An intelligent vehicle lamp control method for dynamically changing a matrix vehicle lamp illumination area

Abstract

The application relates to an intelligent vehicle lamp control method for dynamically changing the illumination area of a matrix vehicle lamp, wherein the Fast R-CNN convolutional neural network is used to acquire the position information of a front vehicle, and a millimeter wave radar is used to acquire the distance and azimuth angle information of the front vehicle; an RBF neural network model is used to predict the extinguished lamp bead area of a fixed LED matrix module and the rotating direction and angle of a movable LED matrix module; a central controller ECU changes the high and low levels of the output of a PWM controller and the pulse output of a PLC controller to respectively adjust the brightness of the lamp bead area in the fixed LED matrix module, the rotating direction and angle of the movable LED matrix module, and the light compensation of the extinguished lamp bead area of the fixed LED matrix module, so that a larger illumination area is achieved. The application solves the problem of the light blind area generated by the traditional matrix LED vehicle lamp when meeting vehicles; the safety visual field demand of the driver at night is met, and the night driving safety is improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent vehicle lighting technology, specifically relating to an intelligent vehicle lighting control method that dynamically changes the illumination area of ​​matrix headlights. Background Technology

[0002] Existing intelligent headlights for nighttime driving typically employ the ADB (Adaptive High Beam) method. This method uses visual sensors to detect external information, which is then processed by the ECU to control the on / off state of some LEDs. When a vehicle encounters oncoming traffic at night, it automatically switches between high and low beams, automatically turning off a portion of the high beam to prevent glare for the driver of the vehicle ahead. However, when vehicles meet oncoming traffic at night, the sudden switching off of some LEDs illuminates the oncoming vehicle, significantly reducing the driver's visibility. Furthermore, the abrupt switching off of some lights can cause a visual shock to the driver, greatly compromising nighttime driving safety.

[0003] This invention utilizes external visual perception and radar detection to train a Fast R-CNN convolutional neural network with the perceived information. The convolutional neural network consists of convolutional layers, pooling layers, and fully connected layers. It is trained by extracting features from the processed image and outputs the extracted vehicle position data. The RBF neural network is a feedforward neural network consisting of an input layer, hidden layers, and an output layer. The transformation function of the neurons in the hidden layer is the radial basis function. The RBF algorithm mainly consists of two stages: a self-organizing learning stage, which is an unsupervised learning process that solves for the center vector and variance of the hidden layer basis functions; and a supervised learning stage, which solves for the weights between the hidden layer and the output layer, adjusting the cluster centers to adjust the basis function centers.

[0004] The PLC controller outputs a certain number of pulses via commands from the ECU to control the servo motor. The servo motor rotates the movable LED matrix module, causing it to rotate at a specific angle and direction. The PWM controller modulates the pulse width to control the output of different high and low frequencies, thereby controlling the brightness of some LEDs in the fixed LED matrix module. By changing the illumination angle of the movable matrix module, the illumination area of ​​the corresponding extinguished LEDs in the fixed matrix module is compensated, maximizing the light illumination area. This invention effectively solves the problem of driving safety visibility faced by traditional LED matrix vehicle lights. Summary of the Invention

[0005] To improve vehicle safety when meeting oncoming traffic at night, this invention provides an intelligent headlight control method that dynamically changes the illumination area of ​​matrix headlights. This method can be applied to various operating conditions when vehicles are driving at night, improving the driver's safe visibility when meeting oncoming traffic at night; preventing blind spots and visual shock caused by the sudden shut-off of some lights, and enhancing the driver's safe driving performance at night.

[0006] To achieve the above objectives, the specific technical solution of the present invention is as follows:

[0007] A smart headlight control method for dynamically changing the illumination area of ​​matrix headlights includes the following steps:

[0008] 1) Data preparation: Collect a dataset of vehicles in front of the vehicle. The dataset includes image data of vehicles in front of the vehicle obtained by a binocular camera installed above the windshield of the vehicle; vehicle speed data obtained by the vehicle speed sensor; distance and azimuth data of vehicles in front obtained by millimeter-wave radar; and divide the headlight LED matrix module into fixed LED matrix modules and movable LED matrix modules.

