Intelligent headlamp adaptive control method and system based on multi-modal data fusion

Abstract

The invention relates to the technical field of automobile driving, and particularly discloses an intelligent headlamp self-adaptive control method and system based on multi-modal data fusion. The method comprises the following steps: acquiring an ambient light value outside a vehicle and a vehicle pose vector in real time; requesting a road side unit to obtain a tunnel or garage feature point set based on the ambient light value through an Internet of vehicles V2X communication protocol; based on the vehicle pose vector and the tunnel or garage feature point set, calculating a relative distance D by using an improved Euclidean distance, introducing a speed-adaptive distance compensation coefficient, and generating a corrected distance D '; the corrected distance D'is compared with a first preset early warning distance threshold K1 and a second preset early warning distance threshold K2, and if D 'is smaller than or equal to K1, a progressive lighting scheme is started; and if D 'is greater than or equal to K2, starting a light brightness attenuation scheme. The invention aims to provide a high-precision and safe mode, so that the self-adaptive control of the automobile headlamp is closer to the actual driving requirement of the automobile.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle driving, and particularly to an intelligent headlight adaptive control method and system based on multi-modal data fusion. Background Art

[0002] Currently, with the rapid development of the economy, cars have increasingly become a means of transportation for more and more people. According to the latest automotive standard specifications, automatic headlights for cars have become the most important automotive safety configuration and are a standard configuration for most cars leaving the factory.

[0003] In the prior art, most vehicles use a single data source and a fixed threshold to implement the control of turning on and off the lights in tunnel scenarios, without introducing the vehicle speed parameter, resulting in improper control timing due to low distance calculation accuracy in high-speed scenarios. At the same time, the light control is relatively simple, mostly using binary operations of "on or off", and once the light changes suddenly, it is easy to cause discomfort to the eyes of the car driver, thus there are potential safety hazards.

[0004] It should be noted that the information disclosed in the above background art section is only used to enhance the understanding of the background of the present application, and thus may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0005] Aiming at the problems of low safety and low distance calculation accuracy in the existing automotive light control, the present invention provides an intelligent headlight adaptive control method and system based on multi-modal data fusion, aiming to provide a high-precision and safe method to make the adaptive control of automotive headlights more in line with the actual driving needs of the vehicle.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An intelligent headlight adaptive control method based on multi-modal data fusion, the method comprising:

[0008] Step 1, real-time collect the ambient light value outside the vehicle and the vehicle pose vector; wherein, the vehicle GPS (Global Positioning System) position data and the IMU (Inertial Measurement Unit) angular velocity data are fused through the ESKF (Error State Kalman Filter) algorithm to output a corrected vehicle pose vector D v , the vehicle pose vector D v comprises the spatial position coordinates (x, y, z) of the vehicle and the heading angle θ of the vehicle;

[0009] Step 2, if the duration for which the ambient light value is greater than the preset brightness M does not exceed the preset time T2, a request is sent to the roadside unit via the V2X (Vehicle-to-Everything) communication protocol to obtain the tunnel or garage feature point set;

[0010] Step 3, based on the vehicle pose vector D v and the tunnel or garage feature point set, the relative distance D is calculated using an improved Euclidean distance calculation formula, and a speed-adaptive distance compensation coefficient β(v) is introduced to generate the corrected distance D ′ = D × β(v); where β(v) = 1 + tanh(0.02v), and v is the current vehicle speed;

[0011] Step 4, compare the corrected distance D ′ with the preset warning distance threshold K. The preset warning distance threshold K includes a first preset warning distance threshold K1 and a second preset warning distance threshold K2. The first preset warning distance threshold K1 is the warning distance of the vehicle from the tunnel or garage entrance, and the second preset warning distance threshold K2 is the warning distance of the vehicle from the tunnel or garage exit;

[0012] If the corrected distance D ′ is less than or equal to the first preset warning distance threshold K1, it is determined that the vehicle is about to enter the tunnel or garage, and a progressive lighting scheme is activated;

[0013] If the corrected distance D ′ is greater than or equal to the second preset warning distance threshold K2, it is determined that the vehicle has left the tunnel or garage, and a lighting brightness attenuation scheme is activated.

[0014] As a preferred solution of the present invention, the relative distance D is calculated using an improved Euclidean distance calculation formula, and the calculation formula is as follows:

[0015]

[0016] In the formula, w i is the dimension weighting coefficient of the i-th tunnel or garage feature point; n is the number of tunnel or garage feature points; α is the heading angle compensation factor; D i is the pose vector of the i-th tunnel or garage feature point, and the pose vector includes the spatial position coordinates and heading angle θ of the i-th tunnel or garage feature point i ; ‖.‖2 is the three-dimensional Euclidean distance; Δθ = |θ - θ i | is the heading angle deviation between the current vehicle and the i-th tunnel or garage feature point;

[0017] The dimension weighting coefficient w of the i-th tunnel or garage feature point iIncluding: the horizontal direction weighting coefficient w of the i-th tunnel or garage feature point xy,i , and the vertical direction weighting coefficient w of the i-th tunnel or garage feature point z,i . The calculation formula is as follows:

[0018] w xy,i = 0.7 - 0.05×sin(2πt / 86400)

[0019] w z,i = 0.3×e {-0.002h}

[0020] In the formula, t is the current time; h is the height of the tunnel entrance or exit, or the height of the garage entrance or exit.

[0021] As a preferred solution of the present invention, the heading angle compensation factor α has the following calculation formula:

[0022] α = 1.2 - 0.015×|θ - θ nom |

[0023] In the formula, θ nom is the nominal road heading angle.

[0024] As a preferred solution of the present invention, the first preset warning distance threshold K1 and the second preset warning distance threshold K2 have the following calculation formula:

[0025]

[0026] In the formula, v is the current vehicle speed; the preset reaction time threshold T1 is 0.8 - 1 s; the preset reaction time threshold T2 is 0.3 - 0.5 s; the reference distance constant C1 is 30 - 45 m; the reference distance constant C2 is 15 - 30 m; a is the vehicle acceleration, with a negative value indicating deceleration and a positive value indicating acceleration.

