehicles, an on-board Beidou satellite navigation system, and a vehicle-mounted inertial measurement unit (accelerometer, gyroscope, magnetometer). The driver wears a smart bracelet (including a heart rate monitor and accelerometer).

[0139] Step 1: The above sensors collect the vehicle's steering wheel angle, the vehicle's horizontal and vertical coordinates relative to the road, the vehicle's acceleration change information, the distance information to the vehicle in front, the road conditions and the general speed and density of surrounding vehicles, the driver's heart rate and movement swing information.

[0140] Step 2: Process the above data to obtain a low-dimensional driving feature vector, and use the corresponding data as the input of the fuzzy inference system, incorporate it into the fuzzy inference system, make judgments based on fuzzy rules and fuzzy membership functions, and continuously output and save the driver's dangerous driving index value. And use the road condition coefficient to perform weighted processing on it.

[0141] Step 3: The driver encounters an emergency situation and brakes suddenly or steers quickly. The driver assistance system detects that the instantaneous acceleration value exceeds the corresponding safety threshold and immediately determines that the driver is in an emergency situation.

[0142] Step 4: Use the vehicle speaker to send a high-frequency sound warning to the driver and surrounding vehicles and pedestrians, the smart bracelet and mobile phone worn by the driver will vibrate at a high speed, and a screen prompt will appear on the vehicle computer. The warning lasts for 10 seconds.

[0143] Step 5: After the warning ends, the driving control layer will provide feedback to the cognitive layer to restore the adaptively adjusted dangerous and fatigue driving thresholds to the initial minimum value.

[0144] Embodiment 2:

[0145] The vehicle's steering wheel is installed with an angle sensor, millimeter-wave radar, an on-board computer that can be connected to the Internet of Vehicles, an on-board Beidou satellite navigation system, and a vehicle-mounted inertial measurement unit (accelerometer, gyroscope, magnetometer). The driver wears a smart bracelet (including a heart rate monitor and accelerometer).

[0146] Step 1: The above sensors collect the vehicle's steering wheel angle, the vehicle's horizontal and vertical coordinates relative to the road, the vehicle's acceleration change information, the distance information to the vehicle in front, the road conditions and the general speed and density of surrounding vehicles, the driver's heart rate and movement swing information.

[0147] Step 2: Process the above data to obtain a low-dimensional driving feature vector, and use the corresponding data as the input of the fuzzy inference system, incorporate it into the fuzzy inference system, make judgments based on fuzzy rules and fuzzy membership functions, and continuously output and save the driver's dangerous driving index value. And use the road condition coefficient to perform weighted processing on it.

[0148] Step 3: The driver is not paying attention when following the vehicle. In an emergency, he / she brakes suddenly or the vehicle ahead is too close based on the driver's current speed. The driver is immediately judged to be engaging in emergency and dangerous driving behavior.

[0149] Step 4: Use the vehicle speaker to send a high-frequency sound warning to the driver and surrounding vehicles and pedestrians, the smart bracelet and mobile phone worn by the driver will vibrate at a high speed, and a screen prompt will appear on the vehicle computer. The warning lasts for 10 seconds.

[0150] Step 5: After the warning ends, the driving control layer will provide feedback to the cognitive layer to restore the adaptively adjusted dangerous and fatigue driving thresholds to the initial minimum value.

[0151] Embodiment three:

[0152] The vehicle's steering wheel is installed with an angle sensor, millimeter-wave radar, an on-board computer that can be connected to the Internet of Vehicles, an on-board Beidou satellite navigation system, and a vehicle-mounted inertial measurement unit (accelerometer, gyroscope, magnetometer). The driver wears a smart bracelet (including a heart rate monitor and accelerometer).

[0153] Step 1: The above sensors collect the vehicle's steering wheel angle, the vehicle's horizontal and vertical coordinates relative to the road, the vehicle's acceleration change information, the distance information to the vehicle in front, the road conditions and the general speed and density of surrounding vehicles, the driver's heart rate and movement swing information.

