Abnormal state determination method for drunk driving vehicle with unmanned aerial vehicle

Abstract

The application discloses a kind of fusion unmanned vehicle driving abnormal state discrimination method of drunk driving, utilizes data processing and unmanned technology to monitor vehicle driving track anomaly, and identifies drunk driving behavior by scoring.First, data processing module establishes vehicle swing discrimination model, vehicle sudden speed change discrimination model and signal light intersection waiting time discrimination model by multiple feature evaluation indexes, and the track abnormal degree is scored;When total score exceeds threshold value, judge whether there is novice driving sign;Then, data generation module dispatches unmanned aerial vehicle to identify the face state of driver, judges whether driver has the characteristic factor of drunk driving;Finally, terminal information judging module confirms whether overall comprehensive index exceeds specified threshold value, to determine whether there is drunk driving situation.The application solves the problem of existing drunk driving discrimination method, such as missed detection, inspection lag and interference with normal drivers, improves the accuracy, real-time efficiency and coverage of drunk driving discrimination.

Description

Technical Field

[0001] This invention relates to the fields of traffic management and safety technology, and in particular to a method for identifying abnormal driving conditions of drunk vehicles by integrating unmanned aerial vehicles (UAVs). Background Technology

[0002] With the continuous and rapid development of the automobile manufacturing industry, the number of cars on the road has been increasing year by year. People rely on cars for travel, but at the same time, they also face increasingly serious traffic safety problems. Drivers, as a vital part of the road traffic system, account for a large proportion of traffic accidents caused by their dangerous driving behaviors, especially drunk driving, which is a particularly dangerous driving behavior that greatly threatens traffic safety. Drunk driving assessments typically use breathalyzers to test the alcohol content in the breath of the test subject, thereby estimating the blood alcohol concentration. However, breathalyzers can only perform a single test and cannot monitor a driver's drinking status in real time. Furthermore, current road traffic management methods have blind spots, failing to promptly detect vehicles suspected of drunk driving. With the rapid development of drone technology, how to integrate drones into traffic management and safety to effectively screen for suspected drunk driving, addressing the difficulties and blind spots in drunk driving detection, and improving the accuracy and efficiency of detection has become an urgent technical problem to be solved.

[0003] Existing methods for detecting drunk driving have several significant problems and shortcomings. First, while traditional breathalyzer and blood tests can determine whether a driver has been drinking to some extent, they are prone to false negatives. This is because these methods can only be performed at specific times and locations, and cannot provide continuous monitoring of the driver, thus missing driving behavior before the peak blood alcohol concentration. Furthermore, the results of breathalyzer and blood tests are affected by many factors, such as the accuracy of the testing equipment, the operator's expertise, and individual differences among the test subjects, which can also lead to false negatives or missed detections.

[0004] Secondly, existing methods for detecting drunk driving are often delayed. In most cases, the testing process needs to be conducted some time after drinking, making immediate detection impossible. This means that even a considerable period after the peak drinking period, a driver may still be considered to be driving under the influence.

[0005] Furthermore, existing methods for detecting drunk driving can interfere with legitimate drivers. For example, certain non-alcoholic substances may trigger false alarms, leading to innocent drivers being wrongly convicted of driving under the influence. This is not only unfair to the wrongly convicted drivers but also disrupts the normal operation of the entire transportation system.

[0006] More importantly, the accuracy of existing drunk driving detection methods is generally low. Due to various reasons, such as the precision of the detection equipment and differences in operator skills, the actual detection results often deviate significantly from the true situation.

[0007] Furthermore, existing methods for detecting drunk driving are inefficient in real-time. In most cases, the detection process requires the driver to stop for inspection, which undoubtedly increases road congestion and affects the real-time nature of the detection. This is a significant obstacle for traffic situations requiring emergency assistance or time-sensitive situations.

[0008] Furthermore, existing methods for detecting drunk driving have limited coverage. In many areas, insufficient equipment or technological limitations prevent comprehensive drunk driving detection, meaning that in certain regions or under specific circumstances, drunk driving may not be effectively curbed.

[0009] It is evident that existing technologies for detecting drunk driving have problems such as missed detections, delayed inspections, interference with normal drivers, low accuracy, poor real-time efficiency, and narrow coverage. There is an urgent need to improve and perfect them through new technologies and methods. Summary of the Invention

[0010] Purpose of the invention: To address the shortcomings of existing technologies in accurately, in real-time, and specifically detecting drunk driving vehicles, this invention proposes a method for identifying abnormal driving states of drunk driving vehicles that integrates drones. This method solves the problems of missed detections, delayed inspections, and interference with normal drivers in existing drunk driving detection methods, thereby improving the accuracy, real-time efficiency, and coverage of drunk driving detection.

[0011] Technical solution: The present invention integrates a drone-based method for identifying abnormal driving conditions of drunk vehicles, which is implemented through a data processing module and a data generation module; the data processing module includes an information acquisition module, a driver and vehicle judgment module, and a data fusion module; the data generation module includes a preliminary data output module, a signal receiving module, a driver's language and expression judgment module, and a terminal information judgment module.

[0012] The present invention integrates the method for identifying abnormal driving status of drunk vehicles using drones, comprising the following steps:

[0013] Step (1): Within the designated monitoring area, identify and obtain vehicle information of the target vehicle;

[0014] Step (2): Establish anomaly detection models for drunk driving vehicle trajectories in three scenarios. The process is as follows:

[0015] (2.1) Extract suspected drunk driving characteristics of the target vehicle;

[0016] (2.2) The degree of abnormality in the driving status of drunk driving vehicles is scored based on three scenarios;

[0017] (2.3) If the total score exceeds the set value, such as 4.8 points, then determine whether there is a novice driver sign;

[0018] (2.4) If a novice driver sign is present, observe and determine the vehicle's trajectory on other road sections.