[0027] As a preferred solution of the present invention, the progressive lighting scheme is specifically implemented as follows:

[0028] Within the first preset time threshold, linearly increase the position lamp brightness from 30% to 100%, and dynamically adjust the low beam brightness L(v) according to the current vehicle speed v, L(v) = L0×(1 + v / 120) δ ; in the formula, L0 is the initial brightness reference value of the low beam, and δ is the non-linear adjustment factor;

[0029] At the same time, start the glare suppression algorithm based on the front camera, and dynamically adjust the low beam irradiation angle ε, ε = ε0 + 0.03v×sin(θ err ); where ε0 is the initial irradiation angle of the low beam, and θ err is the heading angle deviation.

[0030] As a preferred solution of the present invention, the light brightness attenuation solution is specifically implemented as follows:

[0031] Within the second preset time threshold, the low beam brightness is reduced from 100% to 20% based on an exponential curve, and the daytime running lights are turned on synchronously and maintained at 30% brightness; if the ambient light recovery is not detected after exceeding the second preset time threshold, the low beam is turned off and the daytime running lights are kept on constantly.

[0032] As a preferred solution of the present invention, it further includes an exception handling mechanism:

[0033] When the sampling variance of the light sensor is greater than 50 Lux for 3 consecutive times or the positioning deviation is greater than 3 meters for 10 consecutive seconds, the low beam is forcibly turned on and a three-level audible and visual alarm is triggered, and at the same time, a calibration prompt pop-up window is pushed to the vehicle head unit.

[0034] As a preferred solution of the present invention, the three-level audible and visual alarm includes:

[0035] If the three-level response is triggered continuously for more than 3 times, a maintenance request is automatically generated and the nearest service station is reserved;

[0036] When the warning information is projected on the HUD (Head-Up Display), the backlight color temperature of the combination meter is adjusted to warning red at the same time.

[0037] As a preferred solution of the present invention, it further includes a day-night judgment mechanism, and the specific implementation steps are as follows:

[0038] The vehicle latitude is obtained in real time through the in-vehicle GPS;

[0039] Integrate the astronomical application programming interface API, and use the obtained vehicle latitude and date as input parameters to obtain the sunrise time, sunset time and sunshine duration of that date;

[0040] Convert the obtained sunrise and sunset times into UTC timestamps, and further convert them into local times in the time zone where the vehicle is located;

[0041] Compare the current time with the converted sunrise and sunset times to make a day-night state judgment. The specific judgment logic is as follows:

[0042]

[0043] If it is determined to be night, the low beam is turned on.

[0044] An intelligent headlight adaptive control system based on multi-modal data fusion, the system includes:

[0045] A data acquisition module for real-time acquisition of the ambient light value outside the vehicle and the pose vector of the vehicle;

[0046] An environmental perception and collaborative decision-making module. If the duration for which the ambient light value is greater than the preset brightness M does not exceed the preset time T2, a request is sent to the roadside unit via the vehicle-to-everything (V2X) communication protocol to obtain the tunnel or garage feature point set;

[0047] A dynamic distance correction module for calculating the relative distance D using an improved Euclidean distance calculation formula based on the vehicle pose vector and the tunnel or garage feature point set, and introducing a speed-adaptive distance compensation coefficient β(v) to generate the corrected distance D ′ = D × β(v);

[0048] An intelligent lighting control module for comparing the corrected distance D ′ with the preset warning distance threshold K. The preset warning distance threshold K includes a first preset warning distance threshold K1 and a second preset warning distance threshold K2. If the corrected distance D ′ is less than or equal to the first preset warning distance threshold K1, it is determined that the vehicle is about to enter the tunnel or garage, and a progressive lighting scheme is activated; if the corrected distance D ′ is greater than or equal to the second preset warning distance threshold K2, it is determined that the vehicle has left the tunnel or garage, and a lighting brightness attenuation scheme is activated;

[0049] An exception handling module. If the sampling variance of the optical sensor is greater than 50 Lux for 3 consecutive times or the positioning deviation lasts for 10 seconds and is greater than 3 meters, the low beam is forcibly turned on and a three-level sound and light alarm is triggered. At the same time, a calibration prompt pop-up window is pushed to the vehicle head unit.

[0050] Compared with the prior art, the beneficial effects of the present invention are as follows: Through the error-state Kalman filter algorithm for multi-source data fusion, real-time dynamic correction of the vehicle pose vector is achieved, effectively suppressing GPS signal drift and IMU cumulative error, thereby improving the accuracy of vehicle positioning; Based on the ambient light value, V2X communication is dynamically triggered to obtain the tunnel or garage feature point set in real time, breaking through the limitation of the traditional solution that only uses a pre-stored static database, and at the same time supporting the actual application scenarios of newly added garages or tunnels; By setting the dual determination of the preset brightness M and the duration T2, false triggering caused by short-term light changes is avoided, improving the anti-interference ability; When the vehicle is in dynamic driving, introducing a speed-adaptive distance compensation coefficient can reduce the distance calculation error caused by speed fluctuations, thereby improving the distance calculation accuracy in a dynamic environment; The collaborative judgment using a dual warning threshold is adopted to avoid the risk of incorrect control of the vehicle headlights caused by misjudgment of the vehicle leaving or entering. When the vehicle enters, the brightness curve is smoothly increased, and when it leaves, it is gently dimmed according to the exponential decay curve, eliminating the discomfort of sudden light changes in the switch-type control to the vision of the vehicle driver and improving the safety of vehicle driving.

