Multi-factor Dynamic Change Blind Spot Monitoring Method, Vehicle Control Method and System

Abstract

The present invention belongs to the field of road vehicle control systems, and relates to a multi-factor dynamic change blind area monitoring method, a vehicle control method and a system. The blind area monitoring method collects the driver's head state, vehicle movement data and changes in the surrounding environment through a data acquisition module, and sends them to a data processing and fusion module for data preprocessing and spatio-temporal synchronization to achieve data alignment, and at the same time fuse multi-source data; then divides the main blind areas outside the vehicle into a front blind area, a rear blind area, an A-pillar blind area, a B-pillar blind area, and a C-pillar blind area, calculates the basic blind area angles of each area based on the size of the vehicle's fixed structure and its distance from the driver, and finally establishes independent dynamic models for each blind area and a multi-blind area dynamic linkage model, and calculates the dynamic angle range of each area's blind area based on the multi-blind area dynamic linkage model to solve the accident problems caused by the change of the blind area during vehicle driving.

Description

Technical Field

[0001] The present invention belongs to the field of road vehicle control systems, and particularly relates to a multi-factor dynamic change blind spot monitoring method, a vehicle control method and a system. Background Art

[0002] With the development of the automotive industry, the negative impacts it brings to urban road traffic have become increasingly prominent. The frequent occurrence of traffic accidents caused by the increase in the number of automobiles and illegal driving behaviors has caused casualties and economic losses, becoming a common problem faced by many cities. From the main causes of accidents: being rear-ended by other vehicles in the blind spot of vision, not noticing the side danger when driving at high speed, and the change of the blind spot when turning right are several common types in automotive traffic accident types.

[0003] In order to reduce the occurrence of automotive accidents, relevant departments have taken a series of safety measures, including improving road safety facilities and promoting safe driving behaviors. However, due to the high driving speed of automobiles, irregular blind spots, and complex driving conditions, the accident rate of automobiles has not decreased significantly, and there are still difficult problems to be solved in the intelligent blind spot detection devices of automobiles. In particular, the detection measures for the dynamic visual blind spots of automobiles are particularly lacking. Therefore, strengthening the detection of automotive dynamic visual blind spots is of great significance.

[0004] Currently, to solve the problem of automotive blind spot detection, the measures taken mainly include intelligent blind spot detection devices, strengthening professional training efforts, adding monitoring and alarm systems, and implementing humanized design and service improvements. With the rapid development of artificial intelligence technology, intelligent assisted driving technology has gradually become a hot research and application field. As one of the core technologies of intelligent assisted driving technology, the Blind Spot Detection (BSD) system has made remarkable progress in sensor technology, algorithm optimization and system integration in recent years. Its core goal is to continuously monitor the areas that are out of sight around the vehicle (such as the A-pillar, the rear of the vehicle, the inner wheel difference, etc.) through sensors such as radar, cameras, and ultrasonic waves, and combine with an early warning mechanism (acoustic and optical prompts or automatic intervention) to reduce the collision risk. The current technological development shows the following characteristics: Sensor technology is dominant. Millimeter-wave radar has become the mainstream solution because of its strong anti-interference ability and long detection distance (up to 50 meters), but its cost is relatively high; Although the camera solution has a low cost, it is prone to failure in harsh environments such as rain, fog, and strong light.

[0005] Chinese Patent CN 117818622 A calculates the blind spot angle through the driver's head position data and the vehicle pillar position data, generates real-time detection and warning, and optimizes the detection accuracy by combining the driver's perspective. Chinese Patent CN 222329589 U discloses a rearview mirror integrated camera and blind spot detection system, which embeds the camera into the rearview mirror bracket, monitors the rear blind spot in real time, and reduces potential safety hazards during lane change or parking through a prompting system. Chinese Patent CN 117068049 A invented a low-cost in-vehicle blind spot monitoring method based on the camera of a mobile device, which uses the camera of an in-vehicle mobile device (such as a mobile phone or a tablet) to cover the blind spot in the cabin and generates control instructions through image processing. Chinese Patent CN 117392833 A combines the camera and cloud data to obtain the speed and position information of obstacles in turning or extreme weather scenarios, predicts the collision time and outputs a warning, but there is only a single environmental perception solution. Chinese Patent CN 118907282 A discloses a blind spot monitoring system with multiple prompts, which integrates a monitoring module, a control module and a prompting module to improve safety through dual acoustic and optical warnings. Summarizing the existing blind spot monitoring methods, it is not difficult to find that there are still some problems in the existing methods: (1) Relying on a single sensor: Most systems only use radar or cameras, resulting in incomplete scene coverage. (2) Poor response to dynamic environments and insufficient environmental adaptability: Traditional blind spot monitoring mostly relies on static vehicle structures (such as A-pillars, rearview mirrors) or a single sensor (such as millimeter-wave radar), lacking comprehensive consideration of dynamic factors (vehicle speed, steering angle, driver's line of sight), and unable to dynamically adjust the monitoring range. (3) Lack of a human-machine collaboration mechanism and ignoring driver behavior: Existing technologies rarely integrate the impact of personalized parameters such as the driver's driving state, such as head position, line of sight direction, and dynamic vision, on the blind spot, and cannot meet the blind spot accuracy requirements of future intelligent assisted driving systems. Summary of the Invention

[0006] In view of the above technical problems and deficiencies, the purpose of the present invention is to provide a blind spot monitoring method based on multi-factor dynamic changes. This method realizes real-time dynamic modeling of the blind spot range by integrating vehicle motion parameters, environmental perception data, and driver behavior characteristics, and solves the accident problems caused by blind spot changes during vehicle driving.

[0007] To achieve the above purpose, the present invention adopts the following technical solutions:

[0008] A blind spot monitoring method based on multi-factor dynamic changes, the method includes the following steps:

[0009] Step 1. The data acquisition module collects the driver's head state, vehicle motion data, and changes in the surrounding environment, and sends them to the data processing and fusion module;

[0010] Step 2. The data processing and fusion module preprocesses the raw data obtained from the data acquisition module and performs spatio-temporal synchronization to achieve data alignment, and at the same time fuses multi-source data;

[0011] Step 3. Divide the main blind areas outside the vehicle into the front blind area, the rear blind area, the A-pillar blind area, the B-pillar blind area, and the C-pillar blind area, and calculate the basic blind area angles of each area based on the dimensions of the fixed structure of the vehicle and its distance from the driver; among them, the A-pillar blind area includes the left A-pillar blind area and the right A-pillar blind area, the B-pillar blind area includes the left B-pillar blind area and the right B-pillar blind area, and the C-pillar blind area includes the left C-pillar blind area and the right C-pillar blind area;

[0012] Step 4. Establish independent dynamic models for each blind area and a multi-blind area dynamic linkage model, and calculate the dynamic angle range of the blind area in each area based on the multi-blind area dynamic linkage model; among them, the independent dynamic model for each blind area is:

[0013] ;

[0014] Among them, represents the dynamic blind area angle of different areas of the vehicle at time t, i takes values of A left, A right, B left, B right, C left, C right, front, and rear, corresponding to the left A-pillar, the right A-pillar, the left B-pillar, the right B-pillar, the left C-pillar, the right C-pillar, the front of the vehicle, and the rear of the vehicle respectively; represents the basic blind area angle of different areas of the vehicle; represents the dynamic correction coefficient related to the angle increment of different areas of the vehicle; Weather compensation coefficient; is the field of view attenuation angle; is the blind area angle increment; represents the driver's head steering angle at time t; represents the driver's line of sight transfer information at time t;

[0015] The multi-blind area dynamic linkage model is:

[0016] ;

[0017] Among them: i represents the area of the vehicle blind area that changes under the influence of multiple factors; j represents the blind area that is affected by the change of the remaining blind area; is the coupling coefficient; represents the dynamic angle range of blind area j under the influence of the remaining blind areas at time t; the value ranges of i and j are A left, A right, B left, B right, C left, C right, front, and rear, corresponding to the left A-pillar, the right A-pillar, the left B-pillar, the right B-pillar, the left C-pillar, the right C-pillar, the front of the vehicle, and the rear of the vehicle respectively.