[0009] 2) Input the acquired dataset of vehicles in front of the vehicle into the Fast R-CNN convolutional neural network model for training, and output the position information of the vehicles in front in the image coordinate system; transmit the distance and azimuth data of the vehicles in front acquired by the millimeter-wave radar and the position information of the vehicles in front output by the Fast R-CNN convolutional neural network to the central controller ECU via the CAN bus for data fusion to obtain the coordinates (x0, y0), length l, and width w of the vehicles in front, where the coordinates (x0, y0) are the position coordinates of the vehicles in front centered on the headlights of the vehicle.

[0010] 3) Input the coordinates (x0, y0), length l, and width w of the vehicle ahead obtained in step 2) into the RBF neural network model for prediction, and output the area A of the LED beads to be turned off by the fixed LED matrix module. i The rotation direction of the movable LED matrix module is Γ. j Angle θ j information;

[0011] 4) The central controller ECU will select the LED bead area A in the fixed LED matrix module to be turned off. i Data is transmitted to the PWM controller via the CAN bus. The PWM controller modulates the pulse width to control the output of different high and low levels, thereby adjusting the LED bead area A of the fixed LED matrix module. iThe brightness of the LED beads; the central controller ECU transmits the direction and angle information that the movable matrix module needs to rotate to the PLC controller via the CAN bus. The PLC controller sends pulses to the servo motor to control the servo motor to drive the movable LED matrix module to rotate within the specified direction and angle;

[0012] 5) Dynamically control the matrix headlights based on the vehicle's operating status, the movement of vehicles ahead, and their position information.

[0013] Furthermore, the above control method is implemented based on the following devices: a central controller (ECU), an in-vehicle binocular camera, a millimeter-wave radar, a vehicle speed sensor, a PWM controller, a PLC controller, a servo motor, an LED matrix module driver, a movable LED matrix module, and a fixed LED matrix module. The in-vehicle binocular camera, millimeter-wave radar, and vehicle speed sensor are connected to the central controller (ECU) via a CAN bus. The in-vehicle binocular camera acquires image information of the vehicle in front, the millimeter-wave radar acquires distance and coordinate data of the vehicle in front, and the vehicle speed sensor acquires the speed data of the vehicle itself. The PWM controller and PLC controller are connected to the ECU via a CAN bus. The PWM controller and PLC controller receive information processed by the ECU and output pulse width and pulse signals, respectively. The PWM controller is connected to the fixed LED matrix module and controls the on / off state of LEDs in a specific LED area of ​​the fixed LED matrix module through the output pulse width signal. The PLC controller is connected to the servo motor, and the servo motor is connected to the movable LED matrix module. The servo motor controls the rotation angle of the servo motor through the output pulse signal. The movable LED matrix module rotates in a specified direction and angle under the drive of the servo motor.

[0014] Furthermore, the fixed matrix module mentioned in step 1) above is used to adjust the brightness of the LEDs in the matrix area, and can turn off or reduce the brightness of local LED areas; the movable matrix module adjusts the rotation of the matrix module under the drive of the servo motor, bevel gear and swing arm, and changes the illumination direction and angle of the LEDs in the movable matrix module.

[0015] Furthermore, the process of data fusion in the central controller ECU of the millimeter-wave radar and the Fast R-CNN convolutional neural network for the vehicle information in front in step 2) above includes the following: converting the radar coordinate system into an image coordinate system, mapping the two-dimensional point cloud image output by the millimeter-wave radar to the image coordinate system, stitching the mapped image with the image output by the Fast R-CNN convolutional neural network, and matching the position information of the vehicle image in the mapped image with the position information of the vehicle image in the image trained by the convolutional network; obtaining the coordinate information of the matched vehicle in the radar coordinate system, and converting the coordinate information in the radar coordinate system into coordinates (x0, y0) in the world coordinate system; establishing a pixel coordinate system, obtaining the length and width information of the vehicle in the pixel coordinate system, and converting the length and width information of the vehicle in the pixel coordinate system into the length l and width w of the vehicle in the world coordinate system.

