Driver reaction time detection method for real vehicle scenario, vehicle control method

Abstract

The present application belongs to the field of road vehicle control system, and relates to a driver reaction time detection method and a vehicle control method for real vehicle scene. The driver reaction time detection method first collects the driver's forward field of view data and eye movement data based on double cameras, and obtains vehicle state data. Then, based on a coordinate transformation algorithm, the line of sight direction and the eyeball center coordinates in the infrared camera coordinate system are converted to the scene camera coordinate system and a ray is constructed. The intersection of the ray and the imaging plane of the scene camera is taken as the fixation point, which is projected to the two-dimensional pixel coordinate system of the scene camera imaging. The environmental target information is extracted. The time of driver's danger perception and the operation reaction time are determined in combination with the fixation point and the environmental target information, and the time difference between the two is calculated to obtain the driver's reaction time in real time. This method can realize the automatic and accurate extraction of the danger perception time in the real vehicle environment and obtain the driver's reaction time in real time.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of road vehicle control system, and relates to the calculation or prediction of driving parameters, in particular to a driver reaction time real-time detection method for real vehicle scene and a multi-scene personalized vehicle control method. BACKGROUND

[0002] With the continuous growth of the number of cars, the road traffic flow is increasing, and the safety risks such as rear-end collision, lateral scraping and pedestrian accident occur frequently, which seriously threatens the safety of life and property. In the traffic system composed of people, vehicles and roads, the emergency reaction time of the driver, that is, the whole process of the driver from perceiving the danger stimulus to making an operation reaction (focusing on the driver's active reaction behavior), is the core factor to determine the effect of vehicle safety, and its accurate measurement is crucial to the design of assisted driving system and traffic safety optimization.

[0003] The existing driver reaction time measurement technology has obvious shortcomings and cannot balance accuracy and scene adaptability. In a non-real vehicle scene, such as Chinese patent CN 110236575 A, which uses an eye tracker combined with a driving simulator, although it can complete basic measurement, this scheme relies on the virtual environment constructed by the simulator, which is significantly different from the key factors such as traffic flow and road condition changes in real driving, resulting in that the measurement results cannot be transferred to the real vehicle scene, and essentially it is not applicable to reaction time evaluation in real driving environment. In a real vehicle scene, although Chinese patent CN 107126223 A realizes real vehicle measurement, it still has the following problems: (1) Although the eye tracker is used to capture the changes of the driver's eye focus point in the seven visual areas of the front, left window, right window, left rearview mirror, right rearview mirror, rearview mirror and instrument panel to determine the perception state, it can only capture the moment when the eye gaze area changes, and cannot determine the moment when the driver perceives danger; (2) The eye tracker needs to be worn by the driver and needs complex pre-debugging, which not only has long preparation time and high use threshold, but also causes discomfort to the driver during the wearing process, limits the head movement, and thus interferes with the stability of eye movement data collection. Therefore, it is necessary to provide a new driver reaction time detection method to realize real-time detection in a real vehicle scene. SUMMARY

[0004] In order to solve the problem of inaccurate measurement of existing driver reaction time and dependence on special equipment such as eye tracker, the present application provides a driver reaction time detection method for real vehicle scene, which uses a double-camera vision scheme to replace the special eye tracker to collect the fixation point and realize coordinate mapping, so as to realize the automatic and accurate extraction of the moment of danger perception, and then compare it with the moment of operation reaction to obtain the driver reaction time in real time, and realize the accurate measurement of the driver reaction time.

[0005] To achieve the above purpose, the present application adopts the following technical solutions:

[0006] A method for detecting driver reaction time in real-world vehicle scenarios, comprising the following steps:

[0007] Step S1. Multi-source data acquisition: The driver's forward field of view data is acquired through the scene camera, the driver's eye movement data is acquired through the infrared camera, and the vehicle status data of the vehicle is acquired simultaneously. All acquired data is bound to the unified timestamp of the vehicle system.

[0008] Step S2. Perform distortion correction processing on the infrared and scene camera images; use the cornea-pupil offset to characterize the rotational offset of the eyeball on the infrared imaging plane, thereby constructing the initial gaze vector; obtain the head pose and solve for the rotation matrix of the head relative to the infrared camera. With translation vector Based on the attitude matrix ( , Obtain the coordinates of the eyeball center in the infrared camera coordinate system. Simultaneously through rotation matrix The initial gaze vector is corrected to obtain the gaze direction in the infrared camera coordinate system. Then, based on the coordinate transformation algorithm, the line of sight and eye center coordinates in the infrared camera coordinate system are transformed to the scene camera coordinate system, and a ray is constructed. ; to ray The intersection point with the scene camera's imaging plane is taken as the gaze point, projected onto the scene camera's two-dimensional pixel coordinate system, and the gaze point coordinates are smoothed to output the gaze point coordinates in the scene camera image. And associate with timestamp;

[0009] Step S3. Environmental target information extraction: Use target detection algorithms to identify vehicles ahead, vehicles in adjacent lanes, and pedestrians who have entered the lane. Simultaneously extract the braking status of vehicles ahead, the accurate location and speed of the area where the brake lights are located, the speed of adjacent vehicles, the angle and lateral distance between them and the vehicle, and the position and lateral intrusion speed of pedestrians.

[0010] Step S4. Combining the data obtained in steps S1 to S3, determine the moment when the driver perceives the danger and the moment when the driver reacts, and calculate the time difference between the two to obtain the driver's reaction time in real time.

[0011] As a preferred embodiment of the present invention, the scene camera is fixed at the upper middle position inside the windshield of the vehicle, with the lens facing the road ahead. Its collection range covers the vehicles, lane lines, pedestrians and adjacent lane environment in front of the driver's normal field of vision. It is used to collect continuous video stream data of the road ahead in real time, and each frame of the image is associated with a unified timestamp of the vehicle system.

[0012] The infrared camera is installed in the center above the dashboard, with the lens facing the driver's face and focusing on the eye area. It is used to collect images of the driver's eye movements in real time, record the trajectory of eye movement and the direction of gaze. Each frame of eye movement image is also associated with a unified timestamp of the vehicle system.

[0013] The vehicle status data includes the vehicle speed and brake pedal pressure, and all vehicle status data is bound to a unified timestamp of the vehicle system.