[0019] Step (3): Based on the vehicle power, mass, and average speed data, define the suspected drunk driving risk index for the target vehicle. The process is as follows:

[0020] (3.1) Define three scenarios for abnormal drunk driving detection as swaying, sudden acceleration, and abnormal driving at traffic light intersections;

[0021] (3.2) The main indicators for judging abnormal vehicle trajectory models in swaying scenarios involve curvature, deviation, and rate of change;

[0022] (3.3) The main indicators for judging trajectory anomalies in sudden acceleration scenarios involve time parameters and acceleration;

[0023] (3.4) The main indicators for judging trajectory anomalies in traffic light intersection scenarios involve time parameters and acceleration.

[0024] The trajectory anomaly model in the swaying scenario in step (3.2) is as follows:

[0025] Vehicle trajectory anomalies are identified by indicators such as curvature, deviation, rate of change, speed, distance, and angle. This invention constructs a trajectory anomaly detection model based on three indicators: curvature, deviation, and rate of change, to identify whether drunk driving occurs during vehicle operation. Taking multiple factors into account, this trajectory anomaly detection model assigns three indicators: curvature weight η0, deviation weight η1, and rate of change weight η2. For example, if these are 0.4, 0.3, and 0.3, the output comprehensive index N, representing the credibility of drunk driving behavior, is calculated using the following weighted formula:

[0026] N=η0*k(t0)+η1*d(t0)+η2*r(t0) (1.18)

[0027] Where k(t0) is the trajectory curvature value at time t0, d(t0) is the trajectory deviation value at time t0, and r(t0) is the trajectory change rate value at time t0.

[0028] In this invention, to find the equation representing the vehicle's trajectory curve, a simulation model is established using VISSIM. A vehicle model and related parameters are set on a randomly generated lane. The width of the motor vehicle lane is set to 3.5 meters, and the operating speed is 45 km / h. Based on the vehicle's random trajectory, the coordinates of the time and the vehicle trajectory projection onto the normal vector direction are captured, and a rectangular coordinate system is established. A scatter plot of the time parameter x(t) = t on the horizontal axis and the original data on the y-axis is generated, and Excel is used for fitting calculations. After multiple fittings, the curve equation of the vehicle trajectory with respect to the time parameter t is found to be:

[0029]

[0030] Where A, B, C, and D are constant coefficients.

[0031] (3.2.1) The change in the curvature of the trajectory reflects the driver's operational stability and judgment ability. Automated curvature analysis can reduce manual intervention and judgment, improve detection efficiency, and shorten response time.

[0032] The curvature of the trajectory is between 0.01 and 0.1. If it exceeds this range, it may indicate drunk driving. Neglecting error, this trajectory anomaly detection model sets the curvature threshold to 0.15. If the trajectory curvature is greater than 0.15, the vehicle is judged to be driving under the influence of alcohol; if the trajectory curvature is less than or equal to 0.15, the vehicle is considered normal.

[0033] The curvature of the vehicle's trajectory at this point is expressed as:

[0034]

[0035] Where x(t) and y(t) are the horizontal and vertical coordinates of the trajectory, x'(t) and y'(t) are the first derivatives of the trajectory, x"(t) and y"(t) are the second derivatives of the trajectory, and k(t) is the curvature of the trajectory, i.e. the degree of curvature of the trajectory. It reflects the direction and angle of the vehicle's motion. If the curvature is too large or too small, it indicates that the vehicle's motion is unstable or abnormal.

[0036] If the vehicle trajectory satisfies formula (1.20), it means that the confidence level of judging the vehicle as drunk driving behavior when using curvature as the discrimination index is 40%, and the corresponding trajectory curvature value k(t0) is output.

[0037] Assign corresponding scores based on the set thresholds:

[0038]

[0039] (3.2.2) By using trajectory deviation analysis, the driver's status is monitored in real time during driving to promptly detect potential drunk driving behavior. When abnormal deviation is detected, a warning is immediately issued or corresponding control measures are taken to ensure driving safety.

[0040] Given formula (1.19), after multiple fittings, the general curve equation of the vehicle trajectory with respect to the time parameter t is found to be:

[0041]

[0042] Where y(t) is the equation function of the vehicle trajectory curve with time t as a parameter.

[0043] The deviation of the trajectory is between 0.1 and 0.5. If it exceeds this range, it may indicate drunk driving. The error is negligible. This trajectory anomaly detection model sets the deviation threshold to 0.55. If the trajectory deviation is greater than 0.55, the vehicle is judged to be driving under the influence of alcohol. If the deviation is ≤0.55, the vehicle trajectory is normal.

[0044] The formula for the deviation of a known trajectory is expressed as:

[0045]

[0046] Where x(t) and y(t) are the parametric equations of the time parameter t and the vehicle trajectory with respect to the time parameter t, x r (t) is the x-coordinate of the reference trajectory, and d(t) is the deviation of the trajectory.

[0047] This can be expressed using the following mathematical expression:

[0048]

[0049] The present invention sets the threshold to 0.55. If the formula (1.23) is satisfied, it indicates that the confidence level of the vehicle being judged as drunk driving when the deviation is used as the discrimination index is 30%. At this time, the corresponding trajectory deviation value d(t0) is output.