[0016] Furthermore, in step 3) above, the fixed matrix LED illumination area A i The prediction process is as follows:

[0017] 3.1) Arrange the matrix LEDs in two columns, with n LEDs in each column. Number each matrix LED. The headlight's left and right illumination angle range is 30°, i.e., 15° to the left and right. At this time, the illumination angle of each LED is... Furthermore, the angles of illumination from adjacent LED beads on the left and right sides are continuously formed;

[0018] 3.2) Given the specific coordinates (x0, y0), length l, and width w of the vehicle ahead, obtain the coordinates of the four corners (left and right) of the vehicle's foremost and rearmost ends, as follows:

[0019] 3.3) Select points C and F to obtain the headlight illumination angle value. The angle value δ corresponding to the light illuminating the area around the vehicle should be extinguished.

[0020]

[0021] 3.4) The angle γ from the center of the headlight to the left or right where it is not turned off: Starting from the angle of the headlights remaining lit to the left or right from the center, the calculation begins. The calculated angle δ is the combined angle of the light rays from adjacent LEDs. Based on the light ray angle δ, the LED numbers to be extinguished are obtained, thus determining the extinguished LED region A. i ;

[0022] 3.5) Using the specific coordinates (x0, y0), length l, and width w of the vehicle ahead as a dataset, the RBF neural network model is trained. The trained RBF neural network model is then tested on a test set to predict the area of ​​the LEDs to be turned off; the fixed matrix module rotates in the direction Γ.j The corresponding rotation angle θ j It is expressed as follows:

[0023] Γ j ={Left turn, right turn, tilt, tilt, tilt left, tilt left, tilt right, tilt right}

[0024] θ j ={α1, α2, β1, β2, α1β1, α1β2, α2β1, α2β2}

[0025] Where α1 is the left turn angle α1; α2 is the right turn angle α2; β1 is the depression angle; β2 is the elevation angle; α1β1 is the left turn angle α1 followed by the depression angle β1; α1β2 is the left turn angle α1 followed by the elevation angle β2; α2β1 is the right turn angle α2 followed by the depression angle β1; α2β2 is the right turn angle α2 followed by the elevation angle β2; the rotation angle α i β i The range is

[0026] Furthermore, the dynamic control strategy in step 5) above includes the following:

[0027] When the millimeter-wave radar detects a vehicle within a range of [100-200m], it predicts the area A of LEDs that the fixed matrix module needs to extinguish. i The direction of rotation of the movable matrix module Γ j and angle θ j Simultaneously, the movable matrix module adjusts the direction and angle, and controls the LED bead area A of the fixed matrix module when it is turned off. i The extinguished illumination area is compensated for to prevent light jumps during light combination and make the light continuity more uniform. Through the cooperation of movable and fixed matrix modules, the illumination angle of the lamp beads corresponding to the area of ​​light that dazzles the driver is adjusted to maximize the illumination area.

[0028] Compared with current technology, this invention divides the LED matrix module into movable and fixed types; by combining the two quickly and efficiently in lighting, a larger illumination area is generated when meeting oncoming traffic or pedestrians, which effectively solves the safety problems such as blind spots caused to the driver's vision when one or both sides turn off part of the lights when meeting oncoming traffic, and visual impact caused to the driver by the sudden extinguishing of the lamp beads. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of a movable LED matrix module device.

[0030] Figure 2 This is a schematic diagram of the intelligent vehicle lighting control device of the present invention.

[0031] Figure 3 This is a flowchart of the intelligent vehicle lighting control method of the present invention. Detailed Implementation

[0032] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. It should be noted that the following detailed description only provides one intelligent vehicle lighting solution and its implementation principle, but the scope of protection of the present invention is not limited thereto.