[0014] As a preferred embodiment of the present invention, in step S2, the distortion coefficients obtained by calibration are used to perform distortion correction processing on the infrared and scene camera images; when calculating the corneal-pupil offset, the coordinates of the corneal reflection center and the geometric center coordinates of the pupil are calculated first, and then the difference between the two is used to characterize the rotational offset of the eyeball on the infrared imaging plane, that is, the corneal-pupil offset.

[0015] As a preferred embodiment of the present invention, the rotation matrix of the head relative to the infrared camera in step S2... With translation vector The method for determining these features is as follows: A set of key points is detected in the facial image captured by the infrared camera. These key points include the corners of the eyes, the tip of the nose, and the corners of the mouth. This is then combined with the intrinsic parameters of the infrared camera. The PnP algorithm is used to establish a mapping relationship between points in a 3D face model and their corresponding 2D image points, thereby solving for the rotation matrix of the head relative to the infrared camera. With translation vector .

[0016] As a preferred embodiment of the present invention, in step S2, the line of sight and eye center coordinates in the infrared camera coordinate system are transformed to the scene camera coordinate system based on a coordinate transformation algorithm, and a ray is constructed. The specific method is as follows:

[0017] By comparing the external parameters of the infrared camera and the scene camera The rotation matrix in the image represents the line-of-sight direction in the infrared camera coordinate system. View direction vector transformed into scene camera coordinate system ;

[0018] By comparing the external parameters of the infrared camera and the scene camera The center point of the eye in the infrared camera coordinate system Transform to the scene camera coordinate system to determine the eye reference point in the scene camera coordinate system. ;

[0019] In the scene camera coordinate system, with the eye reference point Starting from the line of sight vector Constructing ray The expression is:

[0020] ;

[0021] wherein, is a ray parameter.

[0022] As a preferred embodiment of the present application, in the step S2 of smoothing the gaze point coordinates, a first-order low-pass filter algorithm is used to smooth the gaze point coordinates, and the filter coefficient is dynamically adjusted, and the expression is:

[0023] ;

[0024] wherein, is a basic filter coefficient, and is taken as ; is a head angular velocity, which is obtained by calculating the rate of change of the rotation matrix of the continuous frames; is the rate of change of the adjacent frame gaze vector; , is an adjustment coefficient;

[0025] The expression of the adjustment coefficient , is:

[0026] ;

[0027] wherein, , is the 3s , 95% quantile of the current frame.

[0028] As a preferred embodiment of the present application, in the step S3 of extracting the adjacent lane vehicle parameters, it is simultaneously determined whether the adjacent lane vehicle “cuts into” the lane, and when the lateral distance , the rate of change of the included angle of the continuous 3 frames and the speed are met, the adjacent lane vehicle is marked as “cutting in”; and based on the extracted pedestrian parameters, it is determined whether the pedestrian has intruded into the lane, and for the pedestrian who has intruded into the lane, it is determined whether it is in a “dangerous” state; if the overlap ratio of the pedestrian bounding box and the lane region is , it is determined that the pedestrian has intruded into the lane; otherwise, it is determined that it has not intruded; if the lateral intrusion speed of the pedestrian who has intruded into the lane is , it is marked as “dangerous”.

[0029] As a preferred embodiment of the present application, the step S4 includes the following steps:

[0030] Step S41. Driver danger perception time Acquisition:

[0031] ①Sensing the hazard of a vehicle braking ahead:

[0032] At timestamp If both of the following conditions are met in a frame, the current timestamp is recorded and used as the moment of danger perception. :

[0033] The coordinates of the driver's gaze point in this frame match the coordinates of the brake light area at the rear of the vehicle ahead.

[0034] The vehicle in front of this frame is braking;

[0035] ② Danger perception of vehicles cutting into adjacent lanes:

[0036] At timestamp If both of the following conditions are met in a frame, the current timestamp is recorded and used as the moment of danger perception. :

[0037] The driver's gaze coordinates in this frame match the bounding box coordinates of adjacent vehicles.

[0038] The vehicle status in the adjacent lane of this frame is "cutting in";

[0039] ③ Pedestrian intrusion danger perception:

[0040] At timestamp If both of the following conditions are met in a frame, the current timestamp is recorded and used as the moment of danger perception. :

[0041] The driver's gaze point coordinates match the pedestrian bounding box coordinates in this frame;

[0042] The pedestrian's dangerous situation in this frame is "dangerous";

[0043] Step S42. Operation reaction time Acquisition:

[0044] When an upward trend is detected in the brake pedal pressure signal, this moment is recorded as a candidate moment. If the pressure signal shows a sustained and significant increase within a preset time, exceeding a set threshold, it is confirmed that the driver has performed an emergency braking operation, and the previously recorded timestamp is determined as the reaction moment. If the condition for continuous increase is not met, the candidate moment will not be considered as the operational response moment.

[0045] Step S43. Timestamp Alignment Calibration:

[0046] Step S44. Reaction time calculate:

[0047] the moment of danger perception the moment of operation reaction , i.e. the reaction time of the driver .

[0048] The application also provides a vehicle control method for real vehicle scenarios, which formulates personalized vehicle control strategies based on different driver reaction times to improve road traffic safety.

[0049] A vehicle control method for real vehicle scenarios, which comprises the following steps:

[0050] Step A1. Calculate the collision time under different dangerous stimulus scenarios based on the identified dangerous stimulus scenarios respectively;

[0051] Step A2. Calculate the time required for the driver to complete collision avoidance based on the obtained driver reaction time;

[0052] Step A3. For each identified dangerous stimulus scenario, calculate its safety degree according to the collision time and the time required for the driver to complete collision avoidance, and when more than two dangerous stimulus scenarios are monitored, calculate the comprehensive overall safety degree of all dangerous stimulus scenarios , the expression is:

[0053] ;

[0054] Among them, is the set of current dangerous stimulus scenarios, is the safety degree of the corresponding scenario; is the dynamic weight coefficient of the corresponding scenario, satisfying , the expression is:

[0055] ;

[0056] Among them, is a very small positive number, is the calculated collision time of the corresponding scenario;

[0057] Step A4. According to the different calculation results of the comprehensive overall safety degree , execute the differentiated safety intervention strategy adapted to the difference of each safety state.

[0058] As a preferred embodiment of the application, in step A4, no intervention is needed, and the driver completes the operation according to his own judgment, and the system only continuously monitors the vehicle state and the driver's reaction;

[0059] if , triggering an audible and visual warning, including a red dashboard warning light flashing continuously and a high-frequency warning sound emitted by the vehicle's speakers, prompting the driver to take timely braking or avoidance measures;

[0060] If , starting the vehicle automatic collision avoidance mechanism, automatically applying maximum braking force for emergency deceleration, and combining with the steering auxiliary system to fine-tune the direction to avoid collision, and synchronously activating the double flash lights to prompt the surrounding vehicles.

