A vehicle control method and device, vehicle and medium

Abstract

The application discloses a vehicle control method and device, a vehicle and a medium, and relates to the technical field of vehicle control, and specifically discloses a vehicle control method, which comprises the following steps: acquiring a wiper related parameter, a glare related parameter and a real-time video stream of a vehicle-mounted DVR camera of a vehicle; determining an image feature set according to the real-time video stream, the wiper related parameter and the glare related parameter, wherein the image feature set comprises a rainfall intensity value, a glare intensity value and a license plate area definition; and determining a wiper speed regulation instruction and a light compensation regulation instruction according to a preset expert rule base and the image feature set and performing driving control. Based on the real-time video stream of the vehicle-mounted DVR, the wiper related parameter and the glare related parameter, multi-dimensional accurate quantification of rainfall, glare and license plate definition is realized, multi-dimensional features are fused, a wiper speed regulation instruction and a light compensation regulation instruction are dynamically generated, the wiper frequency and the light compensation brightness are effectively controlled in a dynamic and adaptive manner, the driver's field of vision definition and the vehicle-mounted DVR monitoring evidence effect are simultaneously optimized in a rainy night environment, and the wiper frequency and the light compensation brightness are dynamically and adaptively controlled in a rainy night environment.

Description

Technical Field

[0001] This invention relates to the field of vehicle technology, and in particular to a vehicle control method, device, vehicle, and medium. Background Technology

[0002] In rainy night driving scenarios, rain obstruction, insufficient light, and glare interference are key factors affecting driver visibility safety and the reliability of in-vehicle monitoring evidence collection. With the rapid development of intelligent in-vehicle technology, auxiliary functions such as automatic wiper control and in-vehicle supplemental lighting adjustment are gradually becoming more widespread. However, existing technologies mostly use independent control logic, which is difficult to adapt to the complex and ever-changing environment of rainy nights.

[0003] Currently, automatic control of vehicle wipers mainly relies on optical or capacitive rain sensors, which indirectly detect rainfall through changes in infrared reflection and scattering or differences in the dielectric constant of glass electrode layers, thus achieving fixed-speed adjustment. Vehicle supplementary lighting systems mostly operate independently, using preset fixed brightness output and are not linked with the wiper system.

[0004] Existing technical solutions have significant drawbacks: Rainfall detection relies on indirect sensing via sensors, resulting in insufficient accuracy in heavy rainfall scenarios. It can only output a limited number of speed settings and cannot achieve continuous stepless speed adjustment, requiring manual intervention. Furthermore, it cannot identify glare and license plate overexposure issues. The wiper and supplementary lighting systems are independent of each other, and the supplementary lighting in rainy nights is prone to glare due to the superposition of rainwater and wiper movement, obstructing the driver's vision. This also causes the onboard digital video recorder (DVR) to overexpose the license plate of the vehicle in front, losing its monitoring and evidence collection function. Traditional solutions require additional rain and light sensors, resulting in high hardware costs, poor compatibility, and difficulty in adapting to existing onboard domain control architectures, making engineering and promotion challenging. Summary of the Invention

[0005] This invention provides a vehicle control method, device, vehicle, and medium that achieves dynamic, coordinated, and adaptive control of wiper frequency and supplementary light brightness, simultaneously optimizing the driver's visibility and the effectiveness of vehicle-mounted DVR monitoring and evidence collection in rainy night environments.

[0006] According to one aspect of the present invention, a vehicle control method is provided, comprising:

[0007] Acquire vehicle wiper parameters, glare parameters, and real-time video stream from the vehicle's DVR camera;

[0008] Based on the real-time video stream, the wiper-related parameters, and the glare-related parameters, an image feature set is determined, which includes rainfall intensity value, glare intensity value, and license plate area clarity.

[0009] Based on the preset expert rule base and the image feature set, wiper speed adjustment command and supplementary light dimming command are determined and driven.

[0010] According to a second aspect of the present invention, a vehicle control device is provided, comprising:

[0011] The parameter acquisition module is used to acquire vehicle wiper-related parameters, glare-related parameters, and real-time video stream from the vehicle-mounted DVR camera;

[0012] The feature determination module is used to determine an image feature set based on the real-time video stream, the wiper-related parameters, and the glare-related parameters. The image feature set includes rainfall intensity value, glare intensity value, and license plate area clarity.

[0013] The instruction control module is used to determine the wiper speed adjustment instruction and the fill light dimming instruction based on the preset expert rule base and the image feature set, and to perform drive control.

[0014] According to a third aspect of the present invention, a vehicle is provided, the vehicle comprising:

[0015] At least one controller; and

[0016] A memory communicatively connected to the at least one controller; wherein,

[0017] The memory stores a computer program that can be executed by the at least one controller, which enables the at least one controller to perform the vehicle control method according to any embodiment of the present invention.

[0018] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a controller to execute and implement the vehicle control method according to any embodiment of the present invention.

[0019] The technical solution of this invention achieves multi-dimensional and precise quantification of rainfall, glare, and license plate clarity based on real-time video streams from an in-vehicle DVR, wiper parameters, and glare-related parameters. It integrates multi-dimensional features and uses a pre-set expert rule base to achieve multi-objective optimization decisions, dynamically generating wiper speed adjustment and supplementary lighting dimming commands. This effectively suppresses glare in rainy nights and avoids overexposure of license plates on vehicles ahead, achieving dynamic and adaptive control of wiper frequency and supplementary lighting brightness. In rainy night environments, it simultaneously optimizes driver visibility and the effectiveness of in-vehicle DVR monitoring and evidence collection. No additional rain or light sensors are required, resulting in low hardware costs, strong compatibility, and suitability for various in-vehicle scenarios.