[0018] As a preference of the present invention, the adaptive adjustment rule of the coupling coefficient is:

[0019] ;

[0020] Wherein: is the basic coupling strength, is the vehicle speed sensitivity coefficient, v is the vehicle driving speed; is the steering angle attenuation coefficient, is the vehicle steering angle.

[0021] As a preference of the present invention, the basic blind area angle of the A-pillar blind area has the following geometric calculation formula:

[0022] ;

[0023] Wherein, refers to the average width of the A-pillar, refers to the horizontal distance from the A-pillar to the driver's eyes; wherein, the A-pillar includes the left A-pillar and the right A-pillar, the left A-pillar corresponds to the left blind area of the A-pillar, and the right A-pillar corresponds to the right blind area of the A-pillar;

[0024] The basic blind area angle of the B-pillar blind area has the following geometric calculation formula:

[0025] ;

[0026] Wherein, refers to the average width of the B-pillar, refers to the horizontal distance from the B-pillar to the driver's eyes; wherein, the B-pillar includes the left B-pillar and the right B-pillar, the left B-pillar corresponds to the left blind area of the B-pillar, and the right B-pillar corresponds to the right blind area of the B-pillar;

[0027] The basic blind area angle of the C-pillar blind area has the following geometric calculation formula:

[0028] ;

[0029] Wherein, refers to the lateral extension width of the C-pillar, refers to the horizontal distance from the C-pillar to the driver; wherein, the C-pillar includes the left C-pillar and the right C-pillar, the left C-pillar corresponds to the left blind area of the C-pillar, and the right C-pillar corresponds to the right blind area of the C-pillar;

[0030] The basic blind area angle of the front vehicle blind area has the following geometric calculation formula:

[0031] ;

[0032] Wherein, refers to the hood height, refers to the distance from the vehicle head to the driver;

[0033] The basic blind zone angle of the rear blind zone of the vehicle The geometric calculation formula is:

[0034] ;

[0035] Among them, refers to the height of the lower edge of the rear window, refers to the distance from the rear of the vehicle to the driver.

[0036] As a preference of the present invention, the expression formula of the driver's line-of-sight transfer information is:

[0037] ;

[0038] Among them, (m, n) is the original fixation point, ( ) is the pupil coordinate;

[0039] The expression formula of the blind zone angle increment is:

[0040] ;

[0041] The expression formula of the dynamic correction coefficient related to the angle increment in different regions of the vehicle is:

[0042] ;

[0043] Among them, is the reference distance monitored by the blind zone i, is the vehicle steering angle, and B is the vehicle wheelbase;

[0044] The expression formula of the visual field attenuation angle is:

[0045] ;

[0046] Among them, is the visual field attenuation angle; is the visual field angle under static conditions; is the attenuation coefficient; DAV is the dynamic visual acuity.

[0047] As a preference of the present invention, when calculating the dynamic angle range of the blind zones in each region, an online learning strategy is added to optimize the coupling parameters based on the extended Kalman filter algorithm.

[0048] As a further preference of the present invention, the relationship between the dynamic visual acuity and the speed is:

[0049] ;

[0050] Among them, is the static visual acuity, v is the relative motion speed of the vehicle, kis the attenuation coefficient.

[0051] The present invention also provides a vehicle control method, and the method includes the following steps:

[0052] Step A. Calculate the dynamic angle range of each area blind spot based on the above-mentioned multi-factor dynamic change blind spot monitoring method.

[0053] Step B. According to the data transmitted by the data acquisition module and based on the dynamic angle range calculated in Step A, determine the linkage risk index of each blind spot, and sum the weighted linkage risk indices of each blind spot to obtain the global comprehensive risk index.

[0054] Step C. The risk assessment and grading module determines the risk level based on the global comprehensive risk index, grades according to its risk level, and controls the vehicle at the same time.

[0055] As a preference of the present invention, the linkage risk index of each blind spot The calculation method is:

[0056] ;

[0057] Wherein: is the dynamic angle of blind spot j, is the reference distance monitored by blind spot j; is the traffic flow density of blind spot j; is the driver's reaction time of blind spot j; is the sensor coverage rate of blind spot j;

[0058] The calculation method of the global comprehensive risk index is:

[0059] ;

[0060] Wherein, is the weight of blind spot j.

[0061] The present invention also provides a vehicle control system, and the control system is used to implement the above-mentioned vehicle control method, and includes a data acquisition module, a data processing and fusion module, a blind spot independent dynamic model, a multi-blind spot dynamic linkage model, a risk index calculation module, a risk assessment and grading module, and a control module;

[0062] Wherein, the data acquisition module is used to collect the driver's head state, the vehicle's own motion data and the changes in the surrounding environment, send it to the data processing and fusion module, and store it in the database at the same time;

[0063] The data processing and fusion module is used to perform data preprocessing and spatio-temporal synchronization on the original data, realize data alignment, and fuse multi-source data at the same time;

[0064] The blind - area independent dynamic model is used to calculate the dynamic blind areas of different regions of the vehicle;

[0065] The multi - blind - area dynamic linkage model is used to calculate the dynamic blind areas of each region under the mutual influence of different regions of the vehicle;

[0066] The risk - index calculation model is used to calculate the linkage risk index of each blind area, and the weighted sum of the linkage risk indexes of each blind area is obtained to get the global comprehensive risk index;

[0067] The risk assessment and grading module is used to determine the risk level according to the global comprehensive risk index and conduct grade division according to the level;

[0068] The control module is used to give early warnings and control the vehicle according to the risk level.

[0069] As a preference of the present invention, the data acquisition module includes a vehicle - state perception module, an environment - perception module, and a driver - behavior adaptation module; wherein, the vehicle - state perception module obtains vehicle electronic control unit data through the CAN bus and real - time collects vehicle kinematic parameters; the environment - perception module real - time collects vehicle surrounding environment parameters through multi - source sensor fusion, and the multi - source sensors include millimeter - wave radar, lidar, and binocular cameras; the driver - behavior adaptation module real - time monitors the physiological and behavioral characteristics of the driver, and the driver - behavior adaptation module includes an eye tracker and a head - pose detection unit.

[0070] Advantages and beneficial effects of the present invention:

[0071] (1) The present invention provides a multi - factor dynamic - change blind - area monitoring method. By integrating vehicle motion parameters, environment - perception data, and driver - behavior characteristics, it realizes real - time dynamic modeling and accurate early warning of the blind - area range. This method real - time obtains parameters such as vehicle speed, steering angle, and acceleration through the CAN bus, and calculates the dynamic expansion range of the inner - wheel difference blind area in combination with the wheelbase; it detects the type, position, and motion state of obstacles in the blind area through multi - source sensor fusion (including millimeter - wave radar, binocular cameras, and lidar); it monitors the driver's line - of - sight direction, head - deflection angle, and dynamic visual acuity using an eye tracker and a head - tracking device, and adjusts the early - warning threshold personalized; it integrates and weights and analyzes the above - mentioned parameters through machine learning, real - time calculates the blind - area range, and dynamically generates a blind - area heat map to reduce accidents caused by special situations in the blind area for the driver.