[0061] The advantages and benefits of the present application are as follows:

[0062] 1. The present application proposes a technical solution that can obtain the driver's reaction time in real time in a real vehicle environment, so as to calculate the total time required for the driver to complete the risk avoidance in real time according to the driver's perception and operation state, and realize the calculation of safety degree combined with the collision time TTC, thereby realizing the hierarchical safety intervention decision and vehicle control, forming a complete closed loop from danger perception, operation reaction to control intervention, and effectively improving the road traffic safety and system intelligence level.

[0063] 2. The present application proposes a dual-camera solution using an infrared camera and a scene camera. The dual-camera eye movement tracking solution can effectively replace the expensive and complex eye tracker, realizing non-contact and low-cost tracking of the driver's eye movement. This solution not only avoids the defects of complicated operation and interference with driving of the eye tracker, but also automatically extracts the dangerous perception moment through real-time capture of eye and head movement characteristics by stereo vision, reduces the error of manual intervention, and adapts to the real vehicle scene, providing a feasible path for efficient and accurate measurement of the driver's reaction time.

[0064] 3. The present application proposes a gaze point mapping method based on cornea-pupil offset and dual-camera fusion. The initial gaze vector is constructed using an infrared camera, and through external parameter calibration and posture correction, it is uniformly mapped to the imaging plane of the scene camera. Compared with the traditional method which only relies on the geometric relationship of cornea-pupil or monocular imaging, this method can better eliminate the deviation caused by head posture changes, and realize real-time capture of the driver's gaze point in the real road scene image.

[0065] 4. The present application proposes a filter coefficient adaptive adjustment method based on head angular velocity and gaze vector change rate, which replaces the traditional fixed coefficient low-pass filter to realize the dynamic balance of gaze point smoothing processing in "stable gaze" and "rapid saccade" scenes, significantly improving the stability and real-time performance of the gaze point output, and improving the feasibility and practicality of vehicle-mounted applications.

[0066] 5. The present application proposes to use whether the gaze point falls within the target bounding box as the criterion for the driver's perception of danger, to realize automatic identification of the dangerous perception moment, replace manual annotation, and complete the transition from "region level" to "target level", significantly improving the accuracy and real-time performance of the judgment.

[0067] 6. The application calculates the total time required for the driver to complete collision avoidance by combining the driver's reaction time and the vehicle's braking response time in the safety intervention decision-making link, and calculates the safety degree under different dangerous stimulus scenarios according to the collision time and the time required for the driver to complete collision avoidance, and forms a hierarchical safety intervention decision-making mechanism according to the safety degree in a single scenario.

[0068] 7. The application proposes a comprehensive overall safety degree calculation method covering single scenario and multi-scenario superposition, and quantifies the available time adequacy by the ratio of collision time to collision avoidance time in a single scenario, and calculates the comprehensive overall safety degree by dynamically allocating weights according to the collision time of each scenario in a multi-scenario, replacing the traditional direct comparison of collision avoidance time and collision time for a single scenario, realizing the unified quantitative evaluation of single scenario and multi-dangerous superposition scenario safety state, and providing accurate basis for subsequent differentiated safety intervention decision-making. BRIEF DESCRIPTION OF DRAWINGS

[0069] Fig. 1 The application provides a driver reaction time detection method for real vehicle scenarios;

[0070] Fig. 2 The application provides a gaze point coordinate mapping diagram;

[0071] Fig. 3 The application provides a dangerous recognition flowchart;

[0072] Fig. 4 The application provides a vehicle control method for real vehicle scenarios. DETAILED DESCRIPTION

[0073] In order for those skilled in the art to better understand the technical solutions of the application and their advantages, the application will be described in detail below with reference to the drawings, but not used to limit the protection scope of the application.

[0074] Example 1:

[0075] As shown in the figure, the embodiment provides a driver reaction time detection method for real vehicle scenarios, which replaces the special eye tracker with a double camera vision scheme to collect gaze points and realize coordinate mapping, so as to realize automatic and accurate extraction of dangerous perception moment, and then compare it with operation reaction moment to obtain driver reaction time in real time, the specific steps are as follows: Figs. 1 to 3

[0076] ​Step S1. Multi-source data acquisition: Collect the driver's forward view data through a scene camera (also known as a scene camera or scene camera) facing the scene in front of the driver's field of view, collect the driver's eye movement data through an infrared camera (also known as an infrared camera or infrared camera) facing the driver, and simultaneously collect vehicle state data such as vehicle speed and brake pedal pressure. All collected data is bound to the vehicle-mounted system unified timestamp;

[0077] Specifically, in this embodiment, the scene camera is fixed to the middle and upper position inside the vehicle front windshield, with the lens facing the front road. Its collection range covers the front vehicles, lane lines, pedestrians and adjacent lane environments within the driver's normal field of view, and is used to collect continuous video stream data of the front road in real time. Each frame of image is associated with the vehicle-mounted system unified timestamp.

[0078] The infrared camera is installed in the central position above the instrument panel, with the lens directly facing the driver's face and focusing on the eye area, for real-time collection of the driver's eye movement images, recording of the eye rotation trajectory and line of sight direction. Each frame of eye movement image is also associated with the vehicle-mounted system unified timestamp.

[0079] In this embodiment, the vehicle speed sensor is accessed through the vehicle CAN bus interface to obtain the vehicle speed data in real time. A high-precision pressure sensor is integrated at the mechanical connecting rod of the brake pedal to convert the analog signal of the pedal force into a digital signal through a single-chip microcomputer module, and the brake pedal pressure data is collected synchronously. The above vehicle state data is bound to the vehicle-mounted system unified timestamp, and the time reference is consistent with the camera data.

[0080] Step S2. Distortion correction of infrared and scene camera images; use the cornea-pupil offset to represent the rotation offset of the eyeball on the infrared imaging plane to construct the initial gaze vector; obtain the head pose and solve the rotation matrix of the head relative to the infrared camera and the translation vector , based on the pose matrix , to obtain the eyeball center coordinates in the infrared camera coordinate system , and correct the initial gaze vector through the rotation matrix to obtain the line of sight direction in the infrared camera coordinate system ; then based on the coordinate transformation algorithm, convert the line of sight direction and eyeball center coordinates in the infrared camera coordinate system to the scene camera coordinate system and construct the ray ; the intersection of the ray and the imaging plane of the scene camera is taken as the gaze point, which is projected to the two-dimensional pixel coordinate system of the scene camera imaging, and the gaze point coordinates are smoothed to output the gaze point coordinates in the scene camera image and associate the timestamp.