[0020] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a vehicle control method provided according to Embodiment 1 of the present invention;

[0023] Figure 2 This is a flowchart of determining rainfall intensity value in a vehicle control method according to Embodiment 1 of the present invention;

[0024] Figure 3 This is a flowchart of determining glare intensity value in a vehicle control method according to Embodiment 1 of the present invention;

[0025] Figure 4 This is a flowchart of determining the clarity of the license plate area in a vehicle control method according to Embodiment 1 of the present invention;

[0026] Figure 5 This is a flowchart of instruction determination in a vehicle control method according to Embodiment 1 of the present invention;

[0027] Figure 6 This is a schematic diagram of the structure of a vehicle control device according to Embodiment 2 of the present invention;

[0028] Figure 7 This is a structural schematic diagram of a vehicle that implements an embodiment of the present invention. Detailed Implementation

[0029] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0030] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0031] Example 1

[0032] Figure 1 This is a flowchart of a vehicle control method provided in Embodiment 1 of the present invention. This embodiment is applicable to the combined control of vehicle supplemental lighting and windshield wipers in rainy conditions. The method can be executed by a vehicle control device, which can be implemented in hardware and / or software and can be configured in the vehicle. Figure 1 As shown, the method includes:

[0033] S110: Acquire vehicle wiper-related parameters, glare-related parameters, and real-time video stream from the vehicle's DVR camera.

[0034] In this embodiment, "vehicle" can be understood as any motor vehicle equipped with an onboard domain controller, onboard DVR, electric windshield wipers, and adjustable onboard supplemental lighting. Wiper-related parameters refer to parameters used to dynamically define the region of interest (ROI) for windshield wiper detection, such as the current wiper arm angle, swing position, physical coverage area, and wiper motor operating status. Glare-related parameters are used to dynamically define the glare assessment ROI, and may include the onboard supplemental lighting installation location, supplemental lighting angle, driver's field of vision, and front-view camera's field of view. The onboard DVR camera refers to the vehicle's front-view dashcam camera. The real-time video stream refers to the sequence of color video images continuously output by the DVR at a fixed frame rate (e.g., 25 frames / second), containing information such as raindrops on the windshield, ambient light, and the license plate of the vehicle in front, providing raw data for subsequent image processing.

[0035] Specifically, the vehicle-mounted controller can read the vehicle's wiper operating status and supplementary lighting configuration parameters to generate wiper-related parameters and glare-related parameters, and can receive real-time video streams from the forward-looking DVR via the MIPI interface.

[0036] S120. Determine the image feature set based on the real-time video stream, wiper-related parameters, and glare-related parameters.

[0037] In this embodiment, the image feature set refers to a set of structured environmental parameters extracted and quantized from the real-time video stream, including three core features: rainfall intensity value, glare intensity value, and license plate area clarity. The rainfall intensity value can be understood as a continuous quantization value normalized to 0-1, where 0 represents no rain and 1 represents heavy rain, accurately distinguishing between light, moderate, heavy, and torrential rain, replacing the discrete levels of traditional sensors. The glare intensity value can be understood as a continuous quantization value normalized to 0-1, where 0 represents no glare and 1 represents strong glare. The license plate area clarity can be understood as a continuous score normalized to 0-1, where 0 represents completely blurred and 1 represents clearly discernible.

[0038] Specifically, the controller can dynamically define the Region of Interest (ROI) for windshield wiping using wiper-related parameters; perform grayscale conversion and 5×5 Gaussian filtering for noise reduction on the ROI image; extract raindrop edges using an adaptive Canny operator (Sobel gradient, non-maximum suppression, and adaptive dual thresholding); eliminate interference through geometric filtering of perimeter / compactness / aspect ratio and inter-frame motion consistency verification; and determine the rainfall intensity value by fusing spatial density, temporal rate of change, and historical trends, followed by nonlinear mapping and amplitude limiting. The controller can also dynamically define the driver's field of vision (ROI) using glare parameters; convert the image to a hue-saturation-value (HSV) color space and extract the luminance channel; calculate the proportion of high luminance and luminance distribution entropy to obtain luminance features; extract edges using glare-optimized Canny, and calculate edge density, directional disorder, and continuity to obtain edge features; after weighted fusion, perform Sigmoid mapping and temporal smoothing to output a continuous glare intensity value of 0–1. The controller can locate the license plate of the vehicle in front from the real-time video stream; and determine the clarity of the license plate area through image grayscale conversion, CLAHE contrast enhancement, and bilateral filtering to preserve edges and reduce noise.

[0039] S130. Based on the preset expert rule base and image feature set, determine the wiper speed adjustment command and the supplementary light dimming command and perform drive control.

[0040] In this embodiment, the preset expert rule base is a set of fuzzy inference rules built upon extensive real-vehicle testing and rainy night driving experience. It covers all scenarios including low / medium / high rain, low / medium / high glare, and low / medium / high resolution. It includes multiple rules that specify "if the rain / glare / resolution is at a certain level, then output the corresponding wiper / lighting command," used for multi-objective optimization decision-making. The wiper speed control command is a continuous control signal normalized to 0-1, where 0 is stop and 1 is maximum speed. The light dimming command is a continuous control signal normalized to 0-1, where 0 is off and 1 is brightest.

[0041] Specifically, the controller can use fuzzy logic algorithms to perform fuzzy reasoning and defuzzification processing on real-time video streams, wiper-related parameters, and glare-related parameters according to a preset expert rule base, dynamically generating wiper speed adjustment commands and supplementary light dimming commands to achieve coordinated adaptive control of wiper wiping frequency and supplementary light brightness, thereby optimizing the imaging effect of the in-vehicle DVR while ensuring clear visibility for the driver.

[0042] The technical solution of this invention achieves multi-dimensional and precise quantification of rainfall, glare, and license plate clarity based on real-time video streams from an in-vehicle DVR, wiper parameters, and glare-related parameters. It integrates multi-dimensional features and uses a pre-set expert rule base to achieve multi-objective optimization decisions, dynamically generating wiper speed adjustment and supplementary lighting dimming commands. This effectively suppresses glare in rainy nights and avoids overexposure of license plates on vehicles ahead, achieving dynamic and adaptive control of wiper frequency and supplementary lighting brightness. In rainy night environments, it simultaneously optimizes driver visibility and the effectiveness of in-vehicle DVR monitoring and evidence collection. No additional rain or light sensors are required, resulting in low hardware costs, strong compatibility, and suitability for various in-vehicle scenarios.