[0050] Assign corresponding scores based on the set thresholds:

[0051]

[0052] (3.2.3) By analyzing the rate of change in the driver's driving trajectory, it is possible to determine whether the driver is driving under the influence of alcohol within a short period of time. After drinking alcohol, the driver's reaction time and accuracy will be affected, resulting in an abnormal rate of change in the driving trajectory. By monitoring this rate of change, drunk driving behavior can be detected in a timely manner, improving the immediacy of detection.

[0053] Based on a function model of curvature with respect to the time parameter t:

[0054]

[0055] The rate of change of the trajectory is between 0.1 and 0.5. If it exceeds this range, it may indicate drunk driving. Neglecting error, this trajectory anomaly detection model sets the rate of change threshold to 0.55. If the rate of change of the trajectory is greater than 0.55, the vehicle is judged to be driving under the influence of alcohol; if the rate of change is ≤0.55, the vehicle trajectory is considered normal. The trajectory rate of change is expressed as:

[0056]

[0057] Where k(t) is the curvature of the trajectory, t is time, and r(t) is the rate of change of the trajectory.

[0058] This invention sets a threshold of 0.55. If the trajectory change rate is greater than 0.55, the confidence level of the vehicle being suspected of drunk driving is 30%, and the corresponding trajectory change rate value r(t0) is output.

[0059] Assign corresponding scores based on the set thresholds:

[0060]

[0061] Where, x i S is the discrimination index for the abnormal vehicle trajectory model of the i-th vehicle in the swaying scenario. i Let be the score of the vehicle trajectory anomaly discrimination index for the i-th vehicle in the swaying scenario.

[0062] The trajectory anomaly model in the sudden speed change scenario in step (3.3) is as follows:

[0063] Step (3.3.1) involves using a camera to extract abnormal vehicle speed information and analyzing the changes in speed to determine if the driver is driving under the influence of alcohol. The vehicle's travel time is divided into infinitely close time points, and the speed function is expressed as a first-order difference polynomial using the difference quotient method for precise analysis.

[0064] (1) When the road test equipment and the vehicle-mounted equipment exchange information, the communication delay of the UAV is negligible.

[0065] (2) The friction between the tires and the road is negligible when the vehicle is in motion.

[0066] Typically, determining whether a driver is under the influence of alcohol on a straight road segment primarily relies on factors such as vehicle trajectory and speed. This trajectory anomaly detection model mainly analyzes changes in vehicle speed. Cameras extract abnormal vehicle speed information for a specific road segment for analysis and judgment. Taking a smooth straight road segment as an example, the observation time for the vehicle trajectory is set to 60 seconds (0≤t≤60), and the duration will be expressed in seconds below. Let the distance traveled by the vehicle on this road segment be L, and the vehicle speed on the road be v. limit The speed range is 10 km / h to 60 km / h. The vehicle enters the camera's communication range at time t0, with an initial velocity of v0. The instantaneous velocity of the vehicle during its journey is denoted as V. n (n = 1, 2, 3…n), the time during the journey is divided into infinitely close time points t. n (n = 1, 2, 3…n). Express the velocity function as a first-order difference polynomial using the difference quotient method:

[0067]

[0068] Among them, L1(t), L2(t), L3(t)...L n (t) represents the basis functions of the Lagrange interpolation polynomial; the instantaneous velocity of the vehicle during its movement is V. n (n = 1, 2, 3…n), the time during the journey is divided into infinitely close time points t. n (n=1,2,3…n); then the objective function of the instantaneous acceleration a with respect to t during the vehicle's movement is expressed as: Let L n The numerator of (t) is P(t) = (t-t1)(t-t2)...(tt) n-1 ), let L n The denominator of (t) is Q(t) = (t n -t1)(t n -t2)…(t n -t n-1 ),but

[0069]

[0070]

[0071] Step (3.3.2) calculates acceleration based on vehicle mass, taking into account power output and transmission coefficient factors, and calculates net thrust using maximum power, maximum torque, and average speed parameters. Upon entering the camera's communication range, the vehicle model is identified, and specific parameters are retrieved from the database; vehicle acceleration is then calculated using a formula.

[0072] Calculating acceleration based on vehicle mass requires considering several factors, including the vehicle's mass, power output, and transmission coefficient. The following formula is used to calculate vehicle acceleration:

[0073]

[0074] Where a represents the vehicle's acceleration, F represents the vehicle's net thrust, and M represents the vehicle's mass.

[0075] When calculating net thrust, the vehicle's power output and transmission coefficient must be considered. Generally, net thrust is calculated using the vehicle's maximum power and maximum torque, as shown in the following formula:

[0076]

[0077] Where F represents net thrust, P represents maximum power, and the average vehicle speed is taken as the speed value v. limit The transmission coefficient takes into account vehicle transmission efficiency and transmission losses, and its value ranges from 0.8 to 0.9. Considering all these factors, the following calculation formula is obtained:

[0078]

[0079] When a vehicle enters the camera's communication range, the camera identifies the vehicle model and retrieves specific parameters of that model from its database, such as maximum power and weight. Among these parameters, a... rea This indicates that the vehicle's acceleration is within a reasonable range. P represents the vehicle's maximum power (W), M represents the vehicle's mass (kg), v represents the vehicle's speed (m / s), and 0.8 represents a typical value for the transmission coefficient. This formula is used to estimate the vehicle's acceleration and evaluate its performance.

[0080]

[0081] The reasonable range of acceleration for this vehicle model is expressed as follows:

[0082]

[0083] Step (3.3.3) states that the reasonable acceleration range varies depending on the vehicle model and configuration. Three parameters are set within the reasonable acceleration range, dividing the range into four equal parts, and corresponding scores are assigned according to the range's standards.