[0081] Specifically, in the embodiment, the step S2 realizes accurate mapping of the driver's gaze point in the actual field of view picture, and the specific process is as follows:

[0082] Step S21. Double camera (scene camera, infrared camera) parameter calibration:

[0083] According to the pinhole imaging model, the projection of a space point in the camera coordinate system can be expressed as:

[0084]

[0085] wherein, is a homogeneous coordinate scale factor; , is the pixel coordinate in the camera imaging plane; , , is the focal length of the camera in the , axis direction, , is the principal point coordinate; , is the extrinsic matrix and translation vector of the camera; , , is the three-dimensional coordinate of the space point in the camera coordinate system.

[0086] The intrinsic and extrinsic parameters of the infrared camera and the scene camera are calibrated by Zhang Zhengyou calibration method, and the intrinsic parameters of the infrared camera and the distortion coefficient , the intrinsic parameters of the scene camera and the distortion coefficient , and the relative extrinsic parameters of the infrared camera to the scene camera are obtained, which provide the basis for subsequent coordinate conversion, and the expression is:

[0087]

[0088] wherein, is a rotation matrix; is a translation vector.

[0089] In addition, in order to realize the conversion of pixels and actual physical distance, the pixel-actual distance conversion coefficient needs to be solved in the calibration process, and a calibration object or road lane line with a known actual size is selected on the imaging plane of the scene camera, the physical length of which is , and the pixel interval in the image plane is , so:

[0090]

[0091] The coefficient can be approximately kept unchanged when the camera installation height and the pitch angle are fixed, and can be used for directly converting the pixel distance into the actual lateral or longitudinal distance in the vehicle coordinate system.

[0092] Step S22. Image de-warping;

[0093] Due to the radial and tangential distortion of the lens, the de-warping processing is needed for the infrared and scene camera images by using the distortion coefficients obtained by the calibration, and the de-warping formula is as follows:

[0094]

[0095] wherein, is the original pixel coordinate of the spatial point before de-warping in the camera imaging plane; is the normalized pixel coordinate after de-warping; and the focal length are the radial distance from the principal point to the principal point after de-warping of the normalized distortion coordinate ; is the coordinate after de-warping, , , , , is the distortion coefficient.

[0096] Step S23. Extraction of corneal reflection center and pupil center:

[0097] In order to obtain the corneal reflection point, a high-low threshold combination method is adopted, that is, a higher threshold is used to quickly locate the rough area of the light spot in the gray image, and then a lower threshold is used to extract the region of interest (ROI) in the area, and finally the gray weighted centroid formula is used to accurately calculate the corneal reflection center coordinates , the expression is:

[0098]

[0099] wherein, is the pixel gray value of the coordinate in the ROI.

[0100] In the aspect of pupil center extraction, first, the eye region is binarized under different groups of thresholds, and the ratio of the "pupil area / enclosing rectangle area" is calculated. When the ratio is maximum, it is considered that the threshold segmentation result at this time is optimal, and the pupil region is obtained by using the threshold. Subsequently, the gradient operator is used to extract the feature points of the pupil edge, and the geometric center coordinates of the pupil are obtained by using the ellipse fitting method .

[0101] The cornea-pupil offset is obtained , and the expression is

[0102]

[0103] The corneal reflection center is basically stable in the imaging plane and can be used as a reference point; the pupil center is offset with the rotation of the eyeball, so the difference between the two can effectively represent the rotation offset of the eyeball in the infrared imaging plane.

[0104] Step S24. Head pose estimation:

[0105] To eliminate the influence of the driver's head movement on the gaze point calculation, the head pose needs to be obtained at the same time. A set of key points (eye corners, nose tip, mouth corners, etc.) can be detected in the face image collected by the infrared camera (infrared camera). Combined with the intrinsic parameters of the infrared camera (infrared camera) , the PnP algorithm is used to establish a mapping relationship between the three-dimensional face model points and their corresponding two-dimensional image points, so as to solve the rotation matrix and the translation vector of the head relative to the infrared camera, which not only provides a basis for subsequent line-of-sight direction calculation and pose compensation, but also can obtain the head angular velocity by calculating the rate of change of the rotation matrix of the continuous frames, which provides support for adaptive adjustment of the filtering coefficient; at the same time, since the fixed position of the eyeball center in the face coordinate system is predefined in the three-dimensional face model, the position point can be mapped to the infrared camera coordinate system by using the pose matrix , , so as to obtain the eyeball center coordinates , which is used as the starting point for constructing the gaze ray in the scene camera coordinate system, providing a reference for gaze point projection calculation.

[0106] Step S25. Gaze vector construction and coordinate system conversion:

[0107] Based on the above features and pose information, first, the initial gaze vector is constructed, and the expression is

[0108]

[0109] The line-of-sight direction in the infrared camera coordinate system is obtained by correcting the head pose rotation matrix :

[0110]

[0111] Subsequently, the relative extrinsic parameters between the infrared camera and the scene camera are combined with the rotation matrix in the relative extrinsic parameters , and the line-of-sight direction vector in the scene camera coordinate system is converted :

[0112]

[0113] Step S26. Gaze point projection:

[0114] The relative extrinsic parameters obtained by the dual-camera calibration The eye center point in the infrared camera coordinate system is converted to the scene camera coordinate system, and the expression of the eye reference point in the scene camera coordinate system is

[0115]

[0116] In the scene camera coordinate system, the ray is constructed with the line-of-sight direction vector as the starting point, and the expression is

[0117]

[0118] wherein is the ray parameter, which is used to describe the scale factor in space when the eye center extends along the line of sight, and this parameter does not need to be directly measured, but is naturally determined by the intersection condition of the ray and the imaging plane of the scene camera under the imaging geometric constraint.

[0119] The ray intersects the imaging plane of the scene camera to obtain the intersection point , which is taken as the gaze point, and the internal parameter matrix of the scene camera is used to project the intersection point from the three-dimensional coordinate system of the scene camera to the two-dimensional pixel coordinate system of the scene camera imaging, and the expression is

[0120] wherein

[0121] is the initial pixel coordinate of the gaze point in the scene image, is the homogeneous coordinate scale factor.