[0043] Furthermore, based on the above embodiments, the step of determining the image feature set according to the real-time video stream, wiper-related parameters, and glare-related parameters can be refined as follows:

[0044] Based on the real-time video stream and wiper parameters, the rainfall intensity value is determined; based on the real-time video stream and glare parameters, the glare intensity value is determined; the license plate area of ​​the front vehicle is analyzed in the real-time video stream to determine the clarity of the license plate area; the rainfall intensity value, glare intensity value, and license plate area clarity are used as the image feature set.

[0045] Specifically, the controller dynamically delineates the region of interest (ROI) for wiping based on the wiper arm position and physical coverage area. The ROI image is preprocessed with grayscale conversion and Gaussian filtering, and raindrop edges are extracted using an adaptive Canny operator. Interference is eliminated through geometric feature filtering and inter-frame motion consistency verification. Spatial density, temporal rate of change, and historical trends are fused to output continuous rainfall intensity values ​​normalized to 0–1. A glare assessment ROI is delineated based on driver field of vision characteristics. The image is converted to HSV space and the luminance channel is extracted. The proportion of high luminance and luminance distribution entropy are analyzed. Canny edge density, directional disorder, and continuity are optimized in conjunction with glare data. Weighted fusion is performed, followed by nonlinear mapping and smoothing, to output continuous glare intensity values ​​normalized to 0–1. The controller can analyze the license plate area of ​​the vehicle in front of it in real-time video stream to determine the clarity of the license plate area: Haar detection and KCF tracking are used to locate the license plate of the vehicle in front, and the license plate is preprocessed by grayscale conversion, contrast enhancement, and edge-preserving denoising; edge density, connectivity, character integrity, local contrast, and high-frequency energy indicators are extracted, weighted and fused, and then smoothed by mapping to output a license plate area clarity score normalized to 0-1. The obtained rainfall intensity value, glare intensity value, and license plate area clarity are integrated into an image feature set to provide multi-dimensional perception input for subsequent collaborative control.

[0046] Based on the above embodiments, the step of determining the rainfall intensity value according to the real-time video stream and wiper parameters can be refined as follows:

[0047] Based on the physical coverage range of the wipers in the relevant wiper parameters, the dynamic region of interest in the real-time video stream is determined and a sub-image is extracted; the sub-image is preprocessed to determine the preprocessed image; an adaptive edge detection operator is used to extract edges from the preprocessed image to determine the raindrop edge contours in the preprocessed image; the raindrop edge contours are filtered based on the raindrop geometric features and inter-frame motion consistency to determine the effective raindrop features; and the rainfall intensity value is determined based on the spatiotemporal distribution of the effective raindrop features.

[0048] In this embodiment, the physical coverage area of ​​the wiper refers to the area actually swept by the wiper on the windshield, serving as the basis for dynamically defining the detection area. The dynamic region of interest (ROI) is a rectangular area that changes in real-time with the wiper position, limiting raindrop detection. Raindrop analysis is performed only within this area, improving accuracy and reducing computational load. A sub-image refers to a local image corresponding to the dynamic ROI, extracted from the real-time video stream. The preprocessed image is the image after grayscale conversion and Gaussian filtering, providing clean input for edge extraction. The adaptive Canny operator is an edge detection algorithm that includes gradient calculation, non-maximum suppression, and adaptive dual-threshold filtering. It automatically adjusts the threshold based on image content, accurately extracting weak edges. The raindrop edge contour is a continuous set of pixels detected by the adaptive Canny operator, representing the raindrop boundary. Raindrop geometric features refer to morphological parameters such as the perimeter, area, compactness, and aspect ratio of the raindrop contour, used to distinguish raindrops from reflections, water stains, and other interference. Inter-frame motion consistency refers to the continuity of raindrop centroid displacement, motion direction, and velocity across consecutive video frames, used to filter out non-raindrop motion interference. Effective raindrop features refer to the set of edge contours confirmed as real raindrops through both geometric and motion filtering. Spatiotemporal distribution refers to the distribution density of effective raindrops in a single frame, as well as the rate of change and historical trend between consecutive frames.

[0049] Specifically, the controller can dynamically lock the windshield wiping area in each frame of video based on the physical coverage range of the wipers in the wiper-related parameters, and only extract this area as a sub-image. This sub-image is then preprocessed to determine the preprocessed image. First, the color sub-image is converted to grayscale to simplify subsequent calculations. Then, a 5×5 Gaussian kernel with a standard deviation of 1.4 is used for low-pass filtering to suppress high-frequency noise while preserving the raindrop edge structure, resulting in the preprocessed image, providing high-quality input for edge extraction. The controller can use an adaptive Canny operator to extract edges from the preprocessed image, determining the raindrop edge contours. The horizontal and vertical gradients are calculated using the Sobel operator to obtain the gradient magnitude and direction; edges are refined using non-maximum suppression; then, high and low thresholds are adaptively determined based on the gradient magnitude histogram, and a double-threshold hysteresis connection is used to completely extract the raindrop edge contours, accurately capturing raindrop boundary information. Based on the raindrop geometric features and inter-frame motion consistency, the raindrop edge contours are filtered to determine effective raindrop features. The system calculates the perimeter, area, compactness, and aspect ratio of the raindrop profile to eliminate morphological anomalies. It then matches the profiles across consecutive frames to verify that the centroid displacement, direction of motion, and velocity conform to the vertical falling pattern of raindrops, filtering out false profiles such as reflections, water stains, and wiper shadows. Finally, it selects the true and reliable effective raindrop features. The controller can determine the rainfall intensity value based on the spatiotemporal distribution of the effective raindrop features. It statistically analyzes the spatial distribution density of effective raindrops in a single frame, calculates the inter-frame temporal change rate, and combines this with historical trend weighted fusion. After nonlinear mapping and amplitude limiting, it outputs a continuous, stepless rainfall intensity value in the 0-1 range, achieving refined quantification from light rain to heavy rain, providing a precise basis for adaptive wiper speed adjustment.