[0154] Step 2: Process the above data to obtain a low-dimensional driving feature vector, and use the corresponding data as the input of the fuzzy inference system, incorporate it into the fuzzy inference system, make judgments based on fuzzy rules and fuzzy membership functions, and continuously output and save the driver's dangerous driving index value. And use the road condition coefficient to perform weighted processing on it.

[0155] Step 3: If the driver has dangerous driving habits such as swaying driving direction and unstable forward speed, the dangerous driving index value output by the fuzzy reasoning system of the driver assistance system will increase and exceed the judgment threshold value of dangerous driving. The driver is judged to have dangerous driving behavior habits.

[0156] Step 4: The vehicle speaker makes a sound and the vehicle computer displays a picture to remind the driver of dangerous driving behavior. The warning lasts for 10 seconds.

[0157] Step 5: After the warning ends, the on-board computer pops up the threshold adjustment screen. If the driver believes that his behavior just now is due to his driving habits and is relatively safe, he clicks to confirm the system threshold adjustment. The driver control layer feeds back the adjustment information to the cognitive layer, and the dangerous driving threshold is increased once according to a fixed step length. The threshold has an unchangeable maximum safety value. If he judges that he has indeed engaged in dangerous behavior, he will correct his behavior in time and not adjust the threshold to monitor his own driving behavior.

[0158] Embodiment 4:

[0159] The vehicle's steering wheel is installed with an angle sensor, millimeter-wave radar, an on-board computer that can be connected to the Internet of Vehicles, an on-board Beidou satellite navigation system, and a vehicle-mounted inertial measurement unit (accelerometer, gyroscope, magnetometer). The driver wears a smart bracelet (including a heart rate monitor and accelerometer).

[0160] Step 1: The above sensors collect the vehicle's steering wheel angle, the vehicle's horizontal and vertical coordinates relative to the road, the vehicle's acceleration change information, the distance information to the vehicle in front, the road conditions and the general speed and density of surrounding vehicles, the driver's heart rate and movement swing information.

[0161] Step 2: Process the above data to obtain a low-dimensional driving feature vector, and use the corresponding data as the input of the fuzzy inference system, incorporate it into the fuzzy inference system, make judgments based on fuzzy rules and fuzzy membership functions, and continuously output and save the driver's dangerous driving index value. And use the road condition coefficient to perform weighted processing on it.

[0162] Step 3: Determine the driver's dangerous driving index value over a long period of time (10 minutes). Over a long period of time, the driver's average following distance continues to decrease, the vehicle's lateral sway amplitude and frequency increase, and the heart rate is abnormal. The system determines that the driver has serious fatigue driving behavior.

[0163] Step 4: The vehicle speaker makes a sound and the vehicle computer displays a picture to remind the driver that he is driving fatigued and to immediately find a parking area nearby to rest. The warning lasts for 10 seconds.

[0164] Step 5: After the warning ends, the driving control layer will provide feedback to the cognitive layer to restore the adaptively adjusted fatigue driving threshold to the initial minimum value.

[0165] Embodiment five:

[0166] The vehicle's steering wheel is installed with an angle sensor, millimeter-wave radar, an on-board computer that can be connected to the Internet of Vehicles, an on-board Beidou satellite navigation system, and a vehicle-mounted inertial measurement unit (accelerometer, gyroscope, magnetometer). The driver wears a smart bracelet (including a heart rate monitor and accelerometer).

[0167] Step 1: The above sensors collect the vehicle's steering wheel angle, the vehicle's horizontal and vertical coordinates relative to the road, the vehicle's acceleration change information, the distance information to the vehicle in front, the road conditions and the general speed and density of surrounding vehicles, the driver's heart rate and movement swing information.