[0072] (2) The present invention can effectively monitor the dynamic blind areas during vehicle driving in real time and provide hierarchical early warnings, which to a certain extent solves the accident problems caused by the change of blind areas during vehicle driving and has a positive impact on autonomous driving.

[0073] (3) The multi-factor dynamic change blind spot monitoring method provided by the present invention can improve the safety in automobile driving, avoid major traffic accidents and cause tragic consequences. By real-time monitoring the vehicle body state, the changes in the surrounding environment and the driver's head characteristics, the system can timely respond to potential conflicts and improve services, with advantages such as real-time, accuracy, personalized service and data-driven decision-making, and is expected to bring great changes and development to the autonomous driving industry.

[0074] (4) When determining the blind spot, the present invention considers the influence of the driver's driving state (such as personalized parameters like head position, line of sight direction and dynamic vision) on the blind spot, comprehensively considers and models the dynamic factors (vehicle speed, steering angle, driver's line of sight), can dynamically adjust the blind spot monitoring range, has strong environmental adaptability, and meets the blind spot accuracy requirements of future intelligent assisted driving systems.

[0075] (5) The linkage model designed when determining the blind spot of the present invention solves the fundamental problem that traditional static models cannot adapt to changes in vehicle speed, steering and driver behavior through multi-blind spot dynamic coupling and inner wheel difference expansion calculation, and significantly improves the warning accuracy in complex scenarios (turning, high speed, rain and fog).

[0076] (6) The present invention enables the system to have self-adaptability by adding an online learning mechanism, can continuously improve the model according to actual driving feedback, reduce the false alarm rate and enhance the generalization ability across vehicle models and environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Through the following description with reference to the accompanying drawings, and with a more comprehensive understanding of the present invention, other objects and results of the present invention will become more obvious and easier to understand. In the drawings:

[0078] Figure 1 is a schematic diagram of the blind spot of the vehicle of the present invention;

[0079] Figure 2 is a flowchart of the multi-factor dynamic change blind spot monitoring method provided by the present invention;

[0080] Figure 3 is a flowchart of the vehicle control method provided by the present invention;

[0081] Figure 4 is a structural block diagram of the vehicle control system provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0082] To enable those skilled in the art to better understand the technical solutions and advantages of the present invention, the present application will be described in detail below with reference to the accompanying drawings, but is not used to limit the protection scope of the present invention.

[0083] Embodiment 1:

[0084] AsFigure 1 , Figure 2 As shown in Figure 2 , this embodiment provides a multi-factor dynamic change blind spot monitoring method, and the method includes the following steps:

[0085] Step 1. The data acquisition module collects the driver's head state, vehicle motion data, and changes in the surrounding environment, and sends them to the data processing and fusion module;

[0086] Step 2. The data processing and fusion module performs data preprocessing and spatio-temporal synchronization on the raw data obtained from the data acquisition module to achieve data alignment, and at the same time fuses multi-source data;

[0087] Step 3. Divide the main blind spots outside the vehicle into a front blind spot, a rear blind spot (tail blind spot), an A-pillar blind spot, a B-pillar blind spot, and a C-pillar blind spot, and calculate the basic blind spot angles of each area based on the size of the vehicle's fixed structure and its distance from the driver; among them, the A-pillar blind spot includes an A-pillar left blind spot and an A-pillar right blind spot, the B-pillar blind spot includes a B-pillar left blind spot and a B-pillar right blind spot, and the C-pillar blind spot includes a C-pillar left blind spot and a C-pillar right blind spot;

[0088] Step 4. Establish independent dynamic models for each blind spot and a multi-blind spot dynamic linkage model, and calculate the dynamic angle range of each area's blind spot based on the multi-blind spot dynamic linkage model; among them, the independent dynamic model for each blind spot is:

[0089] ;

[0090] Where represents the dynamic blind spot angle of different areas of the vehicle at time t, i takes values of A left, A right, B left, B right, C left, C right, front, and rear, corresponding to the A-pillar left, A-pillar right, B-pillar left, B-pillar right, C-pillar left, C-pillar right, front of the vehicle, and rear of the vehicle respectively; represents the basic blind spot angle of different areas of the vehicle; represents the dynamic correction coefficient related to the angle increment of different areas of the vehicle; weather compensation coefficient; is the visual field attenuation angle; is the blind spot angle increment; represents the driver's head steering angle at time t; represents the driver's line-of-sight transfer information at time t;

[0091] The multi-blind spot dynamic linkage model is:

[0092] ;

[0093] Where: i represents the area of the vehicle's blind spot that changes under the influence of multiple factors; j represents the blind spot area affected by the change in the remaining blind spot areas; is the coupling coefficient; represents the dynamic angle range of blind area j under the influence of other blind areas at time t (the value ranges of i and j are A left, A right, B left, B right, C left, C right, front, and rear, corresponding to the left A-pillar, right A-pillar, left B-pillar, right B-pillar, left C-pillar, right C-pillar, front of the vehicle, and rear of the vehicle respectively).

[0094] In this embodiment, the adaptive adjustment rule of the coupling coefficient is:

[0095] ;

[0096] Among them: is the basic coupling strength, which is obtained by fitting multiple experimental data. For details of the data, see Appendix Figure 1 and Table 1; is the vehicle speed sensitivity coefficient, with a value of 0.05; v is the vehicle driving speed; is the steering angle attenuation coefficient, with a value of 0.1; is the vehicle steering angle.

[0097] Table 1: Basic coupling strength coefficient Experimental calibration value table

[0098]

[0099] In this embodiment, the data acquisition module is used to collect the driver's head state, the vehicle's own motion data, and the changes in the surrounding environment, send them to the data processing and fusion module, and store them in the database at the same time; the data acquisition module includes a vehicle state perception module, an environment perception module, and a driver behavior adaptation module;

[0100] Among them, the vehicle state perception module obtains vehicle ECU (electronic control unit) data through the CAN bus and real-time collects vehicle kinematic parameters. The vehicle kinematic parameters include vehicle speed, steering angle, acceleration, wheelbase, etc.; the vehicle kinematic parameters provide parameters for calculating the blind area caused by the reduction of dynamic visual acuity and the inner wheel difference expansion blind area under high-speed driving conditions.

[0101] The environment perception module real-time collects vehicle surrounding environment parameters, including weather conditions, obstacle conditions, etc.;

[0102] Specifically, the environment perception module detects the type, position, and motion state of obstacles in the blind area through multi-source sensor fusion; the multi-source sensors include millimeter-wave radar (77 GHz), optical radar (LiDAR), and binocular cameras; among them, the millimeter-wave radar is responsible for detecting the obstacle distance (0.1 - 150 meters), relative speed (±200 km / h), and azimuth angle (±60°), and is applicable to rainy, foggy weather and night scenes;

[0103] LiDAR: Generates high-precision 3D point clouds (300,000 points per second, accuracy ±3 cm), used to build a dense environment model and identify low-reflectivity targets (such as black cones).

[0104] Stereo camera: RGB image acquisition (1920×1080 resolution) + disparity ranging (accuracy ±5 cm), supports object classification (YOLOv5, mAP 85%) and lane line detection.