[0050] For example, a specific example can be used to illustrate how rainfall intensity values ​​are determined. Figure 2 A flowchart for determining rainfall intensity values ​​in a vehicle control method is provided in Embodiment 1 of the present invention, as shown below. Figure 2 As shown, the specific steps for determining the rainfall intensity value are as follows: 1. Dynamic ROI extraction: The controller can define a dynamic rectangular region as the region of interest (ROI) in real time in the current video frame F collected by the vehicle DVR based on the current wiper arm position and its physical coverage area, and extract the sub-image I_ROI of the region from F.

[0051] 2. Grayscale Conversion: If I_ROI is a three-channel color image, convert it to a single-channel grayscale image I_gray; if I_ROI is already a single-channel image, directly copy it as I_gray. The grayscale conversion is calculated pixel-by-pixel using the following formula:

[0052] I_gray(x,y)=0.299×R(x,y)+0.587×G(x,y)+0.114×B(x,y)

[0053] In the formula, R(x,y), G(x,y), and B(x,y) are the red, green, and blue channel brightness values ​​of I_ROI at pixel coordinates (x,y), respectively.

[0054] 3. Gaussian Filtering Smoothing: A two-dimensional Gaussian kernel with a size of 5×5 pixels and a standard deviation σ=1.4 is used to perform low-pass smoothing filtering on the grayscale image I_gray, resulting in the filtered image I_blur, to suppress high-frequency noise and preserve the raindrop edge structure. The two-dimensional Gaussian kernel function G(x,y) and the convolution operation are defined as follows:

[0055] G(x,y)=1 / (2×π×σ^2)×exp(-(x^2+y^2) / (2×σ^2))

[0056] I_blur(x,y)=Σ_{i=-2}^{2}Σ_{j=-2}^{2}I_gray(x+i,y+j)×G(i,j)

[0057] Where (x,y) takes the kernel center as the origin, σ=1.4, and the summation range covers a 5×5 neighborhood.

[0058] 4. Edge Feature Extraction Based on Adaptive Canny Operator: Gradient Calculation. For the Gaussian-filtered grayscale image I_blur, the horizontal gradient Gx and vertical gradient Gy are calculated using the central difference Sobel operator. The Sobel kernel size used is 3×3. The specific calculation is as follows:

[0059] Gx(x,y)=[I_blur(x+1,y-1)+2×I_blur(x+1,y)+I_blur(x+1,y+1)]-

[0060] [I_blur(x-1,y-1)+2×I_blur(x-1,y)+I_blur(x-1,y+1)]

[0061] Gy(x,y)=[I_blur(x-1,y+1)+2×I_blur(x,y+1)+I_blur(x+1,y+1)]-

[0062] [I_blur(x-1,y-1)+2×I_blur(x,y-1)+I_blur(x+1,y-1)]

[0063] Then, the gradient magnitude M(x,y) and gradient direction θ(x,y) of each pixel are calculated using the following formula (in degrees):

[0064] M(x,y)=sqrt(Gx(x,y)^2+Gy(x,y)^2)

[0065] θ(x,y)=arctan2(Gy(x,y),Gx(x,y))

[0066] Perform non-maximum suppression on the gradient magnitude M(x,y) in the neighborhood direction corresponding to the gradient direction θ(x,y). If the magnitude of the current point is not the local maximum in its gradient direction, set it to zero; otherwise, keep it, and obtain the refined edge magnitude image M_supp. Specifically, quantize the gradient direction into four main directions: 0°, 45°, 90°, and 135°, and compare within the corresponding 3×1 or 3×3 neighborhood, only keeping the local maximum. Adaptive double-threshold hysteresis connection, adopt an adaptive strategy based on the gradient magnitude histogram to determine the high threshold T_high and the low threshold T_low. First, calculate the cumulative distribution of the gradient magnitudes of non-zero pixels in M_supp, and take the magnitude at the point where the cumulative distribution reaches a preset percentage (such as 70%-80%) as T_high, and take T_low = k×T_high, where k is a preset proportionality coefficient (such as 0.4). Subsequently, divide the pixels into strong edge points (M≥T_high), weak edge points (T_low≤M<T_high), and non-edge points (M<T_low) according to the double thresholds, and incorporate the weak edge points adjacent to strong edge points into the final edge image E through hysteresis connection to achieve the complete extraction of the edge contour.

[0067] 5. Modeling and screening of raindrop morphological and kinematic characteristics include: Morphological opening operation for denoising, perform morphological opening operation on the edge image E using a 3×3 elliptical structuring element to obtain the denoised edge image E_clean. The opening operation first erodes and then dilates, which can effectively isolate small noise points and non-raindrop edge fragments, while keeping the edge structure of the raindrop main body basically unchanged. Adopt a contour extraction algorithm to retrieve all outer contours from E_clean, only keep the outermost contour level, and store the sequence of contour points {C_i} in the form of a simple chain approximation, where i is the contour index. Each contour C_i is represented by a set of ordered boundary point coordinates. Raindrop contour screening based on geometric features, for each contour C_i, calculate its geometric feature parameters, and perform constraint screening according to the raindrop morphology prior model: Calculate the contour perimeter P_i and area A_i, obtained using the cumulative distance of the contour point sequence and the contour polygon area formula. Calculate the compactness C_i of the contour, which is used to measure the shape roundness:

[0068] C_i = P_i^2 / (4×π×A_i)

[0069] Normal raindrops are oblate due to the air resistance during falling, and the projection is approximately elliptical. Its compactness should be within the preset interval [C_min, C_max], with typical values such as C_min = 1.0 and C_max = 1.8.