[0051] The above general description and the following description are only exemplary and explanatory, and are not used to limit this application. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] One or more embodiments are exemplarily illustrated by corresponding drawings. These exemplary illustrations and the drawings do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings are shown as similar elements. The drawings do not constitute a scale limitation, and wherein:

[0053] Figure 1 is a flowchart of an intelligent headlight adaptive control method based on multi-modal data fusion provided by an embodiment of the present disclosure;

[0054] Figure 2 is a flowchart of V2X dynamic request and collaborative decision-making based on ambient light value provided by an embodiment of the present invention;

[0055] Figure 3 is a schematic diagram of the adaptive control logic of tunnel or garage vehicle lights based on dynamic warning distance provided by an embodiment of the present invention;

[0056] Figure 4 is a specific implementation flowchart of a progressive lighting scheme and a lighting brightness attenuation scheme provided by an embodiment of the present invention;

[0057] Figure 5 is a flowchart of another intelligent headlight adaptive control method based on multi-modal data fusion provided by an embodiment of the present invention;

[0058] Figure 6 is a schematic diagram of the modular structure of an intelligent headlight adaptive control system based on multi-modal data fusion provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] In order to be able to understand the features and technical content of the embodiments of the present disclosure in more detail, the implementation of the embodiments of the present disclosure will be described in detail below with reference to the drawings. The attached drawings are for reference and illustration only, and are not used to limit the embodiments of the present disclosure. In the following technical description, for the sake of explanation, numerous details are provided to give a full understanding of the disclosed embodiments. However, one or more embodiments may still be implemented without these details. In other cases, well-known structures and devices may be shown in a simplified manner to simplify the drawings.

[0060] In the description and claims of the embodiments of the present disclosure and the above-mentioned drawings, the terms "first", "second", etc. are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so as to implement the embodiments of the present disclosure described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion.

[0061] Unless otherwise specified, the term "plurality" means two or more.

[0062] In the embodiments of the present disclosure, the character " / " indicates that the objects before and after are in an "or" relationship. For example, A / B means: A or B.

[0063] The term "and / or" is a description of the association relationship of objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or, the three relationships of A and B.

[0064] The term "corresponding" may refer to an association relationship or a binding relationship. A corresponding to B means that there is an association relationship or a binding relationship between A and B.

[0065] As Figure 1 shown, it is an embodiment of the present invention. This embodiment provides an intelligent headlight adaptive control method based on multi-modal data fusion, including the following steps:

[0066] S1, Collect the ambient light value outside the vehicle and the vehicle pose vector in real time; wherein, the vehicle-mounted GPS position data and the IMU angular velocity data are fused by the ESKF algorithm to output the corrected vehicle pose vector D v , the vehicle pose vector D v includes the spatial position coordinates (x, y, z) of the vehicle and the heading angle θ of the vehicle;

[0067] In one embodiment, step S1 specifically includes:

[0068] S11, Define the system state and perform initialization processing on the system; wherein, the initial position is provided by the GPS first-frame data; the initial attitude is obtained by performing gravity alignment calculation when the IMU is in a stationary state; the zero bias b g , b a is initialized to 0, and the covariance matrix P0 is set as a diagonal matrix.

[0069] Nominal state vector X:

[0070]

[0071] Error state vector δX:

[0072]

[0073] where p w ∈R 3 is the position vector of the vehicle body in the world coordinate system, v w ∈R 3 is the velocity of the vehicle body in the world coordinate system, q wb ∈SO(3) is the attitude quaternion of the vehicle body in the world coordinate system, b g ∈R 3 is the zero bias of the IMU gyroscope, b a ∈R 3 is the zero bias of the accelerometer, δθ wb ∈R 3 is the attitude error of the vehicle body;

[0074] S12, read the IMU data, and use the read IMU data for nominal state prediction and error state covariance prediction; among them, the IMU data includes angular velocity and acceleration; nominal state prediction includes attitude update and velocity and position update; error state covariance prediction includes the calculation of the state transition matrix F k and update the covariance based on F k ; the calculation formula is as follows:

[0075] Angular velocity: w m = w true + b g + n g

[0076] Acceleration: a m = R wb (a true - g w )+ b a + n a

[0077] where w m is the measured value of the IMU angular velocity, w true is the true angular velocity of the vehicle body, n g is the Gaussian white noise of the gyroscope, n g ~N(0, Q g ), Q g is the gyroscope noise covariance matrix, a m is the measured value of the IMU acceleration, R wb is the rotation matrix from the world coordinate system to the IMU body coordinate system, a true is the true acceleration of the vehicle body in the world coordinate system, g is the gravity acceleration vector in the world coordinate system, n a is the Gaussian white noise of the accelerometer, Qa is the accelerometer noise covariance matrix;

[0078] If the IMU is stationary, then a true = 0, and at this time a m = R wb (-g w ) + b a + n a , that is, the accelerometer measurement value is the projection of the gravity in the opposite direction onto the IMU coordinate system.

[0079] Attitude update:

[0080]

[0081] Velocity and position update:

[0082]

[0083] In the formula, is the quaternion multiplication operator, representing the composition of rotations at the time step; is the attitude quaternion at time k + 1; is the attitude quaternion at time k; Δt is the time step; q((w m - b g )Δt) converts the angular velocity integral to an incremental rotation quaternion; is the velocity vector of the vehicle body in the world coordinate system at time k; is the velocity vector of the vehicle body in the world coordinate system at time k + 1; is the position vector of the vehicle body in the world coordinate system at time k, is the position vector of the vehicle body in the world coordinate system at time k + 1;

[0084] The state transition matrix F k is used to describe how the error propagates over time, and the calculation formula is as follows:

[0085]

[0086] Using the state transition model F k and the process noise Q (mapped to the state space through G k ), the prior covariance matrix P k|k for the next moment is derived from the current posterior error covariance matrix P k+1|k ;

[0087]