[0122] Step S27. Gaze point smoothing and output:

[0123] A first-order low-pass filtering algorithm is used to smooth the gaze point coordinates to reduce the influence of high-frequency jitter, and the expression is

[0124] ​​​​

[0125] in, The optimized gaze point coordinates from the previous frame. These are the filter coefficients.

[0126] Furthermore, to overcome the shortcomings of traditional fixed filter coefficients in balancing stability and real-time performance, this invention proposes an adaptive modeling method for filter coefficients based on the driver's dynamic state:

[0127] When the driver's head rotation angular velocity is large, the line of sight changes rapidly, at which point increasing... To reduce filtering lag; when the rate of change of the gaze vector is large (rapid saccades), it also improves This is to ensure the timeliness of gaze tracking.

[0128] Taking all the above factors into account, the filter coefficients Dynamically adjusted to:

[0129]

[0130] in, Based on the basic filter coefficients, take ; This refers to the magnitude of the head's angular velocity. The rate of change of the gaze vector (view direction vector) between adjacent frames; , This is the adjustment coefficient.

[0131] Adjustment coefficient , The expression is:

[0132]

[0133] in, , 3 seconds before the current frame , Choosing the 95th percentile strikes a balance between robustness and sensitivity. Compared to the mean or median, the 95th percentile avoids interference from outliers and, compared to the upper quartiles, better reflects the high-amplitude dynamic characteristics of drivers under rapid gaze shifts or abrupt head movements, providing a reference close to the true extreme values. Modeling with this can more scientifically characterize extreme but common fluctuations in drivers, thereby improving the rationality and applicability of adaptive adjustment of the filter coefficients.

[0134] Step S3. Environmental target information extraction: Based on the forward field of view data collected in step S1, the target detection algorithm is used to identify vehicles in front (including the rear of the vehicle and the taillight area), vehicles in adjacent lanes, and pedestrians who have entered the lane. Key information such as the speed of the vehicles in front, the angle between the adjacent vehicles and the vehicle, and the position of the pedestrians are extracted simultaneously.

[0135] In particular, the step S3 comprises the following steps:

[0136] Step S31. Lane line detection and current lane region definition:

[0137] Identify the left and right boundary line pixel coordinates of the current lane through the lane line detection algorithm, fit to generate left and right boundary equations, left line , right line , construct the lane region pixel mask as , the expression is:

[0138]

[0139] Synchronize the output of the fitting parameters of the left and right lane lines for subsequent adjacent lane definition and target position judgment.

[0140] Step S32. Multi-target detection and tracking:

[0141] Use the YOLOv11 algorithm to detect each frame of the front view image, identify three types of targets and output the boundary box pixel coordinates:

[0142] Front vehicle: labeled label "car_front", boundary box coordinates , , , ), wherein , is the horizontal and vertical pixel coordinates of the top left corner of the front vehicle boundary box, , is the horizontal and vertical pixel coordinates of the right bottom corner of the front vehicle boundary box;

[0143] Adjacent lane vehicle: labeled label "car_adjacent", boundary box coordinates , , , ), wherein , is the horizontal and vertical pixel coordinates of the top left corner of the adjacent lane vehicle boundary box, , is the horizontal and vertical pixel coordinates of the right bottom corner of the adjacent lane vehicle boundary box;

[0144] Pedestrian: labeled label "person", boundary box coordinates , , , ), wherein , horizontal and vertical pixel coordinates of the top-left corner of the bounding box of the pedestrian, , horizontal and vertical pixel coordinates of the bottom-right corner of the bounding box of the pedestrian.

[0145] The SORT algorithm is used to track the three types of targets respectively: Kalman filter is used to predict the motion trajectory of each target, Hungarian algorithm is used to associate the detection boxes of consecutive frames with the existing trajectories, and a unique ID is assigned to each target to ensure that it is continuously tracked in the video sequence. The coordinates of the bounding box of all targets are associated with the current frame timestamp .

[0146] Step S33. Front vehicle feature extraction: based on the extracted features, the braking state of the front vehicle is determined.

[0147] For the target with label "car_front", the tail features and motion parameters are extracted:

[0148] The RGB image of the tail area of the front vehicle is intercepted, converted to HSV color space, and the HSV threshold is set to extract the brake light area. The image is binarized, and the closed operation is performed on the segmented binary image. The change rate of the white pixel ratio in the brake light area in consecutive frames is calculated. By comparing the change of the white pixel ratio in consecutive frames, the working state of the brake light is obtained. At the same time, the brake light is accurately positioned within the bounding box of the tail of the front vehicle. The accurate position of the brake light area is identified by a deep learning model (such as NanoDet model) , , , ), wherein , horizontal and vertical pixel coordinates of the top-left corner of the bounding box of the brake light, , horizontal and vertical pixel coordinates of the bottom-right corner of the bounding box of the brake light;

[0149] Based on the longitudinal coordinates of the center of the vehicle bounding box in the last two frames , , combined with the camera intrinsic parameter conversion to actual longitudinal distance change, the speed of the front vehicle is calculated , the expression is:

[0150]

[0151] wherein, is the pixel-actual distance conversion coefficient.

[0152] Step S34. Adjacent lane vehicle parameter extraction: based on the extracted adjacent lane vehicle parameters, it is determined whether the adjacent lane vehicle "cuts into" the lane.

[0153] For the target with label "car_adjacent", the following key parameters are extracted:

[0154] Based on the center coordinates of the bounding box of the two consecutive frames 、 , the actual displacement is calculated to obtain the speed, and the expression is:

[0155]

[0156] Taking the center of the vehicle head as the origin, an polar coordinate system is established, the lateral distance of the center of the adjacent vehicle bounding box is , and the longitudinal distance is , then the expression of the included angle is:

[0157]

[0158] The shortest pixel distance between the adjacent vehicle bounding box and the lane boundary line is calculated, and the actual distance is converted into the lateral distance ;

[0159] When the lateral distance , the change rate of the included angle for 3 consecutive frames and the speed are met, the adjacent lane vehicle is marked as "cut in".

[0160] Step S35. Pedestrian parameter extraction, based on the extracted pedestrian parameters, judge whether the pedestrian has crossed into the lane, and for the "pedestrian has crossed into the lane", judge whether it is in "dangerous" state;

[0161] The pixel set of the pedestrian bounding box is The expression is:

[0162] ;

[0163] Calculate the overlap ratio with the lane area The expression is:

[0164]

[0165] If , it is determined that "the pedestrian has crossed into the lane"; otherwise, it is determined that "has not crossed".