[0070] Calculate the minimum bounding rectangle of the profile to obtain the aspect ratio AR_i = max(width, height) / min(width, height), reflecting the elongation of the shape. Raindrop aspect ratios are usually close to 1, but can be relaxed depending on the falling posture, retaining profiles that satisfy AR_min ≤ AR_i ≤ AR_max. Additional constraints: Set perimeter threshold intervals [P_min, P_max] and area threshold intervals [A_min, A_max] to exclude excessively large or small interference objects (such as wiper coverage areas, large puddles reflecting light, etc.). Only profiles that simultaneously satisfy all the above geometric conditions are identified as candidate raindrop profiles and drawn onto a blank image E_filtered of the same size as E, with non-profile areas set to zero.

[0071] Based on motion consistency, inter-frame verification further eliminates false raindrops by utilizing motion information between consecutive video frames. The candidate raindrop contour sets of the current frame t and the previous frame t-1 are matched. By calculating the centroid displacement and shape similarity of the contours, it is determined whether the motion conforms to the characteristic of raindrops falling vertically at a near uniform speed. Specifically: (1) Calculate the centroid coordinates of each contour and find corresponding contour pairs between frames by data association based on spatial proximity and similarity. (2) For the corresponding contour pairs that are successfully matched, calculate the displacement vector d_i=(Δx,Δy). It is expected that the direction of raindrop falling is basically along the approximately vertical direction of the image, and the displacement magnitude Δy matches the raindrop falling speed and frame rate, and Δx fluctuates very little. Set the constraint conditions: Δy∈[v_min×Δt,v_max×Δt], and Δx<ε_x, where Δt is the frame interval, v_min and v_max are the preset raindrop falling speed ranges, and ε_x is the horizontal shaking tolerance threshold. (3) Only retain the contours with continuous motion trajectories that satisfy the above motion constraints, and remove the false contours with inconsistent motion from E_filtered to obtain the final reliable raindrop edge image E_final, which is used as the input feature for subsequent rainfall intensity quantization.

[0072] 6. Multi-dimensional feature fusion and rainfall intensity quantization output: Single-frame spatial distribution density calculation: Within the dynamic ROI region, count the number of non-zero pixels N_rain in the final raindrop edge image E_final, and calculate its ratio to the total ROI area A_ROI to obtain the spatial distribution density D_s(t) of the current frame.

[0073] D_s(t)=N_rain(t) / A_ROI

[0074] Where A_ROI = W_ROI × H_ROI, W_ROI and H_ROI are the width and height of the ROI (in pixels), respectively.

[0075] The inter-frame temporal change rate is calculated based on a spatial density sequence of multiple consecutive frames. The temporal change rate R_t(t) of the current frame t is calculated to reflect the instantaneous trend of rainfall intensity. The central difference method is used, utilizing the spatial density values ​​of the previous frame t-1 and the current frame t:

[0076] R_t(t) = D_s(t) - D_s(t-1)

[0077] To enhance noise immunity, a longer sliding time window (such as 3 or 5 frames) can be used for least-squares linear fitting, with the slope of the fitted line as the rate of change over time.

[0078] Historical trend moving average, applied exponentially to the spatial density series, yields the historical trend component H(t), reflecting the overall level of rainfall intensity over a period of time. The recursive calculation formula is as follows:

[0079] H(t)=α×D_s(t)+(1-α)×H(t-1)

[0080] In the formula, α is the smoothing coefficient, with a value range of 0 < α < 1, and a typical value is α = 0.2. The initial value H(0) can be the spatial density value of the first frame or a preset empirical value.

[0081] Multi-feature weighted fusion is used to linearly weight and fuse the spatial distribution density D_s(t), the temporal rate of change R_t(t), and the historical trend H(t) to obtain the raw value of the comprehensive rainfall intensity I_raw(t):

[0082] I_raw(t)=α1×D_s(t)+β1×R_t(t)+γ1×H(t)

[0083] Wherein, α1, β1, and γ1 are pre-calibrated fusion weight coefficients, satisfying α1 + β1 + γ1 = 1. The weight coefficients are determined based on measured data through an optimization algorithm to maximize the correlation between I_raw(t) and the actual rainfall sensor readings.

[0084] Nonlinear mapping applies a nonlinear mapping function f_nl to I_raw(t) to compensate for the nonlinear relationship between the original signal and the actual rainfall perception. The mapping function is defined as follows:

[0085] I_mapped(t)=f_nl(I_raw(t))

[0086] Specifically, piecewise linear mapping, the sigmoid function, or a pre-calibrated lookup table mapping can be used. One alternative is a power function mapping:

[0087] I_mapped(t) = I_raw(t)^η

[0088] Wherein, η>0 is the mapping exponent. When η>1, it enhances the sensitivity of the light rainfall range; when 0<η<1, it improves the distinguishability of the heavy rainfall range.

[0089] Limit the output, and limit the mapped rainfall intensity value to the normalized interval [0,1] to obtain the final quantized rainfall intensity value I_final(t):

[0090] I_final(t)=clamp(I_mapped(t),0,1)

[0091] That is, if I_mapped(t)>1, then take 1; if I_mapped(t)<0, then take 0; otherwise, remain unchanged. I_final(t) is output as a continuous quantized value of rainfall intensity, with a value range from 0 (no rain) to 1 (maximum rainfall), which is used to subsequently drive the adaptive and coordinated control of the wiper frequency and the brightness of the supplementary light.

[0092] Based on the above embodiments, the step of determining the glare intensity value according to the real-time video stream and glare-related parameters can be refined as follows:

[0093] Based on the driver's field of vision in the glare-related parameters, the region of interest for glare assessment in the real-time video stream is determined; the region of interest for glare assessment is converted to the HSV color space and the luminance channel is extracted to determine the luminance distribution image; based on the luminance distribution image and the proportion of high luminance, the luminance glare index is determined; edge feature analysis is performed on the luminance distribution image to determine the edge glare index; based on the luminance glare index and the edge glare index, the glare intensity parameter is determined.