[0088] In the formula, I is the identity matrix; a m - b ais the true specific force after subtracting the bias from the accelerometer measurement; -R wb [a m -b a Δt is the influence of the acceleration error on the velocity error after being transformed to the world coordinate system by the rotation matrix; -R wb Δt is the contribution of the accelerometer bias error δb a to the velocity error; w m -b g is the true angular velocity after subtracting the bias from the gyroscope measurement; [·] is the skew-symmetric matrix of the angular velocity or acceleration for attitude error propagation; -IΔt reflects the influence of the gyroscope bias error on the attitude error; Q = diag(Q g , Q a ) is the process noise covariance matrix, i.e., the IMU noise covariance, where Q a is the accelerometer noise covariance, Q g, is the gyroscope noise covariance; G k is the noise drive matrix for mapping the process noise from the noise space to the state space;

[0089] S13. Based on the read GPS observation data, the Kalman gain is calculated by constructing an observation model to update the error state, thereby updating the GPS data; among them, the observation data includes the global position and the heading angle θ gps ,

[0090] Construct the observation model

[0091] Observation equation:

[0092]

[0093] Observation matrix H:

[0094]

[0095] The position error δp w can be extracted through the first row of the observation matrix H, which corresponds to the GPS position observation; the attitude error δθ wb can be extracted through the second row of the observation matrix H, which corresponds to the heading angle observation;

[0096] In the formula, is the global position measurement value provided by the GPS; θ gps is the heading angle measurement value; n z is the observation noise, following a Gaussian distribution X is the state vector, usually including position, velocity, attitude error, and bias, etc.; z is the observation vector, including the GPS position and the heading angle θgps ;

[0097] Calculation of Kalman gain K:

[0098] K = P k|k-1 H T (HP k|k-1 H T + R z ) -1

[0099] State update:

[0100] X k|k = X k|k-1 + K(z - HX k|k-1 )

[0101] In the formula, R z is the GPS observation noise covariance matrix, used to describe the noise intensity of GPS and heading angle measurements; X k|k-1 is the predicted state at time k, calculated through the state transition model F k and the posterior state X k-1|k-1 at the previous moment; z - HX k|k-1 is the observation residual, used to reflect the difference between the actual observation value and the predicted observation value;

[0102] S14. Inject the error into the nominal state, correct the position / velocity, attitude and zero bias respectively, and reset the covariance at the same time;

[0103] Position / velocity correction:

[0104] p w ← p w + δp w

[0105] v w ← v w + δv w

[0106] Attitude correction:

[0107]

[0108] In the formula, q(δθ wb ) is the conversion of the error rotation vector to a quaternion;

[0109] Zero bias correction:

[0110] b g ← b g + δb g

[0111] b a ← b a + δba

[0112] Covariance reset:

[0113] P k|k = (I - KH)P k|k-1

[0114] S15. Extract the heading angle θ from the quaternion q wb and output the pose vector;

[0115]

[0116] where (x, y, z) are the spatial position coordinates of the vehicle body, and θ is the heading angle of the vehicle body;

[0117] Force quaternion normalization: q wb ← q wb / ||q wb || for numerical stability processing;

[0118] S16. Align the IMU and GPS data to the same timestamp through interpolation, calibrate the rotation offset R from the IMU to the vehicle body coordinate system bi , calibrate the lever arm value t from the GPS antenna phase center to the IMU bi , and at the same time, perform adaptive noise adjustment, which includes GPS signal quality detection and IMU motion detection, as follows:

[0119] GPS signal quality detection: When the number of satellites is less than 4, increase R z to reduce the GPS weight; when the HDOP (Horizontal Dilution of Precision) is greater than 2, multiply the horizontal position noise covariance by 2;

[0120] IMU motion detection: When the acceleration variance is greater than the threshold, temporarily increase Q a .

[0121] Furthermore, the ambient light value outside the vehicle is collected by an optical sensor array, and the specific implementation steps are as follows:

[0122] Install multiple optical sensors outside the vehicle in an array layout so that the collected light intensity can cover different directions. For example, install multiple optical sensors on the roof, and / or, the front bumper, and / or, the rearview mirror, and / or, the rear windshield. The sensor housing adopts an IP67 protection level, a splash guard is added at the installation position on the vehicle bottom, and the sensors are automatically calibrated by a standard light source.

[0123] The sensor parameters are set by the vehicle-mounted processor, and the sensor parameters include sampling frequency, range, and calibration coefficient;

[0124] The ABS (Anti-lock Braking System) wheel speed pulses are obtained through the CAN (Controller Area Network) bus. Whenever N pulses are accumulated, a full-array sampling can be triggered. Among them, due to differences in different vehicle configurations, the pulse counting threshold N needs to be adjusted accordingly;

[0125] Assume that the circumference of the car tire is 2m. When 64 pulses correspond to 10 meters, a full-array sampling can be triggered every time 64 pulses are accumulated.

[0126]

[0127] In the formula, the number of teeth of the ABS gear ring is usually set to 48 - 100 teeth;

[0128] The 16-bit digital signals of each sensor are read in parallel using the DMA (Direct Memory Access) double-buffer mode; for example, the numerical range 0x0000 - 0xFFFF corresponds to 0 - 100 klux.

[0129] Assume there are 10 sensors, and each sensor needs to read 16-bit digital signals. When the 2 CAN rate is 400 kbps, the time t required for a single read is:

[0130]

[0131] 0.4ms < 50ms

[0132] Therefore, the time consumption of this single read meets the requirement that the total delay of the vehicle's single read data does not exceed 50 ms.

[0133] The 16-bit digital signals of each sensor in the sensor array are integrated and processed through a weighted average algorithm, and then the final ambient light value is output.

[0134] Assume that the error of a single sensor is ±5%. Then the total error γ after weighted fusion of the 16-bit digital signals of 10 sensors:

[0135]

[0136] In the formula, w i is the weight of the i-th optical sensor, including forward weight, lateral weight, and backward weight; γ i is the error of the i-th optical sensor;

[0137] For example, for sensors located on the roof and front bumper, the forward weight usually accounts for 60%, for sensors located on the rearview mirror, the lateral weight usually accounts for 20%, and for sensors on the rear windshield, the rearward weight usually accounts for 20%.