[0084] The objective function of the instantaneous acceleration 'a' of the vehicle during travel with respect to time 't' in this trajectory anomaly detection model is expressed as:

[0085]

[0086] Because the power P and mass M vary across different vehicle models and configurations, the optimal acceleration range also differs for each model. Therefore, three parameters are set within this optimal acceleration range. Will Divided into four equal parts, among which They correspond to as Points at 1 / 2 and 3 / 4, Assume the standard for the interval of sudden speed change is: Based on the interval criteria, assign corresponding scores, as follows:

[0087]

[0088] Among them, T i The fraction assigned to a vehicle when it suddenly changes gears, y i It represents the instantaneous acceleration of a vehicle during its movement.

[0089] In step (3.4), the time parameters and acceleration involved in the trajectory anomaly model discrimination index in the traffic light intersection scenario are as follows:

[0090] (1) When the road test equipment and the vehicle-mounted equipment exchange information, the communication delay of the UAV is negligible.

[0091] (2) The friction between the tires and the road is negligible when the vehicle is in motion.

[0092] For ease of analysis, the traffic light status is divided into green and red (including the yellow light period). The green light duration is given as 30 seconds, and the red light duration as 33 seconds. Let L be the communication range between the vehicle and the camera, and let v be the vehicle speed on a typical road. limit Assuming the speed range is between 10 km / h and 60 km / h, and the vehicle enters the camera's communication range at time t. i At this moment, the initial velocity of the vehicle is v. i The driver's reaction time at the stop line at the traffic light is t. j The driver's reaction time after the stop line at the traffic light is t. q The time when the vehicle arrives at the stop line at the intersection is t. p The time of crossing the signalized intersection is t. r .

[0093] Calculating acceleration based on vehicle mass requires considering several factors, including the vehicle's mass, power output, and transmission coefficient. The following formula is used to calculate vehicle acceleration:

[0094]

[0095] Where a represents the vehicle's acceleration, F represents the vehicle's net thrust, and M represents the vehicle's mass.

[0096] When calculating net thrust, the vehicle's power output and transmission coefficient are considered. The net thrust is calculated using the vehicle's maximum power and maximum torque, as shown in the following formula:

[0097]

[0098] Where F represents net thrust, P represents maximum power, and the average vehicle speed is taken as the speed value v. limit The transmission coefficient takes into account vehicle transmission efficiency and transmission losses, and its value ranges from 0.8 to 0.9. Considering all these factors, the following calculation formula is obtained:

[0099]

[0100] Among them, a rea This indicates that the vehicle's acceleration is within a reasonable range. P represents the vehicle's maximum power (W), M represents the vehicle's mass (kg), v represents the vehicle's speed (m / s), and 0.8 represents a typical value for the transmission coefficient. This formula is used to estimate the vehicle's acceleration and evaluate its performance.

[0101]

[0102] (3.4.1) When a vehicle is approaching an intersection, if the traffic light ahead is green but the duration is very short or it is a red light phase that has just started, the vehicle will be unable to pass through the intersection before the traffic light phase ends. This process takes into account the vehicle model, mass and maximum power, and is evaluated by setting a reasonable acceleration range and assigning a score.

[0103] When a vehicle enters the communication range of the intersection's cameras, the system retrieves the vehicle's specific model from the database (determining the vehicle's mass and maximum power) and calculates the reasonable acceleration for that model. If the system detects that the traffic light ahead is green for a very short time, or that the traffic light has just turned red (i.e., the remaining green light time is less than 10 seconds, or the remaining red light time is more than 10 seconds), then the vehicle cannot pass through the intersection before the current phase cycle ends, regardless of acceleration or deceleration. The vehicle must stop and wait for the next green light to turn on before proceeding. In other words, sudden braking or running a red light is considered drunk driving. This process is described by the following mathematical expression:

[0104]

[0105] Among them, v j The reaction time t is the time it takes for a driver to react when they see a green light phase that has only been in the camera's line of sight for a very short time or when a red light phase has just started. j The corresponding instantaneous velocity; v p To reach the stop line t p The instantaneous velocity corresponding to a given moment.

[0106] Because the power P and mass M vary across different vehicle models and configurations, the optimal acceleration range also differs for each model. Therefore, three parameters are set within this optimal acceleration range. Will Divided into four equal parts, among which They correspond to as Points at 1 / 2 and 3 / 4, Assume the standard for the interval of sudden speed change is: Based on the interval criteria, assign corresponding scores, as follows:

[0107]

[0108] Among them, U i The fraction assigned to z when a vehicle suddenly changes gears. i It represents the instantaneous acceleration of a vehicle during its movement.

[0109] (3.4.2) When the vehicle reaches the control area of ​​the intersection, retrieve the specific vehicle model from the database (specifying the vehicle's mass and maximum power) and calculate the reasonable acceleration 'a' for that vehicle model. rea If a driver learns that the remaining time on the green light ahead is between 10 and 30 seconds, and accelerates through normally, this is considered reasonable behavior; otherwise, it is considered drunk driving. A reasonable driving procedure can be described by the following mathematical expression:

[0110] V(t)=L1(t)·V1+L2(t)·V2+L3(t)·V3+…+L n (t)·V n

[0111]

[0112]

[0113] st0≤t j ≤10

[0114] 10≤t r ≤30

[0115]

[0116]

[0117] Among them, v j The reaction time t is the time taken when the driver sees the green light phase after entering the camera's line of sight and there is enough time to proceed. j The corresponding instantaneous velocity; v r To help drivers speed through traffic light intersectionsr The instantaneous velocity corresponding to a given moment.