[0094] In this embodiment, the driver's field of vision refers to the core visual area that the driver can clearly observe ahead while the vehicle is in motion, serving as the benchmark for defining the glare assessment range. The region of interest for glare assessment refers to a local area of ​​a video frame specifically used for glare quantification analysis, defined based on the driver's field of vision. The HSV color space decomposes an image into three independent channels: hue (H), saturation (S), and luminance (V), facilitating separate analysis of luminance information and eliminating color interference. The luminance channel refers to the V channel in the HSV space, directly reflecting the brightness of each pixel in the image. The luminance distribution image is a grayscale image obtained after extracting the luminance channel, visually presenting the luminance distribution characteristics of the image. The high luminance percentage refers to the proportion of pixels in the luminance distribution image whose luminance value exceeds a preset threshold (e.g., 200), characterizing the degree of overexposure. The luminance glare index is a quantized value calculated based on the luminance distribution and the high luminance percentage. Edge feature analysis refers to extracting edge features from the luminance distribution image using an optimized Canny operator. The edge glare index is a quantized value obtained by fusing edge density, directional disorder, and edge continuity, reflecting the degree of edge distortion and scattering caused by glare.

[0095] Specifically, the controller can dynamically delineate a rectangular evaluation area in the real-time video frame based on the driver's clear view of the core field of vision ahead. It focuses on the key field of vision range where glare is formed by raindrop scattering of supplementary light, eliminating interference from non-core areas and improving the accuracy of glare assessment. By separating brightness and color information through color space conversion, only the brightness channel reflecting the brightness of the image is retained, resulting in a brightness distribution image. This eliminates color interference and provides clean data for glare brightness characteristic analysis. A brightness threshold is set to filter high-brightness pixels, calculate the proportion of high-brightness pixels, and combine it with brightness distribution entropy to quantify the overexposure range and brightness concentration of the image, comprehensively obtaining a brightness glare index to characterize the degree of overexposure caused by glare. The Canny operator, optimized for glare, is used to extract edges, calculate edge density, directional disorder, and edge continuity, quantifying the degree of edge density, directional disorder, and edge fragmentation caused by glare, and fusing them to obtain an edge glare index to characterize the edge scattering distortion caused by glare. By weighted fusion of brightness glare index and edge glare index, and through Sigmoid nonlinear mapping and time smoothing filtering, a continuous glare intensity parameter normalized to 0 to 1 is output, which accurately quantifies the glare interference level in rainy nights and provides a core basis for adaptive dimming of supplementary lighting.

[0096] For example, a specific example can be used to illustrate how glare intensity values ​​are determined. Figure 3 A flowchart for determining glare intensity values ​​in a vehicle control method is provided in Embodiment 1 of the present invention, as follows: Figure 3 As shown, the steps for glare intensity value can be as follows: 1. Dynamic delineation of ROI for glare assessment: Based on the installation position of the vehicle-mounted front-view camera and the characteristics of the driver's field of vision, a rectangular sub-region representing the core area of ​​the driver's field of vision is dynamically delineated in the current video frame as the region of interest (GlareROI) for glare assessment, and the sub-image I_glare of this region is extracted from the original image F.

[0097] 2. RGB to HSV color space conversion: Convert I_glare from the RGB color space to the HSV color space to obtain the HSV image I_hsv, so as to separate luminance and color information. The conversion maps the red, green, and blue channel values ​​of each pixel into three independent channels: hue (H), saturation (S), and lightness (V).

[0098] 3. V channel extraction: The third channel, namely the V channel (luminance component), is separated from I_hsv and used as the basis image I_V for glare analysis. This channel directly reflects the luminance distribution of the scene and is the input for subsequent glare feature extraction.

[0099] 4. Brightness Feature Analysis: Calculation of the proportion of high-brightness pixels. For the V channel image I_V, a brightness threshold T_bright=200 is set, and it is binarized to generate high-brightness region points M_high.

[0100] M_high(x,y)=1,ifI_V(x,y)≥200

[0101] M_high(x,y)=0,ifI_V(x,y)<200

[0102] Calculate the total number of pixels in I_V, N_total = W × H, and the number of non-zero pixels in M_high, N_high. Calculate the overexposed area ratio, R_over.

[0103] R_over=N_high / N_total

[0104] The value of R_over ranges from [0,1]. The larger the value, the more overly bright areas there are in the image, and the more severe the potential glare.

[0105] 5. Calculation of brightness distribution entropy

[0106] Calculate the normalized grayscale histogram of I_V to obtain the probability of occurrence of each grayscale level k (0≤k≤255) p(k)=h(k) / N_total, where h(k) is the number of pixels with grayscale level k. Then calculate the luminance distribution entropy H:

[0107] H = -Σ_{k=0}^{255}p(k)×log2(p(k)), where it is specified that if p(k)=0, then this term is 0.

[0108] Normalize the entropy value to the [0,1] interval to obtain the normalized entropy H_norm:

[0109] H_norm=H / 8

[0110] The maximum entropy for a uniform distribution of 256 gray levels is log2(256) = 8. A lower H_norm indicates a more concentrated brightness distribution; H_norm approaching 1 indicates a uniform brightness distribution. When R_over is high and H_norm is low, strong glare is considered to be present.

[0111] 6. Edge Feature Extraction and Analysis: Glare-Optimized Canny Edge Detection. The Canny edge detection algorithm, optimized for glare scenes, is applied to the V-channel image I_V. A low threshold T_low=60, a high threshold T_high=150, and a Sobel operator aperture of 3 are set to extract high-contrast edges, resulting in a binary edge image E_glare. Edge pixels are set to 1, and non-edge pixels are set to 0. Edge density calculation: The number of all non-zero edge pixels N_edge in E_glare is counted, and the edge density D_edge per unit area is calculated.