[0138] The system regularly checks the health status of the sensor. When a sensor fails three times in a row, it will broadcast a fault code through the CAN bus. For example, the fault node can be found by comparing the data of adjacent sensors, and the sliding variance of a single sensor for 10 minutes can be counted. If the sliding variance is greater than the threshold (±200lux), the fault code will be broadcasted through the CAN bus. 2 ), the sensor is deemed to be inoperative.

[0139] S2: If the duration of the ambient light value being greater than the preset brightness M does not exceed the preset time T2, a request is made to the roadside unit through the vehicle-to-vehicle (V2X) communication protocol to obtain a tunnel or garage feature point set;

[0140] like Figure 2 As shown, in one embodiment, step S2 specifically includes:

[0141] S21, when the collected ambient light value exceeds the brightness M, start the timer; if the duration of the ambient light value exceeding M does not reach the preset time T2, trigger the request process;

[0142] S22, encapsulating request information using the vehicle-mounted unit according to the V2X communication protocol standard; wherein the request information includes a vehicle identifier, vehicle location information, and a request type (such as a "tunnel feature point set" or a "garage feature point set");

[0143] S23, using cellular vehicle networking technology, broadcasts the request message to the adjacent roadside units, and selects the communication frequency band of 5.9 GHz;

[0144] S24, the roadside unit checks the legality of the vehicle identity (such as certificate verification), confirms whether the requested area is within the service range, and calls the high-precision map feature point set of the corresponding area from the cloud. The feature point set includes 3D point cloud data (such as tunnel walls, lane lines, signs, etc.), semantic information (such as speed limits, exit locations, obstacle markings) and dynamically updated data (such as temporary construction areas);

[0145] S25, the roadside unit encapsulates the feature point set data into a response message in a standardized format (such as ASN.1 encoding), and sends it to the requesting vehicle through the V2X protocol to ensure data integrity and encrypted transmission (such as TLS);

[0146] S26, the vehicle system receives and analyzes the feature point set data, and fuses the feature point set with the data of the vehicle sensor (such as laser radar, camera) to enhance the environmental perception capability;

[0147] S27. If communication fails or data verification shows an error, the vehicle starts a retry mechanism (using an exponential backoff strategy), while recording a fault log and reporting it to the cloud maintenance system.

[0148] Further, M is 1000 - 1500 Lux. Through experimental testing, when M < 1000 Lux, the system's false trigger rate increases. When M > 1500 Lux, there is a significant response delay. M = 1200 Lux is the optimal value, at which time the false trigger rate is the lowest.

[0149] Further, T2 is 5 - 15 s. The optical sensor requires a certain amount of time from detecting a change in light to triggering. A duration of 5 - 15 s can cover the signal processing delay, thus ensuring the complete execution of the action. Among them, the optimal value of T2 is 10 s.

[0150] S3. Based on the vehicle pose vector D v and the tunnel or garage feature point set, use an improved Euclidean distance calculation formula to calculate the relative distance D, and introduce a speed - adaptive distance compensation coefficient β(v) to generate the corrected distance D ′ = D × β(v); where β(v) = 1 + tanh(0.02v), and v is the current vehicle speed;

[0151] In one embodiment, step S3 specifically includes:

[0152] S31. Based on the vehicle pose vector D v and the tunnel or garage feature point set, use an improved Euclidean distance calculation formula to calculate the relative distance D. The calculation formula is as follows:

[0153]

[0154] In the formula, w i is the dimension - weighted coefficient of the i - th tunnel or garage feature point; n is the number of tunnel or garage feature points; α is the heading angle compensation factor; D i is the pose vector of the i - th tunnel or garage feature point, and the pose vector includes the spatial position coordinates and its heading angle θ of the i - th tunnel or garage feature point i ; ‖.‖2 is the three - dimensional Euclidean distance; Δθ = |θ - θ i | is the heading angle deviation between the current vehicle and the i - th tunnel or garage feature point;

[0155] The dimension - weighted coefficient w of the i - th tunnel or garage feature point i includes: the horizontal - direction weighted coefficient w of the i - th tunnel or garage feature point xy,i and the vertical - direction weighted coefficient w of the i - th tunnel or garage feature point z,i The calculation formula is as follows:

[0156] w xy,i = 0.7 - 0.05 × sin(2πt / 86400)

[0157] w z,i = 0.3 × e {-0.002h}

[0158] In the formula, t is the current time; h is the height of the tunnel entrance or exit, or the height of the garage entrance or exit;

[0159] The heading angle compensation factor α is calculated as follows:

[0160] α = 1.2 - 0.015 × |θ - θ nom | e

[0161] In the formula, θ nom is the nominal road heading angle;

[0162] S32. Introduce the speed - adaptive distance compensation coefficient β(v) = 1 + tanh(0.02v), and use the formula D ′ = D × β(v) to generate the corrected distance D ′ .

[0163] Assume that the vehicle pose vector is D v The spatial position coordinates are (0, 0, 0) m, and the heading angle is 30°; the spatial position coordinates of the feature point D1 are (10, 5, 3) m, and the heading angle is 25°; the current time t is 36000 seconds; the height of the tunnel entrance or exit or the garage entrance or exit h is 5 m; the nominal road heading angle θ nom is 25°, the base of the natural logarithm e is 2, and the vehicle speed is 16.67 m / s.