[0033] like Figure 1 The movable LED matrix module shown, including the aforementioned angle and direction transformation system, is implemented based on the following method: It includes: 1-movable LED matrix module, 2-LED beads, 3-swing arm, 4-servo motor, 5-bevel gear, and 6-support body. External data is used to predict the rotation direction and angle information of the movable matrix module via a Fast R-CNN convolutional neural network and an RBF neural network, and transmitted to the ECU for data processing. The processed data is then transmitted to the PLC controller via a CAN bus. The PLC controller outputs certain pulse signals to the servo motor to control the servo motor to drive the bevel gear and swing arm to rotate. The movable LED matrix module rotates within a specific direction and angle through one or a combination of these movements. Figure 2 The method for dynamically changing the illumination area of ​​matrix headlights is shown in the figure. The above control method is implemented based on the following system: a central controller ECU, a sensor detection system, a data transmission system, an LED array control method, and a motor controller system.

[0034] In a preferred embodiment of the present invention, the control method of the present invention is implemented based on the following devices: a central controller ECU, an in-vehicle binocular camera, a millimeter-wave radar, a vehicle speed sensor, a PWM controller, a PLC controller, a servo motor, an LED matrix module driver, a movable LED matrix module, and a fixed LED matrix module; the in-vehicle binocular camera, millimeter-wave radar, and vehicle speed sensor are connected to the central controller ECU via a CAN bus. The in-vehicle binocular camera acquires image information of the vehicle in front, the millimeter-wave radar acquires distance and coordinate data of the vehicle in front, and the vehicle speed sensor acquires speed data of the vehicle itself; the PWM controller and PLC controller are connected to the ECU via a CAN bus, and the PWM controller and PLC controller receive information processed by the ECU and output pulse width and pulse signals respectively; the PWM controller is connected to the fixed LED matrix module and controls the on / off state of the LEDs in a specific LED bead area in the fixed LED matrix module through the output pulse width signal; the PLC controller is connected to the servo motor, and the servo motor is connected to the movable LED matrix module. The servo motor controls the rotation angle of the servo motor through the output pulse signal, and the movable LED matrix module rotates in a specified direction and angle under the drive of the servo motor.

[0035] To achieve the above objectives, the specific technical solution of the present invention is as follows: A smart vehicle headlight control method for dynamically changing the illumination area of ​​matrix headlights, comprising the following steps:

[0036] 1) Data preparation: Collect a dataset of vehicles in front of the vehicle. The dataset includes image data of vehicles in front of the vehicle obtained by a binocular camera installed above the windshield of the vehicle; vehicle speed data obtained by the vehicle speed sensor; distance and azimuth data of vehicles in front obtained by millimeter-wave radar; and divide the headlight LED matrix module into fixed LED matrix modules and movable LED matrix modules.

[0037] As a preferred embodiment of the present invention, the fixed matrix module is used to adjust the brightness of the LEDs in the matrix area, and can turn off or reduce the brightness of a local LED area; the movable matrix module adjusts the rotation of the matrix module under the drive of the servo motor, bevel gear and swing arm, and changes the illumination direction and angle of the LEDs in the movable matrix module.

[0038] 2) Input the acquired dataset of vehicles in front of the vehicle into the Fast R-CNN convolutional neural network model for training, and output the position information of the vehicles in front in the image coordinate system; transmit the distance and azimuth data of the vehicles in front acquired by the millimeter-wave radar and the position information of the vehicles in front output by the Fast R-CNN convolutional neural network to the central controller ECU via the CAN bus for data fusion to obtain the coordinates (x0, y0), length l, and width w of the vehicles in front, where the coordinates (x0, y0) are the position coordinates of the vehicles in front centered on the headlights of the vehicle.

[0039] As a preferred embodiment of the present invention, the process of data fusion in the central controller ECU of step 2) of the forward vehicle information obtained by the millimeter-wave radar and the Fast R-CNN convolutional neural network includes the following: converting the radar coordinate system into an image coordinate system, mapping the two-dimensional point cloud image output by the millimeter-wave radar to the image coordinate system, stitching the mapped image with the image output by the Fast R-CNN convolutional neural network, and matching the forward vehicle image position information in the mapped image with the forward vehicle image position information in the image trained by the convolutional network; obtaining the coordinate information of the matched forward vehicle in the radar coordinate system, and converting the coordinate information in the radar coordinate system into coordinates (x0, y0) in the world coordinate system; establishing a pixel coordinate system, obtaining the length and width information of the forward vehicle in the pixel coordinate system, and converting the length and width information of the forward vehicle in the pixel coordinate system into the length l and width w of the vehicle in the world coordinate system.