[0112] D_edge = N_edge / N_total

[0113] Where N_total=W×H is the total number of pixels within the ROI for glare assessment. D_edge reflects the density of halos or scattering edges caused by glare; the higher the value, the more significant the glare interference.

[0114] Edge orientation uniformity is determined by calculating the gradient direction θ(x,y) (in degrees) of each pixel based on the luminance channel I_V, focusing only on pixels marked as edges in E_glare. The gradient direction is discretized into eight intervals within the range of 0° to 180°, each interval spanning 45°, and an edge orientation histogram h_dir(k), k=0,1,...,7, is constructed. The number of edge pixels falling into each interval is counted, and the orientation probability distribution is normalized.

[0115] p_dir(k)=h_dir(k) / Σh_dir(k), when Σh_dir(k)=0, p_dir(k)=0.

[0116] Calculate the edge direction distribution entropy H_dir:

[0117] H_dir=-Σ_{k=0}^{7}p_dir(k)×log2(p_dir(k)), with the convention that 0×log2(0)=0.

[0118] Normalizing the entropy value to the [0,1] interval, we obtain the disorder degree U_dir in the edge direction:

[0119] U_dir=H_dir / 3

[0120] The maximum entropy of the uniform distribution in 8 directions is log2(8)=3. The closer U_dir is to 1, the more chaotic and disordered the edge directions are, which is consistent with the directional distribution characteristics caused by glare.

[0121] Edge continuity analysis involves contour extraction from the edge image E_glare, yielding a series of connected contours {C_j}, j=1,...,N_c, where N_c is the total number of contours. The pixel length l_j of each contour (i.e., the pixel count of the contour point sequence) is calculated, and the average contour length L_avg is obtained.

[0122] L_avg=(1 / N_c)×Σ_{j=1}^{N_c}l_j, if N_c=0 then set L_avg=0.

[0123] Glare often causes edge breakage and fragmentation, reducing its continuity, which is reflected in a smaller average profile length L_avg. L_avg can be used directly as a continuity indicator, or it can be normalized to a continuity score C_cont using a preset reference length L_ref.

[0124] C_cont=clamp(L_avg / L_ref,0,1)

[0125] L_ref is calibrated based on the average edge length under normal lighting conditions. The lower C_cont is, the higher the degree of edge fragmentation and the more severe the glare interference.

[0126] 7. Edge continuity quantization and contour extraction: Perform binary image contour detection on the glare edge image E_glare, retrieving only the outermost contour and storing the contour point sequence in a simple compressed form to obtain the contour set {C_j}, j=1,2,…,N_c, where N_c is the total number of detected contours. Calculate the average contour length: if N_c>0, count the number of pixels in the point sequence for each contour as the contour length l_j, and calculate the average length L_avg.

[0127] L_avg=(1 / N_c)×Σ_{j=1}^{N_c}l_j

[0128] If N_c=0 (no edges detected), then set L_avg=0.

[0129] Continuity normalization involves comparing the average contour length L_avg with a preset reference length L_ref=100 and limiting it to [0,1] to obtain the edge continuity index C_cont.

[0130] C_cont=min(L_avg / 100,1)

[0131] A higher value for C_cont indicates a longer average edge length and better continuity; a lower value indicates a higher degree of edge fragmentation, which is consistent with the characteristics of glare causing edge breakage and discontinuity.

[0132] 8. Multi-feature fusion and glare intensity quantization: Brightness and edge feature weighted fusion is performed. The brightness feature score S_lum and the edge feature score S_edge are linearly weighted according to preset weights to obtain the initial glare value G_raw.

[0133] G_raw=w_lum×S_lum+w_edge×S_edge

[0134] In the formula, w_lum and w_edge are the fusion weights, which satisfy w_lum+w_edge=1, with typical values ​​of w_lum=0.6 and w_edge=0.4.

[0135] Sigmoid nonlinear mapping: The Sigmoid function is used to perform nonlinear mapping on G_raw to enhance the distinguishability of mid-range glare intensity values, resulting in the mapped glare value G_map.

[0136] G_map=1 / (1+exp(-k×(G_raw-0.5)))

[0137] Where k is the kurtosis coefficient, typically k=5.0. This mapping compresses the input to the (0,1) interval and provides the maximum slope near G_raw=0.5.

[0138] Temporal smoothing filtering is applied to G_map to suppress fluctuations in single-frame estimation and avoid abrupt changes in control commands. If the historical frames are less than L frames, the existing elements are averaged. A historical queue Q of length L=5 frames is maintained. Each time a frame is updated, G_map is added to the tail of the queue, and the old value at the head of the queue is removed. The arithmetic mean of all elements in the queue is taken as the smoothed glare intensity value G_smooth(t) for the current frame.

[0139] G_smooth(t)=(1 / L)×Σ_{i=t-L+1}^{t}G_map(i)

[0140] The glare intensity value is output, with the G_smooth(t) value serving as the final glare interference level signal. The value range is [0,1], where 0 indicates no glare and 1 indicates the strongest glare interference. This continuous, stepless glare intensity value signal is used for subsequent coordinated control of the supplementary lighting brightness and wiper operation strategy.

[0141] Based on the above embodiments, the step of analyzing the license plate area of ​​the real-time video stream to determine the clarity of the license plate area can be refined as follows:

[0142] Based on the detection and tracking strategy, the license plate region of the vehicle in front is analyzed in the real-time video stream to determine the license plate region of the vehicle in front; the license plate region is preprocessed to determine the preprocessed license plate image; the character edges in the preprocessed license plate image are extracted and multi-dimensional sharpness index is calculated to determine the sharpness of the license plate region.