[0168] Step 2: Process the above data to obtain a low-dimensional driving feature vector, and use the corresponding data as the input of the fuzzy inference system, incorporate it into the fuzzy inference system, make judgments based on fuzzy rules and fuzzy membership functions, and continuously output and save the driver's dangerous driving index value. And use the road condition coefficient to perform weighted processing on it.

[0169] Step 3: If the driver is drowsy, distracted, emotional, or has similar fatigue driving behaviors, the short-term dangerous driving index will not be abnormal, but the average value of the long-term fuzzy reasoning output will increase to above the fatigue driving threshold, and the system will determine that the driver is experiencing fatigue driving behavior.

[0170] Step 4: The vehicle speaker makes a sound and the vehicle computer displays a screen to remind the driver of fatigue driving behavior, remind the driver to concentrate during driving and suggest the driver to take a rest. The warning lasts for 10 seconds.

[0171] Step 5: After the warning ends, the onboard computer pops up the threshold adjustment screen. If the driver believes that he is not tired or has been concentrating on driving, he can click to confirm the system threshold adjustment. The driver control layer will feed back the adjustment information to the cognitive layer, and the fatigue driving threshold will increase once according to a fixed step length. The threshold has an unchangeable maximum safety value. If the driver determines that he is indeed driving fatigued, he will correct his behavior in time and not adjust the threshold to monitor his own driving behavior.

[0172] Embodiment six:

[0173] The vehicle's steering wheel is installed with an angle sensor, millimeter-wave radar, an on-board computer that can be connected to the Internet of Vehicles, an on-board Beidou satellite navigation system, and a vehicle-mounted inertial measurement unit (accelerometer, gyroscope, magnetometer). The driver wears a smart bracelet (including a heart rate monitor and accelerometer).

[0174] Step 1: The above sensors collect the vehicle's steering wheel angle, the vehicle's horizontal and vertical coordinates relative to the road, the vehicle's acceleration change information, the distance information to the vehicle in front, the road conditions and the general speed and density of surrounding vehicles, the driver's heart rate and movement swing information.

[0175] Step 2: Process the above data to obtain a low-dimensional driving feature vector, and use the corresponding data as the input of the fuzzy inference system, incorporate it into the fuzzy inference system, make judgments based on fuzzy rules and fuzzy membership functions, and continuously output and save the driver's dangerous driving index value. And use the road condition coefficient to perform weighted processing on it.

[0176] Step 3: The driver wants to supervise himself to drive more smoothly and properly or the vehicle has a driver change.

[0177] Step 4: At any time, if the driver wants to reset the dangerous and fatigue driving judgment threshold, he can operate the system interface and reset it manually. The drive control layer will feed back the adjustment information to the cognitive layer to adapt to different drivers or supervise the driver to drive safely and in accordance with regulations.

[0178] The above description is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiment. Any equivalent modifications or changes made by ordinary technicians in this field based on the contents disclosed by the present invention should be included in the protection scope recorded in the claims.