[0040] 3) Input the coordinates (x0, y0), length l, and width w of the vehicle ahead obtained in step 2) into the RBF neural network model for prediction, and output the area A of the LED beads to be turned off by the fixed LED matrix module. i The rotation direction of the movable LED matrix module is Γ. j Angle θ j information.

[0041] In a preferred embodiment of the present invention, the fixed matrix LED illumination area A in step 3) i The prediction process is as follows:

[0042] 3.1) Arrange the matrix LEDs in two columns, with n LEDs in each column. Number each matrix LED. The headlight's left and right illumination angle range is 30°, i.e., 15° to the left and right. At this time, the illumination angle of each LED is... Furthermore, the angles of illumination from adjacent LED beads on the left and right sides are continuously formed;

[0043] 3.2) Given the specific coordinates (x0, y0), length l, and width w of the vehicle ahead, obtain the coordinates of the four corners (left and right) of the vehicle's foremost and rearmost ends, as follows:

[0044] 3.3) Select points C and F to obtain the headlight illumination angle value. The angle value δ corresponding to the light illuminating the area around the vehicle should be extinguished.

[0045]

[0046] 3.4) The angle γ from the center of the headlight to the left or right where it is not turned off: Starting from the angle of the headlights remaining lit to the left or right from the center, the calculation begins. The calculated angle δ is the combined angle of the light rays from adjacent LEDs. Based on the light ray angle δ, the LED numbers to be extinguished are obtained, thus determining the extinguished LED region A. i ;

[0047] 3.5) Using the specific coordinates (x0, y0), length l, and width w of the vehicle ahead as a dataset, the RBF neural network model is trained. The trained RBF neural network model is then tested on a test set to predict the area of ​​the LEDs to be turned off; the fixed matrix module rotates in the direction Γ. j The corresponding rotation angle θ j It is expressed as follows:

[0048] Γ j ={Left turn, right turn, tilt, tilt, tilt left, tilt left, tilt right, tilt right}

[0049] θ j={α1, α2, β1, β2, α1β1, α1β2, α2β1, α2β2}

[0050] Where α1 is the left turn angle α1; α2 is the right turn angle α2; β1 is the depression angle; β2 is the elevation angle; α1β1 is the left turn angle α1 followed by the depression angle β1; α1β2 is the left turn angle α1 followed by the elevation angle β2; α2β1 is the right turn angle α2 followed by the depression angle β1; α2β2 is the right turn angle α2 followed by the elevation angle β2; the rotation angle α i β i The range is

[0051] 4) The central controller ECU will select the LED bead area A in the fixed LED matrix module to be turned off. i Data is transmitted to the PWM controller via the CAN bus. The PWM controller modulates the pulse width to control the output of different high and low levels, thereby adjusting the LED bead area A of the fixed LED matrix module. i The brightness of the LED beads; the central controller ECU transmits the direction and angle information of the movable matrix module to the PLC controller via the CAN bus. The PLC controller sends pulses to the servo motor to control the servo motor to drive the movable LED matrix module to rotate within the specified direction and angle.

[0052] 5) Dynamically control the matrix headlights based on the vehicle's operating status, the movement of vehicles ahead, and their position information.