[0118] Because the power (P) and mass (M) vary across different vehicle models and configurations, the optimal acceleration range also differs. Therefore, three parameters are set within this optimal acceleration range. Will Divided into four equal parts, among which They correspond to as Points at 1 / 2 and 3 / 4, Assume the standard for the interval of sudden speed change is: Based on the interval criteria, assign corresponding scores, as follows:

[0119]

[0120] Among them, U i The fraction assigned to z when a vehicle suddenly changes gears. i It represents the instantaneous acceleration of a vehicle during its movement.

[0121] (3.4.3) When a vehicle is traveling within the control area of ​​an intersection and it is known that the traffic light ahead is red with less than 10 seconds remaining, and the vehicle is maintaining its current speed and traveling to the stop line at the traffic light intersection, the next green light has not yet been turned on. At this time, the vehicle must stop and wait for the next green light to be turned on to proceed; that is, first decelerate to the stop line and then accelerate a few seconds after the green light turns on. If the vehicle stops or accelerates suddenly, it is considered drunk driving.

[0122] Based on the experiment by Zheng Xin et al. on the impact of red light waiting time on driver reaction time at intersections, this study uses a cognitive psychology reaction time research paradigm to explore the influence of red light waiting time on drivers. By setting different red light presentation times, the reaction time of participants upon seeing the green light was recorded, and the significance of the different red light presentation times on driver reaction time was compared. The results showed that the average reaction time was fastest when the red light waiting time was 80 seconds. From 40 to 60 seconds, the average reaction time decreased, reaching its fastest point at 80 seconds. From the red light presentation time of 40 to 80 seconds, the average reaction time showed a decreasing trend, reaching a minimum at 80 seconds. When the red light presentation time was 100 seconds, the average reaction time increased again, reaching its maximum at 40 seconds. We assigned weights of 40% and 60% to deceleration and acceleration behaviors before and after the stop line, respectively.

[0123] This process can be described by the following mathematical expression:

[0124] Deceleration behavior before the stop line:

[0125]

[0126] Among them, v j The reaction time t is when the driver enters the camera's line of sight and the red light phase is insufficient for passage. j The corresponding instantaneous velocity; v p To reach the stop line t p The instantaneous velocity corresponding to a given moment.

[0127] The behavior of vehicles accelerating when the green light turns on:

[0128]

[0129] Among them, v q The reaction time t is when the driver accelerates through as soon as the green light turns on. q The corresponding instantaneous velocity; v r To help drivers speed through traffic light intersections r The instantaneous velocity corresponding to a given moment.

[0130] Because the power P and mass M vary across different vehicle models and configurations, the optimal acceleration range also differs for each model. Therefore, three parameters are set within this optimal acceleration range. Will Divided into four equal parts, among which They correspond to as Points at 1 / 2 and 3 / 4, Assume the standard for the interval of sudden speed change is: Based on the interval criteria, assign corresponding scores, as follows:

[0131]

[0132] Among them, U i The fraction assigned to z when a vehicle suddenly changes gears. i It represents the instantaneous acceleration of a vehicle during its movement.

[0133] Step (4): Compare the suspected drunk driving risk indicator with the set suspected drunk driving risk indicator threshold. If the threshold is exceeded, a tracking signal is sent to the target vehicle.

[0134] (4.1) Set thresholds for the three discrimination indicators in the swinging scenario, such as 0.15; 0.55; 0.55;

[0135] (4.2) In the variable speed scenario, the time parameter is controlled within the set value, and the instantaneous acceleration during vehicle driving is compared with the range of normal acceleration.

[0136] (4.3) At traffic light intersections, classify and analyze the color and remaining duration of traffic lights, and compare the driver's reaction time and instantaneous acceleration with the range of normal reaction time and normal acceleration.

[0137] Step (5): Send out a drone to fly over the vehicle suspected of being under the influence of alcohol, identify the driver's face, and determine whether the driver has any characteristics of drunk driving.

[0138] The process of step (1) is as follows:

[0139] (1.1) Identify the license plate model of the vehicle driven by the driver using a camera;

[0140] (1.2) Upload the specific vehicle model to the cloud to compare with all vehicle models in the database, and retrieve vehicle power, mass, average speed and other parameter information;

[0141] (1.3) Generate data tables to classify and store vehicle power, mass and average speed data.

[0142] The data in step (1.3) are vehicle power, mass, and average speed.

[0143] In step (4.2), the time parameter is controlled within one minute in the sudden speed change scenario, and the instantaneous acceleration during the vehicle's driving process is compared with the range of normal acceleration.

[0144] The process of step (5) is as follows:

[0145] (5.1) Receive location information of suspected drunk driving vehicles from the traffic monitoring system via the wireless communication module;

[0146] (5.2) Control the drone to fly above the suspected drunk driving vehicle based on the location information, and take a face photo of the driver inside the vehicle using the camera;

[0147] (5.3) The captured facial images are sent to the cloud server via a wireless communication module;

[0148] (5.4) Run a face recognition algorithm on a cloud server to analyze the face image and determine whether the driver has the characteristics of drunk driving;

[0149] (5.5) If it is determined that the driver has the characteristics of drunk driving and the score is greater than the threshold of the overall comprehensive index, such as 6.8 points, then the driver is confirmed to be drunk driving. The driver sends a command to the drone through the wireless communication module to make the drone issue a warning signal and send an alarm message to the traffic monitoring system at the same time.

[0150] (5.6) If it is determined that the driver does not have the characteristics of drunk driving, a control signal is sent to the drone through the wireless communication module to make the drone return to its original position.

[0151] In step (5.4), the characteristic factors of drunk driving include facial expression, eye state and head posture.