[0166] For the "pedestrian has crossed into the lane", based on the lateral coordinates of the center of the bounding box of the two consecutive frames 、 , the lateral intrusion speed is calculated, and the expression is:

[0167]

[0168] If , it is marked as "dangerous".

[0169] Step S4. Combine the gaze point mapping result obtained in step S2, the environmental target information extracted in step S3, and the brake pedal pressure data in step S1 to determine the moment of driver's danger perception and the moment of operation reaction, calculate the time difference between the two to obtain the driver's reaction time in real time.

[0170] Specifically, step S4 includes the following steps:

[0171] Step S41. Obtain the moment of driver's danger perception :

[0172] ① Front vehicle braking danger perception:

[0173] When the following two conditions are met in the frame with timestamp , record the current timestamp and take it as the moment of danger perception :

[0174] The gaze point coordinates of the driver in this frame match the brake light region coordinates of the rear of the front vehicle, that is, ;

[0175] The braking state of the front vehicle in this frame is "braking" (determined by the working state of the brake light of the front vehicle, if the brake light is on, the front vehicle is braking, otherwise it is not braking).

[0176] ② Adjacent lane vehicle cut-in danger perception:

[0177] When the following two conditions are met in the frame with timestamp , record the current timestamp and take it as the moment of danger perception :

[0178] The gaze point coordinates of the driver in this frame match the adjacent vehicle bounding box coordinates, that is, ;

[0179] The state of the adjacent lane vehicle in this frame is "cut-in".

[0180] ③ Pedestrian intrusion danger perception:

[0181] When the following two conditions are met in the frame with timestamp , record the current timestamp and take it as the moment of danger perception :

[0182] The gaze point coordinates of the driver in this frame match the pedestrian bounding box coordinates, that is, ;

[0183] The pedestrian danger state is "dangerous".

[0184] Step S42. Operation reaction time acquisition:

[0185] The pressure analog signal of the high-precision pressure sensor integrated in the brake pedal is collected in real time by the single-chip microcomputer system, and is converted into a digital quantity and transmitted to the processing unit. The processing unit monitors the received pressure digital signal in real time. When the rising trend of the pressure signal is detected, the time at this moment is recorded as a candidate time. If the pressure signal shows a continuous and significant rise within a preset time (short time) thereafter, and exceeds the set threshold, it is determined that the driver has performed an emergency braking operation, and the previously recorded time stamp is determined as the operation reaction time . If the continuous rising condition is not met, the candidate time is not the operation reaction time.

[0186] Step S43. Time stamp alignment calibration:

[0187] The video data collected by the camera and the pressure analog signal data collected by the pressure sensor are time-stamped and calibrated by the unified system clock of the vehicle-mounted computing platform when the data is received. The time stamp of the dangerous perception time marked in the camera video stream is read by the vehicle-mounted computing unit , and the operation reaction time timestamp transmitted by the brake pedal pressure sensor, directly extract the timestamp value under the unified reference, eliminate the time reference difference between devices, and ensure that and are in the same time coordinate system.

[0188] Step S44. Reaction time calculation:

[0189] Calculate the difference between the dangerous perception time and the operation reaction time , that is, the reaction time of the driver , and the calculation formula is:

[0190]

[0191] If the calculation result is positive, it means that the operation reaction is after the dangerous perception, which meets the normal reaction logic, and the result is output. If the result is negative, it is determined that the time synchronization is abnormal, and the result is not recorded.

[0192] Example 2:

[0193] The embodiment is based on the personalized vehicle control method formulated according to the reaction time of different drivers obtained in embodiment 1, which calculates the time required by the driver to complete collision avoidance according to the real-time obtained driver reaction time, calculates the collision time (TTC) according to the identified dangerous stimulus scene, calculates the safety degree according to the time required by the driver to complete collision avoidance and the collision time, dynamically judges the risk level, and formulates and executes differentiated safety intervention strategies.

[0194] In the embodiment, the collision time is calculated for the three types of dangerous stimulus scenes identified in step S3, as follows:

[0195] ① The TTC of the front vehicle braking scene is:

[0196] The speed of the host vehicle is obtained , the speed of the front vehicle is obtained , and the distance between the vehicles is obtained . The collision time TTC in this scene is calculated by the relative motion model , and the expression is:

[0197]

[0198] ② The TTC of the adjacent lane vehicle cutting-in scene is:

[0199] The lateral distance between the adjacent lane vehicle and the host vehicle is obtained , the lateral relative speed is obtained , the longitudinal distance is obtained , and the longitudinal relative speed is obtained . The collision time TTC in this scene is calculated , and the expression is:

[0200]

[0201] ③ The TTC of the pedestrian cutting-in scene is:

[0202] The longitudinal speed of the host vehicle is obtained , the lateral intrusion speed of the pedestrian is obtained , the lateral distance between the pedestrian and the lane boundary is obtained , and the longitudinal distance between the host vehicle and the current position of the pedestrian is obtained . The collision time TTC in this scene is calculated , and the expression is:

[0203]

[0204] In the embodiment, in order to reasonably divide the differentiated strategies, the time required by the driver to complete collision avoidance and the collision time of the corresponding scene need to be considered comprehensively. The expression of the time required by the driver to complete collision avoidance is as follows:

[0205]

[0206] wherein, is the driver reaction time; is the vehicle braking response time, is taken as 1.0s, is taken as 2.1s, is taken as 3.0s.

[0207] In addition, in order to overcome the intervention error that may occur in a single scene decision in a complex traffic environment, on the basis of obtaining the time required by the driver to complete the danger avoidance and the collision time of each single scene , a safety degree calculation method covering single scene and multiple scenes is introduced in the embodiment, which can perform cooperative safety state evaluation when any one, two or three of the three scenes of front vehicle braking, adjacent lane vehicle cutting in and pedestrian intrusion occur at the same time.

[0208] Specifically, for each dangerous stimulus scene identified in step S3, the safety degree is calculated, and the calculation formula is:

[0209]

[0210] wherein, is the calculated corresponding scene collision time, corresponding to front vehicle braking, adjacent lane vehicle cutting in and pedestrian intrusion, respectively.

[0211] The safety degree directly represents the adequacy of the available time relative to the required time. When , it indicates that the available time is adequate, and is in a safe state; when , the available time is only slightly more than the required time, and the margin is limited, and is in a critical state; when , it indicates that the available time is insufficient, and is in a dangerous state.