[0053] As a preferred embodiment of the present invention, the dynamic control strategy in step 5) includes the following:

[0054] When the millimeter-wave radar detects a vehicle within a range of [100-200m], it predicts the area A of LEDs that the fixed matrix module needs to extinguish. i The direction of rotation of the movable matrix module Γ j and angle θ j While the fixed matrix module is extinguishing the LEDs in its designated area, the movable matrix module adjusts its direction and angle, and controls the LED area A extinguished by the fixed matrix module. i The system compensates for the extinguished illumination areas to prevent light jumps during light combination and ensures more uniform light continuity. Through the coordination of movable and fixed matrix modules, it adjusts the illumination angle of the LEDs corresponding to areas that cause glare to the driver, reducing the visual impact on the driver ahead and maximizing the illuminated area, thus minimizing blind spots. This control method reduces the visual impact and blind spots experienced by drivers when meeting oncoming traffic at night due to simultaneous or partial extinguishing of LEDs by both vehicles, significantly improving nighttime driving safety.

[0055] In a preferred embodiment of the present invention, step 2) involves preprocessing the acquired image information, including image distortion correction and median filtering for noise reduction. The preprocessed image dataset is then transmitted to a Fast R-CNN convolutional neural network model for training. Gradient descent is used to obtain the minimum value ε of the risk function on the training set. Overfitting is reduced through empirical risk minimization and parameter regularization. The accuracy Acc achieved on the test set reaches 97%, making it applicable to real-world driving scenarios. The formulas for image distortion correction, median filtering for noise reduction, gradient descent, the minimum value ε of the risk function, and the accuracy Acc are as follows:

[0056] Radial distortion correction formula: (x0, y0) represents the original location of the distorted point in the image. The radius is the center of the origin.

[0057] Tangential distortion correction formula: p2 and p1 are tangential distortion coefficients.

[0058] Median filtering:

[0059] g(x, y)=med{f(xk, yl), (k, l∈W)}

[0060] g(x, y) and f(x, y) are the processed image and the original image, respectively, and W is a two-dimensional template.

[0061] Minimum ε of the risk function of gradient descent:

[0062]

[0063] Where: ε t Let be the parameter value at the t-th iteration, and α be the learning rate.

[0064] Accuracy (Accel):

[0065] in: Test set size, I(*) is the indicator function, f (*) (x) is the optimal function.

[0066] Image features are extracted from the convolutional neural network layer. RPN is used to generate candidate boxes for vehicles on the feature map. For each candidate box, Fast R-CNN is used to classify it to determine whether it is a vehicle and to provide the vehicle's location information. The vehicle's location feature information is extracted by the Fast R-CNN convolutional neural network model and is combined with the location information of the vehicle in front obtained by the millimeter-wave radar. The data is then transmitted to the ECU via the CAN bus for data fusion to provide the specific location information of the vehicle in front.

[0067] In a preferred embodiment of the present invention, step 3) uses an RBF neural network model to take the specific location information of the vehicle as input, and obtains the region A illuminating the position of the vehicle in front through model training. i The area A of the fixed matrix headlights that is turned off i The direction of rotation of the movable LED matrix headlights Γ j and angle θ j As output, the algorithm formula for the RBF neural network is as follows:

[0068] The Gaussian function is used as the activation function in an RBF neural network.

[0069]

[0070] Where: ||x i -ω i || is the Euclidean norm, ω i Let σ be the center of the Gaussian function, and σ be the variance of the Gaussian function. For the P-th input sample, Let p be the center vector of the radial basis function of the i-th hidden layer node, where p = 1, 2, ..., P, and P is the total number of samples.

[0071] RBF neural network output:

[0072]

[0073] Where: ω ij 2 Let y be the weights from the i-th node in the hidden layer to the j-th node in the output layer. j This represents the actual output of the j-th output node in the neural network.

[0074] In a preferred embodiment of the present invention, in step 4), the LED bead area A of the fixed LED matrix module is turned off. i The direction of rotation of the movable LED matrix module Γ j and angle θ j Data is transmitted via CAN bus to the PWM and PLC controllers respectively; the PWM controller modulates the output pulse width and transmits it to the LED matrix module driver to control area A in the fixed LED matrix module. i The brightness of some LED beads; the PLC controller sends a certain number of pulses to the servo motor according to instructions to control the servo motor's rotation. Left and right rotation is controlled by bevel gears, and pitch rotation is controlled by a swing arm. Alternatively, a combination of gears and swing arms can be used to rotate the movable matrix module. j Direction and angle θ j .