[0152] In step (5.5), the alarm information includes the vehicle's license plate number, location, and speed.

[0153] Working Principle: This invention utilizes data processing and drones to monitor abnormal vehicle trajectories and identifies drunk driving behavior through scoring. First, the data processing module establishes models for vehicle swaying, sudden acceleration, and waiting time at traffic light intersections using multi-feature evaluation indicators, and assigns scores to the degree of trajectory abnormality. When the total score exceeds a threshold, it determines whether a novice driver is present. Next, the data generation module deploys a drone to recognize the driver's facial features and determine if the driver exhibits characteristics of drunk driving. Finally, the terminal information judgment module confirms whether the overall comprehensive index exceeds a specified threshold to determine if drunk driving has occurred.

[0154] The specific process is as follows: Within the designated monitoring area, the camera identifies and acquires vehicle information of the target vehicle, such as license plate number, vehicle model, and speed, and sends this information to the data processing module. Based on the vehicle information, the data processing module establishes three abnormal trajectory discrimination models for drunk driving vehicles under three scenarios: vehicle swaying discrimination model, vehicle sudden speed change discrimination model, and traffic light intersection waiting time discrimination model, to detect anomalies in the target vehicle's trajectory.

[0155] Based on the obtained data, the data processing module defines suspected drunk driving risk indicators for the target vehicle, such as the curvature, deviation, and rate of change of the trajectory. It assigns scores to these risk indicators and calculates the target vehicle's total drunk driving risk score. The data processing module then compares the total drunk driving risk score with a set threshold for drunk driving risk indicators. If the threshold is exceeded, a tracking signal is issued to the target vehicle and sent to the data generation module.

[0156] Based on the tracking signal, the data generation module dispatches a drone to fly over the vehicle suspected of being under the influence of alcohol, identifies the driver's face, and determines whether the driver has characteristics of drunk driving, such as facial expression, eye state, head posture, etc., and sends the judgment results to the data processing module.

[0157] Based on the judgment result, the data processing module verifies the drunk driving status of the target vehicle. If drunk driving is confirmed, it reports to the relevant department; otherwise, it terminates the tracking.

[0158] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0159] (1) The present invention can determine whether the vehicle’s driving trajectory is abnormal, thereby timely detecting vehicles suspected of drunk driving and improving the efficiency of identifying drunk driving vehicles.

[0160] (2) This invention integrates multiple feature discrimination indicators, such as the curvature, deviation and rate of change of the trajectory, to establish a discrimination model for abnormal driving trajectory, and assigns scores to the degree of trajectory abnormality, thereby quantitatively assessing the degree of drunk driving risk of the vehicle and improving the risk assessment efficiency of drunk driving vehicles.

[0161] (3) The present invention sets a threshold for suspected drunk driving based on the scoring of the degree of trajectory abnormality, thereby effectively distinguishing drunk driving vehicles from non-drunk driving vehicles and improving the reliability of drunk driving vehicle identification.

[0162] (4) This invention uses drones to capture the driver’s facial state, such as facial expression, eye state, head posture, etc., to further determine whether the driver has the characteristic factors of drunk driving, thereby improving the accuracy of drunk driving vehicle identification. Attached Figure Description

[0163] Figure 1 This is a flowchart of the method for identifying abnormal driving status of drunk vehicles using drones, which is based on the present invention.

[0164] Figure 2 This is a flowchart illustrating the process of obtaining driver behavior information according to the present invention.

[0165] Figure 3 This is a model diagram of the characteristics of drunk driving according to the present invention;

[0166] Figure 4 This is a diagram of the vehicle acceleration box of the present invention;

[0167] Figure 5 This is a fitting diagram of the vehicle swaying trajectory of the present invention;

[0168] Figure 6 This is a schematic diagram of the structure of the UAV of the present invention. Detailed Implementation

[0169] Example:

[0170] like Figure 1 As shown, the present invention integrates a method for identifying abnormal driving status of drunk vehicles using drones. It uses GIS technology to obtain the real-time location coordinates of the vehicle. The method is implemented using seven modules: vehicle information acquisition module, driver and vehicle operation judgment module, data fusion module, preliminary data output module, signal receiving module, driver language and expression judgment module, and terminal information judgment module.

[0171] The signal receiving module includes a signal receiver, which is used to receive information transmitted by the signal transmitter.

[0172] The terminal information judgment module includes an information storage device and a judgment device. The information storage device is used to store real-time two-dimensional image information, including vehicle location information, and the judgment device is used to comprehensively judge whether the driver is driving under the influence of alcohol.

[0173] This invention takes the Volkswagen Jetta VA3230TSI as an example, with a maximum power ranging from 150,000W to 280,000W and a weight range from 1375kg to 1740kg. The road width is set to 3.5 meters, the operating speed to 45 km / h, and the observation time to 60 seconds. Based on the vehicle's random driving trajectory, the coordinates of the time and the vehicle trajectory projection onto the normal vector direction are captured, establishing a rectangular coordinate system. A scatter plot of X and Y is generated by combining the time parameter x(t) = t on the horizontal axis with the original data on the y-axis, and then fitted using Excel. After multiple fitting operations, as shown... Figure 5 As shown, the equation of the vehicle trajectory curve with respect to the time parameter t is expressed as:

[0174] y = 0.0002t 3 -0.0078t 2 +0.0876t+1.5738(1.50)

[0175] Based on the sway model calculations, the vehicle's curvature value between 2.85 and 4.75 seconds is greater than 0.15, earning 10 points; the deviation is less than 0.55, earning 0 points; and the rate of change is greater than 0.55, earning 10 points. Figure 3 The overall output index is 7 points.