Claims

1. A driver assistance system for identifying and warning of dangerous and fatigued driving behavior, characterized in that: The system comprises a sensor layer, a perception layer, a cognitive layer, a warning execution layer and a drive control layer connected in sequence, wherein the drive control layer is connected to the cognitive layer; The sensor layer collects information about vehicle behavior, driver status, and road conditions through sensor perception; The perception layer processes the initial information obtained by the sensor to form a low-dimensional feature vector corresponding to the vehicle; The cognitive layer is equipped with a fuzzy reasoning system to make a short-term real-time judgment on whether the vehicle characteristic information is dangerous driving. If there is no dangerous driving behavior in the short term, the short-term driving data will be used to make a long-term behavior judgment, including the long-term following distance change and the driver's own condition. If it meets the fatigue driving characteristics, it will be judged as fatigue driving. The warning execution layer, based on the judgment of the cognitive layer, issues high-frequency sound prompts to surrounding vehicles and pedestrians if it is emergency and dangerous driving, and issues high-frequency sound, high-speed vibration, and visual prompts to the target vehicle. If it is emergency and dangerous driving, the driver control layer is fed back to adjust the threshold. If it is ordinary dangerous driving, the driver of the target vehicle is warned with high-frequency sound, high-speed vibration, and visual prompts. If it is fatigue driving, visual and voice prompts are issued to remind the driver to take a rest and concentrate on driving. The driving control layer makes a choice between the system's false positive and false negative judgments based on the driver's judgment of his own driving characteristics and the occurrence of emergency situations, and realizes adaptive adjustment of the thresholds for judgment of dangerous and fatigue driving; The driver assistance system for identifying and warning of dangerous and fatigued driving behavior is based on the following method, including the following steps: Step 1: Use sensors to sense vehicle behavior, driver status, and road conditions; Step 2: Process the initial information obtained by the sensor to form a low-dimensional feature vector corresponding to the vehicle; The step 2 specifically includes the following steps: Step 2-1, using the angle sequence of the steering wheel angle change information to obtain the vehicle lateral sway characteristics; Steering wheel angle sequence within a few seconds ,in n Represents the time-related stream data sequence number and performs fast Fourier transform FFT , and obtain its power spectral density F(x) : Then sum the amplitudes to get the corresponding swing value S : S As a vehicle lateral sway characteristic that reflects the amplitude and frequency of the vehicle's lateral movement, the larger its value is, the more frequent and larger the amplitude of the vehicle's lateral sway is, that is, the more unstable the driving state is; Step 2-2, using the horizontal and vertical coordinates of the vehicle relative to the road to solve the longitudinal speed information parallel to the road and the lateral speed information perpendicular to the road, and averaging them within the same time length as step 2-1; Step 2-3: Divide the acceleration of the vehicle into a certain range, divide the acceleration range into four right-angle sectors, front, back, left, and right, and determine three acceleration ranges to form 16 acceleration sectors, which are numbered with 4 bits to determine the maximum safe acceleration values ​​in the four directions of front, back, left, and right. a fm , a bm , a lm , a rm , and divide the value into four acceleration sectors according to the step length of one third of this value, forming a series of numerical sequences during the vehicle driving process. The acceleration thresholds of the front and rear are different, and the acceleration thresholds of the left and right are the same, that is, a fm > a bm > a lm = a rm ; Step 2-4, determining the percentage difference between the current heart rate and the normal heart rate according to the driver's calm heart rate, and obtaining the driver's heart rate deviation as the driver's swing feature; Step 2-5, averaging the obtained following distances, and saving the average following distance every minute; Step 2-6, obtain traffic information, determine whether the road is congested or not, determine the wetness of the road surface according to weather conditions, and infer the friction conditions. The more congested the traffic is, the lower the road friction is, and the higher the danger level is, the higher this value is; Step 3: Make a short-term real-time judgment on the vehicle characteristic information. If there is an emergency dangerous driving behavior that meets the characteristics in the short term, the driver, surrounding vehicles, and pedestrians on the road will be warned immediately by high-frequency sound, high-speed vibration, and image. The step 3 specifically includes the following steps: Step 3-1: Real-time tracking and judging the following vehicle distance. Different speed ranges have different warning thresholds. The standard is divided into three levels. When the relative speed is 15-20 m / s, the collision threshold is 3.5s, when it is 10-15 m / s, it is 3s, and when it is less than 10 m / s, it corresponds to 2.5s. If it is within the threshold warning range, it is directly judged as emergency dangerous driving; Step 3-2: Analyze the vehicle acceleration block. If the partition is rearward and the acceleration value is greater than the maximum rearward safety acceleration value, a bm , that is, if it is in the rear four zones, the driver is judged to have taken emergency braking behavior, if it is in the left and right four zones, it is a sharp turn, and if it is in the front four zones, it is a sudden acceleration, all of which are judged as emergency and dangerous driving; Step 3-3, except for the abnormal situations determined as emergency dangerous driving, <F1, F2, F4, F5, F6, F7> are all included in the fuzzy reasoning framework to judge dangerous driving behaviors with unclear features. For these six features, the average within a few seconds is taken, and then weighted processing is performed according to F8 road congestion, weather conditions, and road slipperiness, and then normalized according to the maximum value of the corresponding feature attributes of the surrounding vehicles; Step 3-4, judging whether there is dangerous driving behavior according to the dangerous driving index, if it exceeds the threshold value, it is judged as ordinary dangerous driving; Step 3-5, collect the average following distance and vehicle lateral sway characteristics in units of 1 minute. If the average following distance continues to decrease and the vehicle lateral sway continues to increase within ten sets of data, it is determined that the driver has fatigue driving behavior; Step 3-6, collect the dangerous driving index within 10 minutes, if the average value is greater than a certain threshold and all values ​​are less than the dangerous driving threshold, it is determined to be fatigue driving; Step 4: Use the fuzzy inference system to make a short-term real-time judgment on the vehicle feature information. If there is a common dangerous driving behavior that meets the characteristics, the driver will be warned immediately by high-frequency sound, high-speed vibration and screen images; Step 5: If there is no dangerous driving behavior in the short term, the short-term driving data is used to determine the long-term behavior, and the long-term following distance and the driver's own condition are judged. If it meets the characteristics of fatigue driving, the driver is reminded to take a rest through vibration and screen; Step 6: If a driver-confirmed false alarm is generated, an adaptive adjustment is made within a small range that does not affect safety. If an emergency dangerous situation occurs, the threshold is reset. The driver resets the threshold to adapt to the driving style of different drivers.