[0075] The aforementioned system design divides LED matrix headlights into movable and fixed types. It utilizes onboard camera data of vehicles ahead, millimeter-wave radar data of distance and azimuth angle, and vehicle speed data from a speed sensor. After model training, the system outputs information on which LEDs to extinguish in the fixed matrix module and the direction and angle of rotation of the movable matrix module within its corresponding illumination area. By efficiently combining the illumination areas of the movable and fixed matrix lights, a larger light area is illuminated when vehicles meet at night. Therefore, this invention effectively solves the problems of blind spots and the visual impact on drivers caused by the sudden extinguishing of some LEDs when vehicles meet at night in current adaptive headlights. This solution can significantly improve the driver's safe driving vision when meeting oncoming traffic and pedestrians at night, enhancing driver safety performance and showing broad application prospects.

[0076] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the concept and scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the design concept of the present invention should fall within the protection scope of the present invention.

Claims

1. A smart vehicle headlight control method for dynamically changing the illumination area of ​​matrix headlights, characterized in that, Includes the following steps: 1) Data preparation: Collect a dataset of vehicles in front of the vehicle. The dataset includes image data of vehicles in front of the vehicle obtained by a binocular camera installed above the windshield of the vehicle; vehicle speed data obtained by the vehicle speed sensor; and distance and azimuth data of vehicles in front obtained by millimeter-wave radar. The headlight LED matrix module is divided into a fixed LED matrix module and a movable LED matrix module. 2) The acquired dataset of vehicles ahead is input into the Fast R-CNN convolutional neural network model for training, outputting the position information of the vehicles ahead in the image coordinate system; the distance and azimuth data of the vehicles ahead acquired by the millimeter-wave radar are transmitted to the central controller ECU via the CAN bus with the position information of the vehicles ahead output by the Fast R-CNN convolutional neural network for data fusion to obtain the coordinates of the vehicles ahead. ,length ,width Data, including coordinates It is the position coordinate of the vehicle ahead, centered on the headlights; 3) Obtain the coordinates of the vehicle ahead from step 2). ,length ,width The data is input into the RBF neural network model for prediction, and the output is the area of ​​LED beads to be turned off by the fixed LED matrix module. Rotation direction of movable LED matrix module ,angle The information and calculation process are as follows: 3.1) Arrange the matrix LEDs in two columns, each column... Each LED is numbered, and the headlights have a left-right illumination angle of 30° (15° to the left and right). Therefore, the illumination angle of each LED is... Furthermore, the angles of illumination from adjacent LED beads on the left and right sides are continuously formed. 3.2) The specific coordinates of the vehicle ahead are known. ,length ,width The coordinates of the four corners of the foremost and rearmost points of the vehicle ahead are obtained as follows: 3.3) Select The angle value of the headlight illumination is obtained from two points; the angle value corresponding to the light illuminating the area around the vehicle must be turned off. : ； 3.4) Angle from the center of the headlights to the left or right where the lights are not turned off : , The calculation begins from the angle at which the headlights remain lit to the left or right of their center. The angle of the combined light rays from adjacent LED beads, based on the light irradiation angle. Obtain the LED numbers that need to be turned off, thus determining the area of ​​LEDs to be turned off. ; 3.5) Obtain the specific coordinates of the vehicle ahead. ,length ,width The dataset is used to train the RBF neural network model. The trained RBF neural network model is then tested on the test set, and finally the area of ​​the LED beads to be turned off is predicted. Rotation direction of movable matrix module The corresponding rotation angle It is expressed as follows: in, Turn left angle; Turn right angle; The tilt angle; For the tilt angle; Turn left first Angle tilted again angle; Turn left first Turn the angle up again angle; Turn right first Angle tilted again angle; Turn right first Turn the angle up again Angle; rotation angle The range is , 4) The central controller ECU will select the areas of LED beads in the fixed LED matrix module that need to be turned off. Data is transmitted to the PWM controller via the CAN bus. The PWM controller modulates the pulse width to control the output of different high and low levels, thereby adjusting the LED bead area of ​​the fixed LED matrix module. The brightness of the LED beads; the central controller ECU transmits the direction and angle information that the movable matrix module needs to rotate to the PLC controller via the CAN bus. The PLC controller sends pulses to the servo motor to control the servo motor to drive the movable LED matrix module to rotate within the specified direction and angle; 5) Dynamically control the matrix headlights based on the vehicle's operating status, the movement of vehicles ahead, and their position information.