[0105] The driver behavior adaptation module monitors the driver's physiological and behavioral characteristics in real time, including the driver's line of sight direction, head deflection angle, and dynamic visual acuity; the driver behavior adaptation module includes an eye tracker, a head pose detection unit, and a calibration assist device; among them, the eye tracker uses Tobii Pro Glasses 3, with a sampling rate of 120 Hz and an accuracy of ±0.5°, supports the measurement of dynamic visual acuity (DVA), and simultaneously captures the position of the driver's fixation point ( ), the position of the pupil ( ), and the change in diameter, and evaluates the blind spot range;

[0106] The head pose detection unit includes a 9-axis IMU integrated in the seat headrest and a wide-angle infrared camera installed in the vehicle. The 9-axis IMU is used to detect the head deflection angle (accuracy ±1°) and the nodding / shaking frequency; the wide-angle infrared camera is used to assist in calibrating the head position and compensating for the cumulative error of the IMU; the calibration assist device includes an LED calibration board, and the LED calibration board is installed at positions such as the vehicle's A-pillar and dashboard, provides reference points with known spatial coordinates, and is responsible for regularly calibrating the coordinate system alignment error between the eye tracker and the head pose detection unit.

[0107] The data processing and fusion module is used to process and optimize the data collected by each sensor and fuse multi-sensor parameters.

[0108] In this embodiment, the vehicle body state data (vehicle's own motion data), driver data (driver's head state), and surrounding environment state data received by the data acquisition module are preprocessed and synchronized in time and space. Since the data formats, sampling frequencies, coordinate systems, and timestamps of different sensors may vary, data alignment must be achieved through standardization processing to provide consistent input for subsequent fusion algorithms.

[0109] Specifically, for the data obtained through the CAN bus, the acceleration and angular velocity data output by the IMU (Inertial Measurement Unit) are denoised through Kalman filtering (KF) to reduce the error caused by vehicle body vibration;

[0110] i. Define the state variables: The input states to be estimated are the vehicle speed v and the steering angle , the state vector is:

[0111] ;

[0112] where k represents the discrete time step, corresponding to the sampling moment of the CAN bus, and represent the vehicle speed and steering angle at discrete time step k.

[0113] ii. Establish the process model:

[0114] ;

[0115] where and represent the previous and current states of a unit discrete time step, is the process noise, following a Gaussian distribution , is the state transition matrix, is the process noise covariance matrix;

[0116] iii. Establish the observation model. The CAN bus directly measures the vehicle speed and steering angle, and the observation equation and the observation matrix are:

[0117] , ;

[0118] where the observation noise , following a Gaussian distribution , represents the sensor noise; is the observation noise covariance matrix.

[0119] iv. Design of the noise covariance matrix:

[0120] ;

[0121] : Uncertainty of the vehicle speed model; : Measurement noise variance of the vehicle speed sensor.

[0122] : Uncertainty of the steering angle model; : Measurement noise variance of the steering angle sensor.

[0123] v. Implementation steps of the Kalman filter:

[0124] (1) Initialization (initial state and error covariance matrix ​Estimation):

[0125] ;

[0126] Among them, and represent the initial vehicle speed and steering angle of the vehicle; and represent the error covariance between the initial vehicle speed and the estimated value of the initial steering angle;

[0127] (2) Predict the state and error covariance matrix based on the model:

[0128] ;

[0129] ;

[0130] Among them, represents predicting the prior state at time k based on the optimal estimated state at time k - 1 ; represents predicting the covariance of the prior state at time k based on the covariance of the optimal estimated state at time k - 1 ; represents the transpose of the state transition matrix F;

[0131] (3) Fuse the observed data:

[0132] Calculate the Kalman gain:

[0133] ;

[0134] Among them, is the Kalman gain matrix, is the transpose matrix of the observation matrix H;

[0135] State update:

[0136] ;

[0137] Among them, is the posterior state estimate value at time k;

[0138] Covariance update:

[0139] ;

[0140] Among them, is the posterior state error covariance matrix at time k;

[0141] vi. Parameter tuning and implementation details:

[0142] Noise covariance adjustment:

[0143] If the filtering result is too dependent on the model (with obvious lag), increase N or decrease Q;

[0144] If the filtering result follows the observation too much (with insufficient noise suppression), decrease N or increase Q.

[0145] For the data obtained by the millimeter-wave radar (77 GHz), the following preprocessing methods are adopted:

[0146] i. Noise filtering:

[0147] Set the SNR threshold and remove the invalid points with SNR < 10 dB;

[0148] Filter out the targets with speeds exceeding the reasonable range (such as v > 180 km / h);

[0149] ii. Use the DBSCAN algorithm for target clustering:

[0150] iii. Coordinate transformation: Convert Cartesian coordinates to polar coordinates.

[0151] For the data obtained by LiDAR, the following preprocessing methods are adopted:

[0152] i. Input: Original 3D point cloud (300,000 points per second, including coordinates (x, y, z) and reflection intensity);

[0153] ii. Ground point removal: Use the RANSAC algorithm to fit the ground plane equation , and remove the points;

[0154] iii. Outlier filtering: Statistical filtering, remove the points whose distance exceeds the mean by ;

[0155] iv. Target segmentation: Set the threshold of 0.3 m through Euclidean clustering to separate the obstacle point cloud clusters:

[0156] (1) Build a search structure: Store the point cloud in a KD-Tree to accelerate the nearest neighbor query;

[0157] (2) Core clustering process:

[0158] 2.1 Initialize an empty clustering list and an unprocessed point queue;

[0159] 2.2 Randomly select a seed point and create a new cluster;

[0160] 2.3 Search for all neighborhood points within the radius r = 0.3 m of the seed point;

[0161] 2.4 Add the neighborhood points to the current cluster and push them into the queue;

[0162] 2.5 Recursively process the points in the queue until no new points are added;

[0163] 2.6 When the number of clustered points ≥ threshold (e.g., 50 points), save it as a valid obstacle;

[0164] (3) Post - processing:

[0165] 3.1 Filter out too - small clusters (e.g., number of points < 50, may be noise);

[0166] 3.2 Output the point cloud clusters and bounding boxes of each obstacle.

[0167] v. Coordinate transformation: Convert the LiDAR coordinate system to the vehicle coordinate system (apply the extrinsic parameter matrix ).

[0168] For the data obtained by the binocular camera, pre - process it in the following way:

[0169] i. Image enhancement: Enhance the details of low - light areas through CLAHE;

[0170] ii. Distortion correction: Use the intrinsic parameter matrix K and distortion coefficients D to correct lens distortion;

[0171] iii. Object detection: Output the object category (pedestrian / vehicle) and 2D bounding box through the YOLOv5 model;

[0172] iv. Coordinate mapping: Project the 2D bounding box to the vehicle coordinate system using inverse perspective mapping (IPM).

[0173] For the data obtained by the eye tracker, pre - process it in the following way:

[0174] i. Input: Original fixation points (m, n);

[0175] ii. Noise filtering: Use the median filtering method, set the window size to 5 frames, and remove the coordinate jumps caused by blinking;

[0176] iii. Gaze vector calculation: Unit vector transformation:

[0177] ;

[0178] where, ( ) are the pupil coordinates, represents the driver's gaze shift information at time t;

[0179] iv. Coordinate system transformation: Convert the eye tracker coordinate system to the vehicle coordinate system (apply the extrinsic parameter matrix ).

[0180] The data obtained by the head pose detection unit is preprocessed in the following manner:

[0181] i. Input: 9-axis IMU raw data (acceleration , angular velocity , sampling rate 100Hz);

[0182] ii. Attitude solution: Analyze using the quaternion update algorithm;

[0183] iii. Coordinate system conversion: Convert the head coordinate system to the vehicle coordinate system (apply the extrinsic parameter matrix ).