[0164] w xy,1 = 0.7 - 0.05 × sin(2π × 36000 / 86400) = 0.65

[0165] w z,1 = 0.3 × e {-0.002×5} = 0.303

[0166] Assume w i = w xy,i + w z,i Then w1 = w xy,1 + w z,1 = 0.65 + 0.303 = 0.953

[0167] α = 1.2 - 0.015 × |30° - 25°| 2 = 1.2 - 0.015 × 25 = 0.825

[0168] Δθ = |θ v - θ1| = |30° - 25°| = 5°

[0169]

[0170] β(v) = 1 + tanh(0.02v) = 1 + tanh(0.02×16.67) = 1 + 0.322 = 1.322

[0171] D ′ = D×β(v) = 14.97×1.322 = 19.79m

[0172] S4. Compare and correct the distance D ′ with a preset warning distance threshold K. The preset warning distance threshold K includes a first preset warning distance threshold K1 and a second preset warning distance threshold K2. The first preset warning distance threshold K1 is the warning distance of the vehicle from the tunnel or garage entrance, and the second preset warning distance threshold K2 is the warning distance of the vehicle from the tunnel or garage exit;

[0173] If the corrected distance D ′ is less than or equal to the first preset warning distance threshold K1, it is determined that the vehicle is about to enter the tunnel or garage, and a progressive lighting scheme is started;

[0174] If the corrected distance D ′ is greater than or equal to the second preset warning distance threshold K2, it is determined that the vehicle has left the tunnel or garage, and a lighting brightness attenuation scheme is started.

[0175] As Figure 3 shown, in one embodiment, step S4 specifically includes:

[0176] S41. Calculate the first preset warning distance threshold K1 and the second preset warning distance threshold K2. The calculation formulas are as follows:

[0177]

[0178] In the formula, v is the current vehicle speed; the preset reaction time threshold T1 is 0.8 - 1s; the preset reaction time threshold T2 is 0.3 - 0.5s; the reference distance constant C1 is 30 - 45m; the reference distance constant C2 is 15 - 30m; a is the vehicle acceleration, with a negative value indicating deceleration and a positive value indicating acceleration.

[0179] Assume the car drives into the garage at a constant speed, v = 10km / h ≈ 2.78m / s, T1 = 0.8s, C1 = 35m:

[0180] K1 = 2.78×0.8 + 35 = 37.22m

[0181] Suppose when the vehicle accelerates and drives out of a long tunnel, v = 50 km / h ≈ 13.89 m / s, acceleration a = 0.5 m / s, T2 = 1 s, C2 = 45 m:

[0182]

[0183] S42, compare the corrected distance D ′ with the first preset warning distance threshold K1 and the second preset warning distance threshold K2 respectively;

[0184] S43, if the corrected distance D ′ is less than or equal to the first preset warning distance threshold K1, it is determined that the vehicle is about to enter the tunnel or garage, and a progressive lighting scheme is activated; if the corrected distance D ′ is greater than or equal to the second preset warning distance threshold K2, it is determined that the vehicle has driven out of the tunnel or garage, and a lighting brightness attenuation scheme is activated;

[0185] Furthermore, as Figure 4 shown, the specific implementation methods of the progressive lighting scheme and the lighting brightness attenuation scheme are as follows:

[0186] Progressive lighting scheme: within the first preset time threshold, linearly increase the position light brightness from 30% to 100%, and dynamically adjust the low beam brightness L(v) according to the current vehicle speed v, L(v) = L0 × (1 + v / 120) δ ; in the formula, L0 is the initial brightness reference value of the low beam, and δ is the non-linear adjustment factor; at the same time, activate the glare suppression algorithm based on the front camera to dynamically adjust the low beam irradiation angle ε, ε = ε0 + 0.03v × sin(θ err ); where, ε0 is the initial irradiation angle of the low beam, and θ err is the course angle deviation.

[0187] Furthermore, the optimal value of the first preset time threshold is 0.8 s. Research shows that the threshold for the human eye to perceive sudden brightness changes is usually between 0.3 and 1 second, and 0.8 seconds can ensure that the incremental value of each frame of brightness change is lower than the critical value perceptible by the human eye, thus avoiding visual discomfort.

[0188] Lighting brightness attenuation scheme: within the second preset time threshold, based on the exponential curve, reduce the low beam brightness from 100% to 20%, synchronously turn on the daytime running lights and maintain 30% brightness; if the ambient light does not recover after exceeding the second preset time threshold, turn off the low beam and keep the daytime running lights on constantly.

[0189] Furthermore, the optimal value of the second preset time threshold is 2 s. Research shows that the adaptation time of the human eye to sudden brightness changes is about 1.5 - 2.5 seconds. When the low beam brightness drops from 100% to 20% in an exponential curve within 2 seconds, the brightness change rate is just below the human eye sensitivity threshold, avoiding the generation of a visual jump sensation.

[0190] Assume that the initial brightness reference value of the low beam L0 = 200 Lux, the current vehicle speed v = 20 km / h, and the non-linear adjustment factor δ = 0.5:

[0191]

[0192] Assume that the initial irradiation angle of the low beam ε0 = 0°, the current vehicle speed v = 50 km / h, and the heading angle deviation θ err = 5°:

[0193] ε = 0.03 × 50 × sin(5°) = 1.5 × 0.087 = 0.13°

[0194] As Figure 5 shown, in another embodiment, the method further includes an exception handling mechanism: when the sampling variance of the optical sensor is greater than 50 Lux for 3 consecutive times or the positioning deviation is greater than 3 meters for 10 consecutive seconds, the low beam is forced to turn on and a three-level audible and visual alarm is triggered. At the same time, a calibration prompt pop-up window is pushed to the in-vehicle computer.

[0195] Furthermore, the three-level audible and visual alarm includes: if the three-level response is triggered continuously for more than 3 times, a maintenance request is automatically generated and the nearest service station is reserved; when the warning information is projected on the head-up display (HUD), the backlight color temperature of the combination instrument is adjusted to warning red at the same time.

[0196] In another embodiment, the method further includes a day-night judgment mechanism, and the specific implementation steps are as follows:

[0197] The vehicle latitude is obtained in real time through the on-vehicle GPS, such as 40.7128° north latitude;

[0198] Integrate the astronomical API (Application Programming Interface), and use the obtained vehicle latitude and date as input parameters to obtain the sunrise time, sunset time, and sunshine duration of that date;

[0199] Convert the obtained sunrise and sunset times into UTC timestamps, and further convert them into local times in the time zone where the vehicle is located;

[0200] Compare the current time with the converted sunrise and sunset times to make a day-night status judgment. The specific judgment logic is as follows:

[0201]

[0202] If it is determined to be night, turn on the dipped headlights.