2. The intelligent vehicle headlight control method for dynamically changing the illumination area of ​​matrix headlights as described in claim 1, characterized in that, The control method is implemented based on the following devices: a central controller (ECU), an in-vehicle binocular camera, a millimeter-wave radar, a vehicle speed sensor, a PWM controller, a PLC controller, a servo motor, an LED matrix module driver, a movable LED matrix module, and a fixed LED matrix module. The in-vehicle binocular camera, millimeter-wave radar, and vehicle speed sensor are connected to the central controller (ECU) via a CAN bus. The in-vehicle binocular camera acquires image information of the vehicle in front, the millimeter-wave radar acquires distance and azimuth data of the vehicle in front, and the vehicle speed sensor acquires the speed data of the vehicle itself. The PWM controller and PLC controller are connected to the ECU via a CAN bus. The PWM controller and PLC controller receive information processed by the ECU and output pulse width and pulse signals, respectively. The PWM controller is connected to the fixed LED matrix module and controls the on / off state of LEDs in a specific area of ​​the fixed LED matrix module through the output pulse width signal. The PLC controller is connected to the servo motor, and the servo motor is connected to the movable LED matrix module. The servo motor controls the rotation angle of the servo motor through the output pulse signal. The movable LED matrix module rotates in a specified direction and angle under the drive of the servo motor.

3. The intelligent vehicle headlight control method for dynamically changing the illumination area of ​​matrix headlights as described in claim 1, characterized in that... In step 1), the fixed matrix module is used to adjust the brightness of the LEDs in the matrix area, and can turn off or reduce the brightness of local LED areas; the movable matrix module adjusts the rotation of the matrix module under the drive of the servo motor, bevel gear and swing arm, and changes the illumination direction and angle of the LEDs in the movable matrix module.

4. The intelligent vehicle headlight control method for dynamically changing the illumination area of ​​matrix headlights as described in claim 1, characterized in that... In step 2), the process of fusing the information about the vehicle ahead obtained by the millimeter-wave radar and the Fast R-CNN convolutional neural network in the central controller ECU includes the following: converting the radar coordinate system into an image coordinate system, mapping the two-dimensional point cloud image output by the millimeter-wave radar to the image coordinate system, stitching the mapped image with the image output by the Fast R-CNN convolutional neural network, and matching the position information of the vehicle ahead in the mapped image with the position information of the vehicle ahead in the image trained by the convolutional network; obtaining the coordinate information of the matched vehicle ahead in the radar coordinate system, and converting the coordinate information in the radar coordinate system into coordinates in the world coordinate system. Establish a pixel coordinate system, obtain the length and width information of the vehicle in front in the pixel coordinate system, and convert the length and width information of the vehicle in front in the pixel coordinate system into the length of the vehicle in the world coordinate system. ,width .

5. The intelligent vehicle headlight control method for dynamically changing the illumination area of ​​matrix headlights as described in claim 1, characterized in that... The dynamic control strategy in step 5) includes the following: When the millimeter-wave radar detects a vehicle within a range of 100-200 meters, it predicts the area of ​​LEDs that the fixed matrix module needs to extinguish. The direction of rotation of the movable matrix module and angle Simultaneously, the movable matrix module adjusts the direction and angle, and controls the LED areas where the fixed matrix module is turned off. The extinguished illumination area is compensated for to prevent light jumps during light combination and make the light continuity more uniform. Through the cooperation of movable and fixed matrix modules, the illumination angle of the lamp beads corresponding to the area of ​​light that dazzles the driver is adjusted to maximize the illumination area.

Images