[0184] In this embodiment, the coordinate systems of the millimeter-wave radar, LiDAR, binocular camera, eye tracker, and 9-axis IMU are unified to the vehicle coordinate system through the extrinsic parameter calibration method. A checkerboard calibration board is used to calculate the extrinsic parameter matrix between sensors through feature point matching.

[0185] In addition, in this embodiment, data spatio-temporal synchronization aims to align the timestamps and coordinate systems of multi-sensor data to ensure fusion accuracy. Specifically, time is synchronized at both the hardware and software levels. The hardware provides the global UTC time (accuracy ±1ms) through the GPS timestamp, and uses the PXIe-6674T timing module to achieve μs-level time synchronization with a delay <1ms. The software uses the interpolation alignment method to perform linear interpolation on low-frequency data (LiDAR 10Hz) to generate sampling points synchronized with high-frequency data (eye tracker 120Hz).

[0186] It should be noted that the above preprocessing method provided in this embodiment is only for illustrative purposes. Those skilled in the art can also use other methods to preprocess the acquired raw data. The preprocessing method of the raw data and the spatio-temporal synchronization method are not the key points of the present invention.

[0187] In this embodiment, the main blind areas outside the vehicle are divided into the front blind area, the rear blind area, the A-pillar blind areas (left and right), the B-pillar blind areas (left and right), and the C-pillar blind areas (left and right); multi-sensor data fusion is the core link of vehicle blind area monitoring. Combining the advantages of radar (strong anti-interference), LiDAR (high-precision 3D modeling), and camera (target classification), it covers the entire scene. The overall blind area size is obtained through multi-blind area coupling dynamic linkage calculation, and it is predicted and corrected through the extended Kalman filter; the combination of multi-blind area coupling linkage calculation and the extended Kalman filter (EKF) optimizes the blind area calculation, improving the accuracy, robustness, and scene adaptability of the blind area calculation.

[0188] Coordinate system selection: A polar coordinate system is established with the vehicle's center of mass as the origin, the horizontal direction as the polar axis, and the vertical direction as the height axis, covering a 360° range.

[0189] In this embodiment, the basic blind area angle is the static occlusion angle determined by the vehicle's fixed structure (such as the front and rear of the vehicle, A / B / C pillars, etc.). Its geometric model method is calculated through the dimensions of the vehicle's fixed structure and its horizontal distance from the driver's eyes. The specific calculation method is as follows:

[0190] i. A-pillar blind area: The A-pillar is the vertical pillar between the windshield and the left and right front doors, located between the engine compartment and the passenger compartment. Due to the existence of the A-pillar, when the driver turns left or right, part of their line of sight will be blocked, and they cannot see the area behind the A-pillar, thus forming the A-pillar blind area; the basic blind area angle of the A-pillar blind area The geometric calculation formula is:

[0191] ;

[0192] where, refers to the average width of the A-pillar, which is approximately between 0.1 - 0.15 m, refers to the horizontal distance from the A-pillar to the driver's eyes; among them, the A-pillar includes the left A-pillar and the right A-pillar. The left A-pillar corresponds to the left blind area of the A-pillar, and the right A-pillar corresponds to the right blind area of the A-pillar. The reference range of the left blind area of the A-pillar is 6° - 12° (for sedans), and the reference range of the right blind area of the A-pillar is 5° - 8° (for sedans).

[0193] ii. B-pillar blind area: The B-pillar is located between the front and rear seats of the cockpit and is the longitudinal pillar between the two side doors. When the vehicle is in motion, especially when changing lanes or turning, the B-pillar will block the driver's line of sight to oncoming vehicles from the opposite side and rear, forming the B-pillar blind area. The basic blind area angle of the B-pillar blind area The geometric calculation formula is:

[0194] ;

[0195] where, refers to the average width of the B-pillar, which is approximately between 0.1 - 0.12 m, refers to the horizontal distance from the B-pillar to the driver's eyes; among them, the B-pillar includes the left B-pillar and the right B-pillar. The left B-pillar corresponds to the left blind area of the B-pillar, and the right B-pillar corresponds to the right blind area of the B-pillar; the reference range of the left blind area of the B-pillar is 5° - 8° (for sedans), and the reference range of the right blind area of the B-pillar is 3° - 5° (for sedans).

[0196] iii. C-pillar blind area: The C-pillar is located on both sides of the rear seat headrest and is the vertical pillar between the rear windshield and the rear door. The existence of the C-pillar will block the driver's rear line of sight to a certain extent, forming the C-pillar blind area. The basic blind area angle of the C-pillar blind area The geometric calculation formula is:

[0197] ;

[0198] Among them, refers to the lateral extension width of the C-pillar, which is approximately between 0.2 - 0.3 m, refers to the horizontal distance from the C-pillar to the driver; among them, the C-pillar includes the left C-pillar and the right C-pillar, the left C-pillar corresponds to the left blind area of the C-pillar, and the right C-pillar corresponds to the right blind area of the C-pillar; the left blind area of the C-pillar has a reference range of 12° - 20° (for sedans), and the right blind area of the C-pillar has a reference range of 10° - 15° (for sedans), which is significant for MPV / SUV models.

[0199] iv. Front blind area: The front blind area of a vehicle refers to the area directly below the front of the vehicle that the driver cannot directly see while sitting in the driver's seat. The geometric calculation formula for the basic blind area angle of the front blind area is:

[0200] ;

[0201] Among them, refers to the height of the engine hood, which is approximately between 0.8 - 1.2 m, refers to the distance from the front of the vehicle to the driver. The reference range is 15° - 25° (for the longitudinal blind area of 3 - 4 meters at the front of an SUV) and 10° - 18° (for sedans).

[0202] v. Rear blind area: The rear blind area of a vehicle refers to the area directly below the rear of the vehicle that the driver cannot see through the rearview mirror or by direct observation. The geometric calculation formula for the basic blind area angle of the rear blind area is:

[0203] ;

[0204] Among them, refers to the height of the lower edge of the rear window, which is approximately between 0.5 - 0.8 m, refers to the distance from the rear of the vehicle to the driver. The reference range is 20° - 30° (covering the area within 5 meters at the rear of the vehicle).

[0205] In this embodiment, the calculation of the inner wheel difference blind area expansion plays a key role in the calculation of vehicle blind areas. Its core lies in dynamically correcting the change in the blind area range caused by vehicle turning, thereby improving the accuracy and safety of blind area monitoring. By analyzing vehicle steering parameters (such as steering angle, wheelbase) in real time, the blind area boundary is dynamically adjusted to solve the problem that the traditional static model cannot reflect the expansion of the blind area during turning.

[0206] Inner wheel difference length The calculation formula for is:

[0207] ;

[0208] Wherein, is the vehicle steering angle, and B is the wheelbase of the vehicle.

[0209] Blind spot i angle increment The calculation formula is:

[0210] ;

[0211] Wherein, is the reference distance for blind spot i monitoring. When , the calculation formula for the blind spot angle increment can be simplified to:

[0212] ;

[0213] Substitute the relationship between and to derive the steering angle correction coefficient :

[0214] Simplify through trigonometric identities :

[0215] ;

[0216] Combined with the reference distance for blind spot monitoring, the blind spot i angle increment can be approximated as:

[0217] ;

[0218] Thus, the correction coefficient is obtained:

[0219] ;

[0220] The greater the length of the inner wheel difference, the farther the rear wheel trajectory deviates from the front wheel during steering, and the wider the blind spot coverage range.

[0221] In this embodiment, the reference distance r for each blind spot monitoring can be determined according to the model of the actually used sensor. Table 2 shows the recommended values of the reference distance r in different scenarios, and these recommended values are set with reference to the requirements of international standards such as ISO 17387 for blind spot monitoring systems.