[0212] When the system simultaneously monitors two or three dangerous stimulus scenes, the comprehensive overall safety degree of all dangerous stimulus scenes is calculated, which is specifically realized through the following weighted fusion formula:

[0213]

[0214] wherein, is the current set of dangerous stimulus scenes; is the dynamic weight coefficient of the corresponding scene, satisfying .

[0215] The smaller the value of the scene collision time, the greater the weight of the scene in the overall safety degree evaluation, so the dynamic weight coefficient of the multiple scenes is determined by the following formula based on the emergency level:

[0216]

[0217] wherein, is a very small positive number, such as 0.01, to avoid numerical instability problems when tends to zero.

[0218] For different calculation results of the single scene and the overall safety degree of the multi-scene comprehensive whole, the differentiated safety intervention strategy suitable for each safety state is executed: Specifically, if

[0219] , no intervention is needed, and the driver completes the operation according to his own judgment, and the system only continuously monitors the vehicle state and the driver's reaction. If

[0220] , a sound and light warning is triggered, including a red warning light on the dashboard flashing continuously and a high-frequency warning sound emitted by the vehicle-mounted loudspeaker, prompting the driver to take braking or avoidance measures in time. If

[0221] , the vehicle automatic collision avoidance mechanism is started, the maximum braking force is automatically applied for emergency deceleration, and the steering auxiliary system is combined to fine-tune the direction to avoid collision, and the double flash lamp is activated to prompt the surrounding vehicles. In this embodiment, the single scene safety degree

[0222] , the above method realizes the unified quantitative evaluation of the safety state of the single scene and the multi-danger superimposed scene. The application further provides an electronic device, comprising: one or more processors, a memory; wherein the memory is used to store one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors realize the vehicle control method described above.

[0223] The application further provides a computer readable medium, which stores a computer program, and the computer program is executed by a processor to realize the vehicle control method described above.

[0224]

[0225] ​Those skilled in the art can understand that all or part of the functions of various methods / modules in the above embodiments can be implemented by hardware or by a computer program. When all or part of the functions are implemented by a computer program, the program can be stored in a computer readable storage medium, which can include read-only memory, random access memory, magnetic disk, optical disk, hard disk, etc. The above functions are implemented by executing the program by a computer. For example, the program is stored in the memory of the device, and the above functions are implemented by executing the program in the memory by the processor.

[0226] In addition, when all or part of the functions in the above embodiments are implemented by a computer program, the program can also be stored in a storage medium such as a server, another computer, a disk, an optical disk, a flash disk, or a mobile hard disk, and then downloaded or copied to the memory of the local device, or the system of the local device is updated, and then the program in the memory is executed by the processor to implement all or part of the functions in the above embodiments.

[0227] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or replacements within the technical scope disclosed by the present application can be easily thought by those skilled in the art, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope described in the claims.

Claims

1. A driver reaction time detection method for a real vehicle scenario, characterized by, The method comprises the following steps: Step S1. Multi-source data acquisition: collecting the driver's forward field of view data through a scene camera, collecting the driver's eye movement data through an infrared camera, and synchronously collecting the vehicle state data of the vehicle, all of which are bound to the unified timestamp of the vehicle-mounted system; Step S2. Perform de-warping on the infrared and scene camera images; use the cornea-pupil offset to represent the rotation offset of the eyeball on the infrared imaging plane to construct an initial gaze vector; obtain the head pose and solve the rotation matrix R of the head relative to the infrared camera hd and the translation vector t hd , obtain the eyeball center coordinate E hd in the infrared camera coordinate system based on the pose matrix (R hd , t d ), and correct the initial gaze vector by the rotation matrix R hd to obtain the line-of-sight direction g d in the infrared camera coordinate system; then convert the line-of-sight direction and the eyeball center coordinate in the infrared camera coordinate system to the scene camera coordinate system based on the coordinate transformation algorithm and construct the ray P(λ); take the intersection of the ray P(λ) and the imaging plane of the scene camera as the gaze point, project it to the two-dimensional pixel coordinate system of the scene camera imaging, perform smoothing processing on the gaze point coordinate, output the gaze point coordinate (u final , v final ) in the scene camera image and associate the time stamp; Step S3. Environment target information extraction: using a target detection algorithm to identify the front vehicle, adjacent lane vehicle and pedestrian intruding into the lane, and synchronously extracting the braking state of the front vehicle, the accurate position of the brake light area and the speed, the speed of the adjacent vehicle, the angle and lateral distance with the vehicle, and the position and lateral intrusion speed of the pedestrian; Step S4. Combining the data obtained in steps S1 to S3, determining the time of the driver's hazard perception and the time of the operation reaction, and calculating the time difference between the two to obtain the driver's reaction time in real time; In step S2, the gaze point coordinates are smoothed by using a first-order low-pass filtering algorithm, and the filtering coefficient a is dynamically adjusted, and the expression is: a = a0 + k1|ω| + k2Δg; Wherein, a0 is a basic filter coefficient, taking 0.4; ||ω|| is the head angular velocity, which is obtained by calculating the change rate of the rotation matrix R hd of the continuous frames; Δg is the change rate of the adjacent frame gaze vector; k1, k2 are adjustment coefficients; The expressions of the adjustment coefficients k1 and k2 are: where ω 95 , Δg 95 are the quantiles of the previous frame for the 3s∣∣ω∣∣, Δg95% 2. The driver reaction time detection method for a real vehicle scenario according to claim 1, characterized by, The scene camera is fixed to the middle and upper position of the inside of the vehicle front windshield, with the lens facing the front road, and its collection range covers the front vehicle, lane line, pedestrian and adjacent lane environment within the normal field of view of the driver, for real-time collection of continuous video stream data of the front road, and each frame of image is associated with the unified timestamp of the vehicle-mounted system; The infrared camera is installed in the central position above the instrument panel, with the lens facing the driver's face and focusing on the eye area, for real-time collection of the driver's eye movement image, recording of the eye rotation track and line of sight direction, and each frame of eye movement image is also associated with the unified timestamp of the vehicle-mounted system; The vehicle state data includes the vehicle speed and brake pedal pressure, and the vehicle state data is bound to the unified timestamp of the vehicle-mounted system.