[0176] The acceleration value of the vehicle calculated based on the sudden change model is as follows: Figure 4 Within the range of [5.56, 8.64], assign 6 points.

[0177] Based on the traffic light intersection waiting time discrimination model, it is determined that the vehicle has between 10 and 30 seconds remaining before it learns that the traffic light ahead is green. The calculated acceleration value of the vehicle is... Figure 4 Within the range of [8.65, 11.73], assign 8 points. Therefore, based on... Figure 3 The overall score is 5.65.

[0178] The comprehensive index of 2 is greater than the set threshold of 4.8 points. Therefore, further judgment is made as to whether the vehicle has a novice driver's license plate. The result is that there is no novice driver's license plate. Figure 3 The overall output index is 6.65 points.

[0179] A drone was dispatched to fly over the suspected drunk driver's vehicle and perform facial recognition. The results indicated that the driver exhibited characteristics of drunk driving. Figure 3The output comprehensive index 4 is 7.65 points, which is greater than the prescribed threshold of 6.8 points for the overall comprehensive index. Therefore, it is confirmed that the vehicle is driving under the influence of alcohol. The command is sent to the drone through the wireless communication module, so that the drone issues a warning signal and sends an alarm message to the traffic monitoring system at the same time.

Claims

1. A method for identifying abnormal driving conditions of drunk driving vehicles using unmanned aerial vehicles (UAVs), characterized in that: Includes the following steps: Step (1): Within the designated monitoring area, identify and obtain vehicle information of the target vehicle; Step (2): Establish anomaly detection models for drunk driving vehicle trajectories in three scenarios. The process is as follows: (2.1) Extract suspected drunk driving characteristics of the target vehicle; (2.2) The degree of abnormality in the driving status of drunk driving vehicles is scored based on three scenarios; (2.3) If the total score exceeds the set value, determine whether there is a novice driver sign; (2.4) If a novice driver sign is present, observe and determine the vehicle's trajectory on other road sections. Step (3): Define the suspected drunk driving risk indicators for the target vehicle. The process is as follows: (3.1) Define three scenarios for abnormal drunk driving detection as swaying, sudden acceleration, and abnormal driving at traffic light intersections; (3.2) The main indicators for judging abnormal vehicle trajectory models in swaying scenarios involve curvature, deviation, and rate of change; Given three criteria—curvature weight η0, deviation weight η1, and rate of change weight η2—the comprehensive index N of the credibility of drunk driving behavior is calculated using the following weighted formula: N=η0*k(t0)+η1*d(t0)+η2*r(t0)(1.1) Where k(t0) is the trajectory curvature value at time t0, d(t0) is the trajectory deviation value at time t0, and r(t0) is the trajectory change rate value at time t0. The curvature of the vehicle's trajectory is expressed as: Where x(t) is the abscissa of the trajectory, y(t) is the ordinate of the trajectory, x'(t) and y'(t) are the first derivatives of the trajectory, x"(t) and y"(t) are the second derivatives of the trajectory, and k(t) is the curvature of the trajectory, i.e. the degree of curvature of the trajectory. Assign corresponding scores based on the set thresholds: The formula for trajectory deviation is expressed as: Where x(t) is the x-coordinate of the trajectory, and y(t) is the y-coordinate of the trajectory. r (t) is the x-coordinate of the reference trajectory, and d(t) is the deviation of the trajectory. Assign corresponding scores based on the set thresholds: The rate of change of the trajectory is expressed as: Where k(t) is the curvature of the trajectory, t is time, and r(t) is the rate of change of the trajectory; Assign corresponding scores based on the set thresholds: Where, x i S is the discrimination index for the abnormal vehicle trajectory model of the i-th vehicle in the swaying scenario. i The score of the vehicle trajectory anomaly discrimination index for the i-th vehicle in the swaying scenario; (3.3) The criteria for judging trajectory anomalies in sudden acceleration scenarios involve time parameters and acceleration; Express the velocity function as a first-order difference polynomial using the difference quotient method: The instantaneous speed of the vehicle during its movement is V. n (n = 1, 2, 3…n), the time during the journey is divided into infinitely close time points t. n (n=1,2,3…n), L1(t), L2(t), L3(t)…L n (t) are the basis functions of the Lagrange interpolation polynomial; Remember L n The numerator of (t) is P(t) = (t-t1)(t-t2)...(tt) n-1 ), let L n The denominator of (t) is Q(t) = (t n -t1)(t n -t2)…(t n -t n-1 ),but The objective function of the instantaneous acceleration 'a' of a vehicle during travel with respect to time 't' is expressed as: Where P represents the vehicle's maximum power and M represents the vehicle's mass; Three parameters are set within the reasonable range of acceleration. Will Divided into four equal parts, among which Corresponding to Points at 1 / 4, 1 / 2, and 3 / 4, Assume the standard for the interval of sudden speed change is: Based on the interval criteria, assign corresponding scores, as follows: Among them, T i The fraction assigned to a vehicle when it suddenly changes gears, y i Indicates the instantaneous acceleration of a vehicle during its movement; (3.4) When the phase information of the traffic light ahead is green and the duration is short, or the traffic light ahead is a red light phase that has just been turned on, the following mathematical expression can be used to describe it: Among them, v j The reaction time t is the time it takes for a driver to react when they see a green light phase that has only been in the camera's line of sight for a very short time or when a red light phase has just started. j The corresponding instantaneous velocity; v p To reach the stop line t p The instantaneous velocity corresponding to a given moment; The behavior of vehicles accelerating when the green light turns on: Among them, v q The reaction time t is when the driver accelerates through as soon as the green light turns on. q The corresponding instantaneous velocity; v r To help drivers speed through traffic light intersections r The instantaneous velocity corresponding to a given moment; Set three parameters in the acceleration range. Will Divided into four equal parts, among which They correspond to as Points at 1 / 4, 1 / 2, and 3 / 4, Assume the standard for the interval of sudden speed change is: Based on the interval criteria, assign corresponding scores, as follows: Among them, U i The fraction assigned to z when a vehicle suddenly changes gears. i Indicates the instantaneous acceleration of a vehicle during its movement; Step (4): Compare the suspected drunk driving risk indicator with the set suspected drunk driving risk indicator threshold. If the threshold is exceeded, a tracking signal is sent to the target vehicle. (4.1) Thresholds are set for the three discrimination indicators in the swinging scenario; (4.2) In the variable speed scenario, the time parameter is controlled within the set value, and the instantaneous acceleration of the vehicle is compared with the range of normal acceleration; (4.3) At traffic light intersections, classify and analyze the color and remaining time of traffic lights, and compare the driver's reaction time and instantaneous acceleration with the range of normal reaction time and normal acceleration; Step (5): Send out a drone to fly over the vehicle suspected of being under the influence of alcohol, perform facial recognition on the driver, and determine whether the driver has any characteristics of drunk driving.