[0222] Table 2 Recommended values of the reference distance r for each blind spot monitoring in different scenarios

[0223]

[0224] In this embodiment, Dynamic Visual Acuity (DVA) refers to the ability to clearly identify the details of an object in a moving state (such as when oneself or the target is moving), usually expressed by the minimum distinguishable visual angle (angular minute) or the visual acuity chart level. It mainly affects the ability to process information simultaneously in a dynamic environment. The geometric relationship between Dynamic Visual Acuity (DAV) and the visual field range is analyzed as follows:

[0225] Speed and dynamic visual acuity attenuation curve:

[0226] ;

[0227] Where: is the static visual acuity (refer to the international standard visual acuity chart), v is the relative motion speed of the vehicle (km / h), k is the attenuation coefficient, and the reference value is 0.02.

[0228] Relationship formula between the dynamic visual acuity attenuation curve and the visual field attenuation angle:

[0229] ;

[0230] Where: is the visual field attenuation angle; is the visual field angle under static conditions; is the attenuation coefficient, with a value of 0.9 (calibration value based on the dynamic vision experiment of ISO 15007-2 standard), indicating the compression ratio of the visual field due to the decrease in DVA. Table 3 shows the relationship between the visual field attenuation amount of the driver and the driving speed, taking the basic visual field angle of the driver as 180° and the static visual acuity as 1.0 as an example.

[0231] Table 3 Relationship between the visual field attenuation amount of the driver and the driving speed

[0232] Driving speed (km / h) Vision attenuation (°) 10 29.19 20 53.11 30 72.73 40 88.81 50 102.00 60 112.81 70 121.67 80 128.94 90 134.89 100 139.78 110 143.78 120 147.06 130 149.75

[0233] In this embodiment, the conversion formula for the driver's line-of-sight vector is:

[0234] ;

[0235] Where: ([[]]END]] ) are the pupil coordinates, ([[]]END]] m , n ) are the positions of the fixation points in the direction of the eye gaze, represents the line-of-sight transfer information of the driver at time t;

[0236] The steering angle of the driver's head at time t is collected and processed by the head pose detection unit.

[0237] In this embodiment, the design methods of each blind area independent dynamic model and multi-blind area dynamic linkage model are as follows:

[0238] Blind area basic calculation mathematical model:

[0239] ;

[0240] Where: represents the basic blind area angle; represents the reference value of the dynamic correction coefficient, where , , , .

[0241] Environmental factor compensation, introducing the weather compensation coefficient (1.0 for sunny days, 1.2 for rain and fog, dimensionless), to correct the blind area angle:

[0242] ;

[0243] Where, is the corrected blind area angle, represents the blind area angle before correction;

[0244] Each blind area independent dynamic model:

[0245] ;

[0246] Where: represents the dynamic blind area angle of different areas of the vehicle at time t, i takes values of A left, A right, B left, B right, C left, C right, front, and rear, corresponding to the left A-pillar, right A-pillar, left B-pillar, right B-pillar, left C-pillar, right C-pillar, front of the vehicle, and rear of the vehicle respectively; represents the basic blind area angle of different areas of the vehicle; represents the dynamic correction coefficient related to the attenuation of the field of view of different areas of the vehicle due to speed, represents the dynamic correction coefficient related to the angle increment of different areas of the vehicle, represents the dynamic correction coefficient related to the steering angle of the driver's head in different areas of the vehicle, represents the dynamic correction coefficient related to the line of sight direction of the driver in different areas of the vehicle; Weather compensation coefficient;

[0247] Adjust the parameters according to the actual situation:

[0248] For the A-pillar blind area, the vehicle speed v affects the forward shift of the fixation point, and it needs to be corrected to 0.08 ° / km / h through the steering angle The influence is small, set it to 0.

[0249] For the front blind area of the vehicle, the vehicle speed v affects the forward movement of the fixation point. It is necessary to correct it to 0.15 ° / km / h. The front blind area is not sensitive to the steering angle. Set it to 0.

[0250] For the rear blind area of the vehicle, it is mainly determined by the rear window structure and is less affected by dynamic factors. It can be simplified to .

[0251] Independent dynamic models for each blind area after adjustment:

[0252] ;

[0253] Multi-blind area dynamic linkage model:

[0254] ;

[0255] Among them: i represents the area of the vehicle blind area that changes under the influence of multiple factors; j represents the blind area affected by the change of the remaining blind area areas; : coupling coefficient; represents the dynamic angle range of blind area j under the influence of the remaining blind areas at time t (the value ranges of i and j are A left, A right, B left, B right, C left, C right, front, and rear, corresponding to the left A-pillar, right A-pillar, left B-pillar, right B-pillar, left C-pillar, right C-pillar, front of the vehicle, and rear of the vehicle respectively).

[0256] Furthermore, in this embodiment, the blind area calculation is optimized by combining multi-blind area coupling linkage calculation with extended Kalman filter (EKF) to improve the accuracy, robustness, and scene adaptability of the blind area calculation:

[0257] a. State vector expansion:

[0258] ;

[0259] Among them: is the vehicle driving speed; is the vehicle steering angle; is the position of the fixation point in the direction of the eye gaze; is the head turning angle of the driver; d represents the vector dimension.

[0260] b. Establishment of the process model:

[0261] State transition equation:

[0262] ;

[0263] Among them: Represents the blind spot angles of different regions of the vehicle in the subsequent state; Represents the expansion range of the blind spot angles under different influences in the previous state; Is the visual field attenuation angle in the k-th state, Is the increment of the blind spot angle in the k-th state, Is the steering angle of the driver's head in the k-th state, Is the line-of-sight transfer information of the driver in the k-th state, Represents the process noise, which follows a Gaussian distribution , Is the process noise covariance matrix;

[0264] ;

[0265] Among them, Is the variance of each parameter;

[0266] c. Establish an observation model:

[0267] Observation equation:

[0268] ;

[0269] Among them, Is the observation noise, which follows a Gaussian distribution , representing the sensor noise, Is the observation noise covariance matrix.

[0270] d. Steps for implementing the extended Kalman filter:

[0271] (1) Prediction stage

[0272] Prediction of the extended state equation:

[0273] ;

[0274] Among them, Represents the prior state at time k predicted based on the optimal estimated state at time k - 1 ; Is the process noise; Is the non-linear state transition function;

[0275] ;

[0276] Among them, Is the control input vector, which are the acceleration, the vehicle steering angular velocity, and the head and eye steering angular velocities respectively; S is the parameter matrix , Is the time step.

[0277] At each step k, for the non-linear function and perform first-order Taylor expansion to obtain respectively At the Jacobian matrix of and at the Jacobian matrix of , ; where represents partial derivative of with respect to represents partial derivative of with respect to

[0278] Error covariance prediction:

[0279] ;

[0280] wherein represents process noise, represents the covariance of the optimal estimated state at the (k - 1)th moment to predict the covariance of the prior state at the kth moment;

[0281] (2) Update stage:

[0282] Calculate the Kalman gain:

[0283] ;

[0284] wherein represents observation noise;

[0285] State update:

[0286] ;

[0287] Covariance update:

[0288] ;

[0289] Dynamic coupling parameter optimization: Add an online learning strategy:

[0290] , ;

[0291] wherein represents the change in the coupling coefficient, represents partial derivative of with respect to = 0.01, constraint condition: .