2. A driver assistance system for identifying and warning of dangerous and fatigue driving behaviors according to claim 1, characterized in that: The sensor layer includes a steering wheel angle sensor, a Beidou satellite navigation system, a vehicle inertial measurement unit, a wearable accelerometer and heart rate meter, a radar, and a computing device.

3. A driver assistance system for identifying and warning of dangerous and fatigue driving behaviors according to claim 1, characterized in that: In step 1, the steering wheel angle change information, the horizontal and vertical coordinates of the vehicle relative to the road, the vehicle acceleration change information, the driver's motion swing and heart rate status, the front vehicle distance information and the road condition information are obtained.

4. A driver assistance system for identifying and warning of dangerous and fatigue driving behaviors according to claim 1, characterized in that: In step 2, the driving state feature to be detected is composed of < F1, F2, F3, F4, F5, F6, F7, F8 >, that is, the order of the corresponding values ​​of each feature is arranged to make the next step of judgment; among them, F1 is the vehicle lateral sway feature; F2 is the vehicle absolute speed information; F3 is the vehicle lateral speed information; F4 is the vehicle acceleration block; F5 is the driver's swing feature; F6 is the driver's heart rate deviation; F7 is the vehicle's short-term forward following distance; F8 is the road condition coefficient.

5. A driver assistance system for identifying and warning of dangerous and fatigue driving behaviors according to claim 1, characterized in that: The step 5 specifically includes the following steps: Step 5-1: After the system makes a dangerous driving judgment, if the driver believes that his driving behavior is relatively safe, he chooses to judge it as a false positive of the system, and a false alarm occurs. The driver manually increases the dangerous driving threshold; Step 5-2: After the system makes a fatigue driving judgment, if the driver believes that he is not fatigued, it is judged as a false positive by the system, and a false alarm occurs. The driver manually increases the fatigue driving threshold; Step 5-3, if the driver thinks the threshold is too high, the driver manually chooses to reset the driving threshold of the driving control layer to the initial value 0.5 to supervise himself to drive safely; Step 5-4: If any of the reminders in step 3-1, step 3-2, and step 3-5 appears, that is, emergency dangerous driving behavior or obvious fatigue driving behavior, the system will consider that a false negative judgment has occurred and will also feedback to the drive control layer to automatically reset the driving threshold.