2. The method for determining abnormal driving status of drunk driving vehicles using integrated unmanned aerial vehicles according to claim 1, characterized in that: The process of step (1) is as follows: (1.1) Identify the license plate number and model of the vehicle driven by the driver through a camera; (1.2) Upload the model number to the cloud comparison database to retrieve vehicle power, mass, and average speed parameters; (1.3) Generate data tables to classify and store data.

3. The method for determining abnormal driving status of drunk driving vehicles using integrated unmanned aerial vehicles according to claim 2, characterized in that: The data in step (1.3) are vehicle power, mass, and average speed.

4. The method for identifying abnormal driving status of drunk driving vehicles using integrated unmanned aerial vehicles according to claim 1, characterized in that: In step (3.4), when the remaining time of the green light is less than 10 seconds, or the remaining time of the red light is more than 10 seconds, it is represented by the following expression: Among them, v j The reaction time t is the time it takes for the driver to see the green light phase or the red light phase after entering the camera's line of sight. j The corresponding instantaneous velocity; v p To reach the stop line t p The instantaneous velocity corresponding to a given moment.

5. The method for determining abnormal driving status of drunk driving vehicles using integrated unmanned aerial vehicles according to claim 1, characterized in that: In step (3.4), when the remaining time of the green light phase of the current traffic light is between 10 and 30 seconds, the driver's operation process is described by the following mathematical expression: V(t)=L1(t)·V1+L2(t)·V2+L3(t)·V3+…+L n (t)·V n s.t.0≤t j ≤10 10≤t r ≤30 Among them, v j The reaction time t is the time taken when the driver sees the green light phase after entering the camera's line of sight and there is enough time to proceed. j The corresponding instantaneous velocity; v r To help drivers speed through traffic light intersections r The instantaneous velocity corresponding to a given moment.

6. The method for determining abnormal driving status of drunk driving vehicles using fusion drones according to claim 1, characterized in that: In step (3.4), if the remaining time of the red light phase of the traffic light ahead is less than 10 seconds, and the vehicle maintains its current speed to the stop line at the traffic light intersection, the next green light has not yet been turned on. At this time, the vehicle first decelerates to the stop line and then the green light turns on after a few seconds. The vehicle then begins to accelerate. This process is described by the following mathematical expression: Deceleration behavior before the stop line: Among them, v j The reaction time t is when the driver enters the camera's line of sight and the red light phase is insufficient for passage. j The corresponding instantaneous velocity; v p To reach the stop line t p The instantaneous velocity corresponding to a given moment.

7. The method for determining abnormal driving status of drunk driving vehicles using integrated unmanned aerial vehicles according to claim 1, characterized in that: In step (4.2), the time parameter is controlled within a few minutes in the sudden speed change scenario, and the instantaneous acceleration during vehicle driving is compared with the range of normal acceleration.

8. The method for determining abnormal driving status of drunk driving vehicles using fusion drones according to claim 1, characterized in that: The process of step (5) is as follows: (5.1) Receive location information of suspected drunk driving vehicles from the traffic monitoring system via the wireless communication module; (5.2) Control the drone to fly above the suspected drunk driving vehicle based on the location information, and take a face photo of the driver inside the vehicle using the camera; (5.3) The captured facial images are sent to the cloud server via a wireless communication module; (5.4) Run a face recognition algorithm on a cloud server to analyze the face image and determine whether the driver has the characteristics of drunk driving; (5.5) If a driver has characteristics of drunk driving and the value exceeds the threshold of the overall comprehensive index, the driver is confirmed to be driving under the influence of alcohol. The driver sends a command to the drone via the wireless communication module to make the drone issue a warning signal and send an alarm message to the traffic monitoring system at the same time. (5.6) If the driver does not have any characteristics of drunk driving, the drone will be sent a control signal through the wireless communication module to return to its original position.

9. The method for determining abnormal driving status of drunk driving vehicles using integrated unmanned aerial vehicles according to claim 8, characterized in that: In step (5.4), the characteristic factors of drunk driving include facial expression, eye state and head posture.

10. The method for determining abnormal driving status of drunk driving vehicles using fusion drones according to claim 8, characterized in that: In step (5.5), the alarm information includes the vehicle's license plate number, location, and speed.

Images