[0292] Example 2:

[0293] As Figure 3 shown, this embodiment provides a vehicle control method, and the method includes the following steps:

[0294] Step A. Calculate the dynamic angle range of each regional blind area based on the blind area monitoring method of Embodiment 1;

[0295] Step B. According to the data transmitted by the data acquisition module and based on the dynamic angle range calculated in Step A, determine the risk index of each blind area, and perform weighted summation on the linkage risk indices of each blind area to obtain the global comprehensive risk index;

[0296] Step C. The risk assessment and grading module determines the risk level based on the global comprehensive risk index, performs grading according to its risk level, and controls the vehicle at the same time.

[0297] This embodiment adopts the calculation method of the blind area risk index of the SAE J2804 standard, and fuses it with the multi-blind area risk index model to optimize it into a partition risk index model. The specific steps are as follows:

[0298] The SAE J2804 standard defines the calculation method of the blind area risk index, and its core formula is:

[0299] ;

[0300] Among them: The blind area area refers to the blind area coverage area calculated by the model (unit: ㎡); The traffic flow density refers to the number of potential obstacles in the blind area per unit time (unit: number / second); The driver reaction time refers to the time from perceiving danger to taking action (unit: second); The sensor coverage rate refers to the proportion of the effective detection range of the sensor in the total blind area area (unit: %).

[0301] Fusion and optimization of the multi-blind area risk index model:

[0302] (1) Partition risk index calculation, independently calculate the linkage risk index for each blind area (left A-pillar, right A-pillar, left B-pillar, right B-pillar, left C-pillar, right C-pillar, front of the vehicle, rear of the vehicle) :

[0303] ;

[0304] Among them: is the dynamic angle of blind area j (calculated based on the linkage model), is the reference distance monitored by blind area j; is the traffic flow density of blind area j (number of obstacles / second); is the driver reaction time of blind area j (extended when distracted); is the sensor coverage rate of blind area j (reduced in rainy and foggy days).

[0305] (2) Comprehensive risk index. The linkage risk indices of each blind area are weighted and summed to obtain the global comprehensive risk index:

[0306] ;

[0307] Among them, is the weight of blind area j; the specific weight distribution is as follows: A-pillar (left and right) blind area: (the risk is significant during turning); front blind area of the vehicle: (high importance in low-speed scenarios); B and C-pillars (left and right), rear blind area of the vehicle: (rely on the driver's active observation).

[0308] In this embodiment, based on the comprehensive risk index the risk level is determined, classified and warned, and at the same time the vehicle is controlled, specifically as follows:

[0309] If is in the range of 1.0 - 2.0, it is considered that the risk level is level one. At this time, the LED warning light inside the vehicle is controlled to flash at a frequency of 2 Hz for acoustic and optical prompts to warn the driver;

[0310] If is in the range of 2.0 - 3.0, it is considered that the risk level is level two. At this time, the steering wheel of the vehicle is controlled to vibrate at an intensity of 5 Hz; at the same time, a voice broadcast "Danger in a certain blind area" is made;

[0311] If is in the range of > 3.0, it is considered that the risk level is level three (emergency warning). At this time, automatic braking pre-charging (reducing the braking distance by 0.5 meters) and forced lane keeping are performed. This method can effectively control the scale of conflicts in time, avoid large traffic accidents and reduce casualties. In addition, after the emergency braking is triggered, if the driver steps on the accelerator pedal > 50%, the braking intervention will be terminated immediately.

[0312] Embodiment 3:

[0313] As Figure 4 shown, this embodiment provides a vehicle control system, which includes a data acquisition module, a data processing and fusion module, a blind area independent dynamic model, a multi-blind area dynamic linkage model, a risk index calculation module, a risk assessment and classification module, and a control module;

[0314] The data acquisition module is used to collect the driver's head state, the vehicle's own motion data and the surrounding environment changes, send them to the data processing and fusion module, and store them in the database; the data acquisition module includes a vehicle state perception module, an environment perception module, and a driver behavior adaptation module;

[0315] The vehicle state perception module acquires vehicle ECU (electronic control unit) data through the CAN bus and collects vehicle kinematic parameters in real time; the environment perception module collects vehicle surrounding environment parameters in real time through multi-source sensor fusion, and the multi-source sensor includes millimeter wave radar (77GHz), optical radar (LiDAR), and binocular camera; the driver behavior adaptation module monitors the driver's physiological and behavioral characteristics in real time, and the driver behavior adaptation module includes an eye tracker and a head posture detection unit (head tracking device);

[0316] The data processing and fusion module is used to perform data preprocessing and spatiotemporal synchronization on the original data, realize data alignment, and fuse multi-source data at the same time;

[0317] The blind spot independent dynamic model is used to calculate the dynamic blind spots in different areas of the car;

[0318] The multi-blind zone dynamic linkage model is used to calculate the dynamic blind zones of each area under the mutual influence of different areas of the car;

[0319] The risk index calculation model is used to calculate the linkage risk index of each blind spot, and to obtain a global comprehensive risk index by weighted summation of the linkage risk indexes of each blind spot;

[0320] The risk assessment and grading module is used to determine the risk level based on the global comprehensive risk index and to grade the risk level by level;

[0321] The control module is used for early warning and vehicle control according to the risk level.

[0322] The present invention also provides an electronic device, comprising: one or more processors and a memory; wherein the memory is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned vehicle control method.

[0323] The present invention also provides a computer-readable medium having a computer program stored thereon, wherein the computer program implements the above-mentioned vehicle control method when executed by a processor.

[0324] Those skilled in the art can understand that all or part of the functions of the various methods / modules in the above embodiments can be implemented in a hardware manner or in a computer program manner. When all or part of the functions in the above embodiments are implemented in a computer program manner, 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 can be realized by a computer executing this program. For example, the program is stored in the memory of the device, and when the processor executes the program in the memory, the above all or part of the functions can be realized.

[0325] In addition, when all or part of the functions in the above embodiments are implemented in a computer program manner, the program can also be stored in a storage medium such as a server, another computer, magnetic disk, optical disk, flash drive or mobile hard disk, and saved to the memory of the local device by downloading or copying, or the system of the local device can be updated in version. When the processor executes the program in the memory, all or part of the functions in the above embodiments can be realized.

[0326] The above uses specific examples to illustrate the present invention, which is only for helping to understand the present invention and is not intended to limit the present invention. For those skilled in the technical field to which the present invention pertains, according to the idea of the present invention, several simple deductions, deformations or substitutions can also be made. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A blind spot monitoring method based on dynamic changes of multiple factors, characterized in that: The method comprises the following steps: Step 1. The data acquisition module collects the driver's head status, vehicle movement data and surrounding environment changes, and sends them to the data processing and fusion module; Step 2. The data processing and fusion module performs data preprocessing and spatiotemporal synchronization on the raw data obtained from the data acquisition module, realizes data alignment, and fuses multi-source data at the same time; Step 3. Divide the main blind spots outside the car into the front blind spot, the rear blind spot, the A-pillar blind spot, the B-pillar blind spot, and the C-pillar blind spot, and calculate the basic blind spot angle of each area based on the size of the vehicle fixed structure and its distance from the driver; wherein the A-pillar blind spot includes the A-pillar left blind spot and the A-pillar right blind spot, the B-pillar blind spot includes the B-pillar left blind spot and the B-pillar right blind spot, and the C-pillar blind spot includes the C-pillar left blind spot and the C-pillar right blind spot; Step 4. Establish an independent dynamic model for each blind spot and a dynamic linkage model for multiple blind spots, and calculate the dynamic angle range of each blind spot based on the dynamic linkage model for multiple blind spots; wherein, the independent dynamic model for each blind spot is: Among them, θ i (t) represents the dynamic blind spot angles of different areas of the car at time t, i takes the values ​​of A left, A right, B left, B right, C left, C right, front, and rear, corresponding to A pillar left, A pillar right, B pillar left, B pillar right, C pillar left, C pillar right, front of the car, and rear of the car respectively; θ 0,i Represents the basic blind spot angles of different areas of the car; k δ,i Represents the dynamic correction coefficient related to the angle increment of different areas of the car; k w Weather compensation coefficient; θ VD is the field of view attenuation angle; δ(t) is the blind spot angle increment; θ h (t) represents the driver's head steering angle at time t; θ e (t) represents the driver’s line of sight shift information at time t; The multi-blind zone dynamic linkage model is: Where: i represents the blind spot area of ​​the car that changes under the influence of multiple factors; j represents the blind spot area affected by the changes in other blind spots; λ i→j is the coupling coefficient; represents the dynamic angle range of blind spot j under the influence of other blind spots at time t; the value range of i, j is A left, A right, B left, B right, C left, C right, front, and rear, corresponding to A pillar left, A pillar right, B pillar left, B pillar right, C pillar left, C pillar right, front of the vehicle, and rear of the vehicle, respectively, θ j (t) represents the dynamic blind spot angle of the j blind spot under the influence of the other blind spots at time t.