[0203] Further illustrate with a specific scenario:

[0204] Suppose the vehicle is located at 40.7° north latitude and the time is December 25th. Then, the sunrise time obtained through the astronomical API is 07:15, and the sunset time is 16:45. If the current time is 18:00, according to the day-night state judgment logic, it is night at this time. Therefore, it is necessary to turn on the dipped headlights.

[0205] As Figure 6 shown, this is another embodiment of the present invention. This embodiment provides an intelligent headlight adaptive control system based on multi-modal data fusion. The system includes:

[0206] A data acquisition module for real-time collecting the ambient light value outside the vehicle and the pose vector of the vehicle;

[0207] An environment perception and collaborative decision-making module for, if the duration that the ambient light value is greater than the preset brightness M does not exceed the preset time T2, sending a request to the roadside unit through the vehicle-to-everything (V2X) communication protocol to obtain a set of tunnel or garage feature points;

[0208] A dynamic distance correction module for calculating the relative distance D according to the vehicle pose vector and the set of tunnel or garage feature points, and introducing a speed-adaptive distance compensation coefficient β(v) to generate a corrected distance D ′ = D × β(v);

[0209] An intelligent light control module for comparing the corrected distance D ′ with a preset warning distance threshold K. The preset warning distance threshold K includes a first preset warning distance threshold K1 and a second preset warning distance threshold K2. If the corrected distance D ′ is less than or equal to the first preset warning distance threshold K1, it is determined that the vehicle is about to enter a tunnel or a garage, and a progressive lighting scheme is started; if the corrected distance D ′ is greater than or equal to the second preset warning distance threshold K2, it is determined that the vehicle has left the tunnel or the garage, and a lighting brightness attenuation scheme is started;

[0210] An exception handling module for, if the sampling variance of the optical sensor is greater than 50 Lux for 3 consecutive times or the positioning deviation is greater than 3 meters for 10 consecutive seconds, forcibly turning on the dipped headlights and triggering a three-level sound and light alarm, and at the same time pushing a calibration prompt pop-up window to the vehicle head unit.

[0211] In summary, through the error-state Kalman filtering algorithm for multi-source data fusion, the real-time dynamic correction of the vehicle pose vector is realized, effectively suppressing the GPS signal drift and the IMU cumulative error, thereby improving the accuracy of vehicle positioning; based on the ambient light value, the vehicle network communication is dynamically triggered to obtain the tunnel or garage feature point set in real time, breaking through the limitation of the traditional solution that only uses a pre-stored static database, and at the same time supporting the actual application scenarios of newly added garages or tunnels; by setting the double determination of the preset brightness M and the duration T2, false triggering caused by short-term light changes is avoided, and the anti-interference ability is improved; when the vehicle is in dynamic driving, introducing a speed-adaptive distance compensation coefficient can reduce the distance calculation error caused by speed fluctuations, thereby improving the distance calculation accuracy in a dynamic environment; the collaborative judgment of double warning thresholds is adopted to avoid the risk of miscontrolling the car headlights caused by misjudgment when the vehicle exits or enters. When the vehicle enters, the brightness curve is smoothly increased, and when it exits, it is gently dimmed according to the exponential decay curve, eliminating the discomfort of the sudden light change of the switch-type control to the vision of the car driver and improving the safety of car driving.

[0212] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description, and it is intended to include all changes falling within the meaning and scope of the equivalent elements of the claims in the present invention.

Claims

1. An intelligent headlight adaptive control method based on multimodal data fusion, characterized in that, Including: Step 1, collect the ambient light value outside the vehicle and the vehicle pose vector in real time; among them, the vehicle pose vector D is output by fusing the vehicle-mounted global positioning system (GPS) position data and the inertial measurement unit (IMU) angular velocity data through the error state Kalman filter (ESKF) algorithm v , the vehicle pose vector D v includes the spatial position coordinates (x, y, z) of the vehicle and the heading angle θ of the vehicle; Step 2, if the duration for which the ambient light value is greater than the preset brightness M does not exceed the preset time T2, send a request to the roadside unit via the vehicle-to-everything (V2X) communication protocol to obtain the tunnel or garage feature point set. Step 3, based on the vehicle pose vector D v and the tunnel or garage feature point set, use the improved Euclidean distance calculation formula to calculate the relative distance D, and introduce a speed-adaptive distance compensation coefficient β(v) to generate the corrected distance D ′ = D × β(v); where β(v) = 1 + tanh(0.02v), and v is the current vehicle speed; Step 4, compare and correct the distance D ′ with a preset warning distance threshold K, where the preset warning distance threshold K includes a first preset warning distance threshold K1 and a second preset warning distance threshold K2. The first preset warning distance threshold K1 is the warning distance of the vehicle from the entrance of the tunnel or garage, and the second preset warning distance threshold K2 is the warning distance of the vehicle from the exit of the tunnel or garage; If the correction distance D ′ is less than or equal to the first preset warning distance threshold K1, it is determined that the vehicle is about to enter a tunnel or a garage, and a progressive lighting scheme is activated; If the correction distance D ′ is greater than or equal to the second preset warning distance threshold K2, it is determined that the vehicle has left the tunnel or garage, and a lighting brightness attenuation scheme is activated.