3. The real vehicle scenario-oriented driver reaction time detection method according to claim 1, characterized in that, In step S2, the distortion coefficients obtained by calibration are used to perform distortion correction on the infrared and scene camera images; when calculating the cornea-pupil offset, the corneal reflection center coordinates and the geometric center coordinates of the pupil are calculated respectively, and then the difference between the two is used to represent the rotation offset of the eyeball on the infrared imaging plane, i.e. the cornea-pupil offset.

4. The real vehicle scenario-oriented driver reaction time detection method according to claim 1, characterized in that, Rotation matrix R of the head relative to the infrared camera in step S2 hd Translation vector t hd The determination method is as follows: a group of key points are detected in the face image collected by the infrared camera, the key points include the corners of the eyes, the tip of the nose and the corners of the mouth, and the internal parameter K d of the infrared camera is combined to establish a mapping relationship between the three-dimensional face model points and the corresponding two-dimensional image points by using a PnP algorithm, so that the rotation matrix R hd and the translation vector t hd of the head relative to the infrared camera are solved.

5. The real vehicle scenario-oriented driver reaction time detection method according to claim 1, characterized in that, In step S2, based on the coordinate transformation algorithm, the line of sight direction and eyeball center coordinates in the infrared camera coordinate system are converted to the scene camera coordinate system and the specific method of constructing the ray P(λ) is: The line-of-sight direction g in the infrared camera coordinate system is converted to the line-of-sight direction vector g in the scene camera coordinate system by the rotation matrix in the relative extrinsic parameter T between the infrared camera and the scene camera d→f d f ;​​ Through the relative extrinsic parameters T of the infrared camera and the scene camera d→f Convert the eye center point E in the infrared camera coordinate system to the scene camera coordinate system, thereby determining the eye reference point E in the scene camera coordinate system d Convert the eye center point E in the infrared camera coordinate system to the scene camera coordinate system, thereby determining the eye reference point E in the scene camera coordinate system f ; In the scene camera coordinate system, taking the eye reference point E f as the starting point, combining the line-of-sight direction vector g f , a ray P(λ) is constructed, and the expression is: P(λ) = E f + λg f ; Wherein, λ is the ray parameter.

6. The real vehicle scenario-oriented driver reaction time detection method according to claim 1, characterized by, In step S3, after extracting the parameters of the vehicles in the adjacent lanes, it is determined whether the vehicles in the adjacent lanes "cut in" the lane of the host vehicle. When the lateral distance d lat ≤1.5 m, the change rate of the included angle Δθ / Δt>5° / s for 3 consecutive frames, and the speed v adj ≥0, the vehicle in the adjacent lane is marked as "cutting in". Based on the extracted pedestrian parameters, it is determined whether the pedestrian has crossed into the lane of the host vehicle, and whether the pedestrian in the "crossed into the lane" state is in a "dangerous" state. If the overlap ratio R between the pedestrian bounding box and the lane region is ≥20%, it is determined that the pedestrian has crossed into the lane. Otherwise, it is determined that the pedestrian has not crossed into the lane. If the lateral intrusion speed v x of the pedestrian in the "crossed into the lane" state is ≥0 m / s, it is marked as "dangerous".

7. The driver reaction time detection method for a real vehicle scenario according to claim 6, characterized by, Step S4 comprises the following steps: Step S41. Obtaining the driver's hazard perception time t1: ① Front vehicle braking hazard perception: When the following two conditions are met in the frame with timestamp t, the current timestamp is recorded and taken as the hazard perception time t1: The gaze point coordinates of the driver in this frame match the brake light area coordinates of the rear of the front vehicle; The brake state of the front vehicle in this frame is "braking"; ② Adjacent lane vehicle cutting-in hazard perception: When the following two conditions are met in the frame with timestamp t, the current timestamp is recorded and taken as the hazard perception time t1: The gaze point coordinates of the driver in this frame match the adjacent vehicle bounding box coordinates; The state of the adjacent lane vehicle in this frame is "cutting in"; ③Pedestrian intrusion danger perception: When the following two conditions are met in the frame with timestamp t, then record the current timestamp as the danger perception time t1: The frame driver's gaze point coordinates match the pedestrian bounding box coordinates; The frame pedestrian danger state is "danger"; Step S42. Operation reaction time t2 acquisition: When the brake pedal pressure signal starts to show an upward trend, record the time as a candidate time. If the pressure signal shows a continuous and significant upward trend within a preset time, exceeding the set threshold, it is confirmed that the driver has performed emergency braking operation, and the previously recorded timestamp is determined as the operation reaction time t2; If the continuous upward condition is not met, the candidate time is not the operation reaction time; Step S43. Timestamp alignment calibration: Step S44. Reaction time t calculation: Calculate the difference between the danger perception time t1 and the operation reaction time t2, which is the driver's reaction time t.

8. A vehicle control method for a real vehicle scenario, characterized by The method comprises the following steps: Step A1. Calculate the collision time under different dangerous stimulus scenes based on the identified dangerous stimulus scenes; Step A2. Calculate the time required for the driver to complete collision avoidance based on the driver reaction time obtained by the driver reaction time detection method for real vehicle scene in any one of claims 1 to 7; Step A3. For each identified dangerous stimulus scene, calculate its safety degree according to the collision time and the time required for the driver to complete collision avoidance. When more than two dangerous stimulus scenes are monitored, calculate the comprehensive overall safety degree S of all dangerous stimulus scenes, and the expression is: S =∑ i∈A (ω i ·s i ); Wherein, A is the set of current dangerous stimulation scene, s i Safety degree of corresponding scene; ω i Dynamic weight coefficient of corresponding scene, satisfying ∑ω i =1, expression is: where ε is a small positive number, TTC i is the calculated time-to-collision for the corresponding scenario; Step A4. According to the different calculation results of the comprehensive overall safety degree S, execute the differentiated safety intervention strategy adapted to each safety state.

9. The vehicle control method according to claim 8, characterized by, In step A4, if S>1.25, no intervention is needed, and the driver completes the operation according to his own judgment, and the system only continuously monitors the vehicle state and the driver's reaction; If 1<S≤1.25, trigger the sound and light warning, including the continuous flashing of the red warning light on the instrument panel and the high-frequency warning sound emitted by the vehicle-mounted loudspeaker, prompting the driver to take braking or avoidance measures in time; If S≤1, start the vehicle automatic collision avoidance mechanism, automatically apply maximum braking force for emergency deceleration, and at the same time, combine with the steering auxiliary system to fine-tune the direction to avoid collision, and activate the double flash lamp to prompt the surrounding vehicles.

Images