2. The method for monitoring blind spots based on dynamic changes of multiple factors according to claim 1 is characterized in that: The coupling coefficient adaptive adjustment rule is: l i→j =a i→j ·tanh(β·v)·e -γ|ξ| ; Where: α i→j is the basic coupling strength, β is the vehicle speed sensitivity coefficient, v is the vehicle speed; γ is the steering angle attenuation coefficient, ξ is the vehicle steering angle.

3. The blind spot monitoring method based on multi-factor dynamic changes according to claim 1 is characterized in that: The basic blind spot angle of the A-pillar blind spot θ 0,A The geometric calculation formula is: Among them, W A Refers to the average width of the A-pillar, D A Refers to the horizontal distance from the A-pillar to the driver's eyes; the A-pillar includes the left A-pillar and the right A-pillar. The left A-pillar corresponds to the left blind spot of the A-pillar, and the right A-pillar corresponds to the right blind spot of the A-pillar. The basic blind spot angle of the B-pillar blind spot θ 0,B The geometric calculation formula is: Among them, W B Refers to the average width of the B-pillar, D B Refers to the horizontal distance from the B-pillar to the driver's eyes; the B-pillar includes the left B-pillar and the right B-pillar. The left B-pillar corresponds to the left blind spot of the B-pillar, and the right B-pillar corresponds to the right blind spot of the B-pillar. The basic blind spot angle of the C-pillar blind spot θ 0,C The geometric calculation formula is: Among them, W C Refers to the lateral extension width of the C-pillar, D C Refers to the horizontal distance from the C-pillar to the driver; the C-pillar includes the left C-pillar and the right C-pillar. The left C-pillar corresponds to the left blind spot of the C-pillar, and the right C-pillar corresponds to the right blind spot of the C-pillar. Basic blind spot angle θ of the front blind spot 0,前 The geometric calculation formula is: Among them, H 引擎 Refers to the hood height, D 车头 Refers to the distance from the front of the vehicle to the driver; The basic blind spot angle of the rear blind spot θ 0,后 The geometric calculation formula is: Among them, H 后窗 Refers to the height of the lower edge of the rear window, D 后 Refers to the distance from the rear of the vehicle to the driver.

4. The method for monitoring blind spots based on dynamic changes of multiple factors according to claim 1 is characterized in that: The expression of the driver's sight shift information is: Among them, (m, n) is the position of the gaze point in the eye gaze direction, (m0, n0) is the pupil coordinates; The expression of blind spot angle increment is: The expression of the dynamic correction coefficient related to the angle increment of different areas of the car is: Among them, r i is the reference distance for monitoring blind spot i, ξ(t) is the steering angle of the vehicle, and B is the wheelbase of the vehicle; The expression of the field of view attenuation angle is: Among them, θ VD is the field of view attenuation angle; θ s is the visual angle under static conditions; α is the attenuation coefficient; DAV is the dynamic visual acuity, V stastic For static vision.

5. The method for monitoring blind spots based on dynamic changes of multiple factors according to claim 1 is characterized in that: When calculating the dynamic angle range of the blind spots in each area, an online learning strategy is added to optimize the coupling parameters based on the extended Kalman filter algorithm.

6. The method for monitoring blind spots based on dynamic changes of multiple factors according to claim 4 is characterized in that: The relationship between dynamic visual acuity and speed is: DAV=V stastic ×e -kv ; Among them, V stastic is the static vision, v is the relative speed of the car, and k is the attenuation coefficient.

7. A vehicle control method, characterized in that: The method comprises the following steps: Step A. Calculate the dynamic angle range of the blind spot in each area based on the blind spot monitoring method based on multi-factor dynamic changes according to any one of claims 1 to 6; Step B. Determine the linkage risk index of each blind spot according to the data transmitted by the data acquisition module and based on the dynamic angle range calculated in step 1, and perform weighted summation of the linkage risk indexes of each blind spot to obtain a global comprehensive risk index; Step C. The risk assessment and classification module determines the risk level based on the global comprehensive risk index, classifies the vehicle according to its risk level, and controls the vehicle at the same time.

8. A vehicle control method according to claim 7, characterized in that: The linkage risk index for each blind spot The calculation method is: in: represents the dynamic angle range of blind zone j under the influence of other blind zones at time t, r j is the reference distance for blind spot j monitoring; D j is the traffic flow density of blind spot j; T j is the driver reaction time in blind spot j; C j is the sensor coverage of blind spot j; The global comprehensive risk index is calculated as follows: Among them, ω j is the weight of blind zone j.

9. A vehicle control system, characterized in that: The control system is used to implement the vehicle control method according to claim 7 or 8, including a data acquisition module, a data processing and fusion module, a blind spot independent dynamic model, a multi-blind spot dynamic linkage model, a risk index calculation module, a risk assessment and classification module, and a control module; The data acquisition module is used to collect the driver's head status, the vehicle's own motion data and the surrounding environment changes, send them to the data processing and fusion module, and store them in the database; The data processing and fusion module is used to perform data preprocessing and spatiotemporal synchronization on the original data, realize data alignment, and fuse multi-source data at the same time; The blind spot independent dynamic model is used to calculate the dynamic blind spots in different areas of the car; The multi-blind zone dynamic linkage model is used to calculate the dynamic blind zones of each area under the mutual influence of different areas of the car; The risk index calculation model is used to calculate the linkage risk index of each blind spot, and to obtain a global comprehensive risk index by weighted summation of the linkage risk indexes of each blind spot; The risk assessment and grading module is used to determine the risk level based on the global comprehensive risk index and to grade the risk level by level; The control module is used for early warning and vehicle control according to the risk level.

10. A vehicle control system according to claim 9, characterized in that: The data acquisition module includes a vehicle state perception module, an environment perception module, and a driver behavior adaptation module; wherein the vehicle state perception module obtains the vehicle electronic control unit data through the CAN bus and collects the vehicle kinematic parameters in real time; the environment perception module collects the vehicle surrounding environment parameters in real time through multi-source sensor fusion, and the multi-source sensor includes a millimeter wave radar, an optical radar, and a binocular camera; the driver behavior adaptation module monitors the driver's physiological and behavioral characteristics in real time, and the driver behavior adaptation module includes an eye tracker and a head posture detection unit.

Images