2. The intelligent headlight adaptive control method based on multi-modal data fusion according to claim 1, wherein, The relative distance D is calculated using the improved Euclidean distance calculation formula as follows: Wherein, w i is the dimension weighting coefficient of the i-th tunnel or garage feature point; n is the number of feature points of the tunnel or garage; α is the heading angle compensation factor; D i is the pose vector of the i-th tunnel or garage feature point, and the pose vector includes the spatial position coordinates and the heading angle θ of the i-th tunnel or garage feature point i ; ‖.‖2 is the three-dimensional Euclidean distance; Δθ = |θ - θ i | is the heading angle deviation between the current vehicle and the i-th tunnel or garage feature point; The dimensional weighting coefficient w of the i-th tunnel or garage feature point i includes: the horizontal direction weighting coefficient w of the i-th tunnel or garage feature point xy,i , and the vertical direction weighting coefficient w of the i-th tunnel or garage feature point z,i , and the calculation formula is as follows: w xy,i = 0.7 - 0.05 × sin(2πt / 86400) w z,i = 0.3 × e {-0.002h} In the formula, t is the current time; h is the height of the tunnel entrance or exit, or the height of the garage entrance or exit.

3. The intelligent headlight adaptive control method based on multi-modal data fusion according to claim 2, wherein, The heading angle compensation factor α is calculated as follows: α = 1.2 - 0.015 × |θ - θ nom | In the formula, θ nom is the nominal road heading angle.

4. The intelligent headlight adaptive control method based on multi-modal data fusion according to claim 1, characterized in that, The first preset warning distance threshold K1 and the second preset warning distance threshold K2 are calculated as follows: In the formula, v is the current vehicle speed; the preset reaction time threshold T1 is 0.8 - 1 s; The preset reaction time threshold T2 is 0.3 - 0.5 s; the reference distance constant C1 is 30 - 45 m; the reference distance constant C2 is 15 - 30 m; a is the vehicle acceleration, with a negative value indicating deceleration and a positive value indicating acceleration.

5. The intelligent headlight adaptive control method based on multi-modal data fusion according to claim 1, characterized in that, The progressive lighting scheme is specifically implemented as follows: Within the first preset time threshold, linearly increase the position lamp brightness from 30% to 100%, and dynamically adjust the low beam brightness L(v) according to the current vehicle speed v, where L(v) = L0×(1 + v / 120). δ ; where L0 is the initial brightness reference value of the low beam, and δ is the non-linear adjustment factor. At the same time, start the glare suppression algorithm based on the front camera, and dynamically adjust the irradiation angle ε of the low beam, ε = ε0 + 0.03v × sin(θ err ); where ε0 is the initial irradiation angle of the low beam, and θ err is the heading angle deviation.

6. The intelligent headlight adaptive control method based on multi-modal data fusion according to claim 1, characterized in that The lighting brightness attenuation scheme is specifically implemented as follows: Within the second preset time threshold, gradually reduce the low beam brightness from 100% to 20% based on an exponential curve, and simultaneously turn on the daytime running lights and maintain a brightness of 30%; if the ambient light recovery is not detected after exceeding the second preset time threshold, turn off the low beam and keep the daytime running lights on constantly.

7. The intelligent headlight adaptive control method based on multi-modal data fusion according to claim 1, wherein, It also includes an exception handling mechanism: When the sampling variance of the light sensor is greater than 50 Lux for 3 consecutive times or the positioning deviation is greater than 3 meters for 10 consecutive seconds, forcefully turn on the low beam and trigger a three-level audible and visual alarm, and at the same time push a calibration prompt pop-up window to the in-vehicle computer.

8. The intelligent headlamp adaptive control method based on multi-modal data fusion according to claim 7, characterized in that, The three-level audible and visual alarm includes: If the three-level response is continuously triggered more than 3 times, automatically generate a maintenance request and reserve the nearest service station; When projecting the alarm information on the head-up display (HUD), simultaneously adjust the backlight color temperature of the combination meter to warning red.

9. The intelligent headlight adaptive control method based on multi-modal data fusion according to claim 1, wherein, It also includes a day-night judgment mechanism, and the specific implementation steps are as follows: Obtain the vehicle latitude in real time through the on-vehicle GPS. Integrate the astronomical application programming interface (API), and use the obtained vehicle latitude and date as input parameters to obtain the sunrise time, sunset time, and sunshine duration of that date. Convert the obtained sunrise and sunset times into UTC timestamps, and further convert them into the local time of the time zone where the vehicle is located. Compare the current time with the converted sunrise and sunset times to make a day-night status judgment. The specific judgment logic is as follows: If it is determined to be night, turn on the low beam.

10. The control system for the intelligent headlight adaptive control method based on multi-modal data fusion according to any one of claims 1 to 9, characterized in that, The system includes: A data acquisition module for real-time acquisition of the ambient light value outside the vehicle and the vehicle pose vector. An environment perception and collaborative decision-making module for, if the duration for which the ambient light value is greater than the preset brightness M does not exceed the preset time T2, sending a request to the roadside unit via the vehicle-to-everything (V2X) communication protocol to obtain the tunnel or garage feature point set. The dynamic distance correction module is used to calculate the relative distance D according to the vehicle pose vector and the tunnel or garage feature point set, adopt an improved Euclidean distance calculation formula, and introduce a speed-adaptive distance compensation coefficient β(v) to generate a corrected distance D ′ = D × β(v); Intelligent lighting control module, used to compare and correct the distance D ′ with a preset warning distance threshold K, the preset warning distance threshold K includes a first preset warning distance threshold K1 and a second preset warning distance threshold K2. If the corrected distance D ′ is less than or equal to the first preset warning distance threshold K1, it is determined that the vehicle is about to enter a tunnel or a garage, and a progressive lighting scheme is started; if the corrected distance D ′ is greater than or equal to the second preset warning distance threshold K2, it is determined that the vehicle has left the tunnel or the garage, and a lighting brightness attenuation scheme is started; An exception handling module for, if the sampling variance of the light sensor is greater than 50 Lux for 3 consecutive times or the positioning deviation is greater than 3 meters for 10 consecutive seconds, forcefully turn on the low beam and trigger a three-level audible and visual alarm, and at the same time push a calibration prompt pop-up window to the in-vehicle computer.

Images