[0143] In this embodiment, the detection and tracking strategy refers to a joint processing method combining target detection and target tracking algorithms. For example, a Haar cascade classifier can be used for multi-scale license plate detection, while a KCF tracker is introduced to perform continuous frame-by-frame stable tracking of the detected license plates, balancing detection accuracy and tracking stability. The license plate region refers to the rectangular region detected in the video frame that belongs to the license plate of the vehicle ahead; the candidate box with the largest area is selected as the main license plate region. The preprocessed license plate image refers to the license plate image after grayscale conversion, contrast enhancement, and edge-preserving denoising, serving as input for edge extraction and sharpness calculation. Character edges refer to the contour boundaries of the license plate characters, extracted using the adaptive Canny operator, and are a key feature for judging sharpness. Multi-dimensional sharpness indicators include edge density, edge connectivity, character integrity, local contrast, and high-frequency energy, comprehensively reflecting the license plate imaging quality from different dimensions.

[0144] Specifically, the controller can use a pre-trained Haar cascade classifier to perform multi-scale target detection on video frames, outputting candidate license plate rectangles; select the candidate box with the largest area as the initial license plate region; introduce a KCF tracker to continuously track the initial license plate region frame by frame to prevent missed detections and drift, and stably output the license plate region of the vehicle ahead. The color image of the license plate region is converted into a grayscale image; the CLAHE algorithm is used to enhance local contrast and strengthen character edge details; then, bilateral filtering is used to smooth noise and preserve edge sharpness, finally obtaining the pre-processed license plate image. The controller can use an adaptive Canny operator to extract character edge contours; calculate five sharpness indicators in sequence: edge density, edge connectivity, character integrity, local contrast, and high-frequency energy; after weighted fusion of each indicator, it is processed by Sigmoid nonlinear mapping and time smoothing to output the license plate region sharpness normalized to 0-1, providing a basis for adaptive dimming of supplementary lighting.

[0145] For example, a specific example can be used to demonstrate how to determine the clarity of the license plate area. Figure 4 This invention provides a flowchart for determining the clarity of the license plate area in a vehicle control method, as shown in Embodiment 1. Figure 4 As shown, the license plate region sharpness can be determined through the following steps: 1. License plate detection and tracking: License plate detection based on Haar cascade. On the current video frame, a pre-trained Haar cascade classifier is used for multi-scale object detection to locate the license plate region of the vehicle ahead. The detection parameters are set as follows: scaling factor 1.1, minimum neighborhood number 3, and minimum search window size 80×25 pixels. The detection result is a set of candidate license plate rectangles {R_i}, where each R_i is defined by the top-left corner coordinates (x_i, y_i), width w_i, and height h_i. Main license plate region filtering: If the detection result {R_i} is not empty, the rectangle with the largest area is selected as the main license plate region R_plate of the current frame.

[0146] A_i=w_i×h_i

[0147] R_plate = argmax_{R_i}(A_i)

[0148] If the detection result is empty, there is no valid license plate area. Record R_plate=∅ and directly output the default clarity score of 0.2, which represents the extremely low clarity state that the license plate image cannot be obtained at present.

[0149] Tracking assistance and region stabilization (optional): A KCF (Kernelized Correlation Filter) tracker is introduced between consecutive frames to continuously track and update the detected license plate region R_plate to avoid region loss caused by missed detection in a single frame. When the detector fails intermittently, the region positioning is maintained based on the tracking results, thereby stabilizing the output of the tracked license plate region R_track.

[0150] 2. License Plate Image Preprocessing: Grayscale Conversion. The extracted license plate area image R_plate is converted from a color image to a single-channel grayscale image I_gray. If R_plate is a three-channel BGR image, the grayscale value is calculated pixel by pixel using the following formula:

[0151] I_gray(x,y)=0.299×B(x,y)+0.587×G(x,y)+0.114×R(x,y)

[0152] This is to ensure grayscale consistency with subsequent edge detection and feature calculation.

[0153] CLAHE contrast enhancement: The grayscale image I_gray is locally enhanced using Limiting Contrast Adaptive Histogram Equalization (CLAHE). A cropping threshold of 2.0 and a local block size of 8×8 pixels are set, outputting the enhanced image I_enhanced. This process effectively enhances the contrast of license plate character edges while suppressing excessive amplification of local noise. Bilateral filtering edge-preserving denoising: A bilateral filter is applied to the enhanced image I_enhanced for smoothing. The filter kernel diameter d = 5 pixels, spatial domain standard deviation σ_s = 50, and grayscale domain standard deviation σ_r = 50. Bilateral filtering smooths random noise in uniform regions while preserving the sharpness of license plate character edges, resulting in the preprocessed license plate image I_denoised, which serves as input for subsequent edge extraction.

[0154] 3. Sharpness Index Calculation: Edge density calculation is performed based on the binary edge image E_char obtained from adaptive Canny edge detection. The number of non-zero edge pixels N_edge is counted, and its ratio to the total number of pixels in the license plate area N_total = W × H is calculated to obtain the edge density D.

[0155] D = N_edge / N_total

[0156] D reflects the density of character edges; clear images typically have a high edge density.

[0157] Edge connectivity analysis involves performing eight-neighbor connected component analysis on E_char, filtering out tiny fragmented regions with a pixel area ≤10, and then calculating the total area A_total and the number K of the remaining effective connected regions. If K>0, the average area A_avg = A_total / K of the effective connected regions is calculated, and the edge connectivity index C_conn is obtained through normalization.

[0158] C_conn=min(A_avg / 50,1)

[0159] If K=0, then C_conn=0. The larger this index is, the more complete the connected components the edges tend to form, and the more complete the character structure.

[0160] Character integrity assessment involves binarizing the preprocessed license plate image I_denoised using Otsu's method to obtain a binary image B. The mean sequence of its horizontal projections along the vertical direction, P(y), where y = 1, ..., H, is then calculated. Local maxima peaks on P(y) that simultaneously satisfy the following conditions are detected:

[0161] (a) P(y)>P(y-1) and P(y)>P(y+1);

[0162] (b) P(y)>0.3 (the normalized threshold).

[0163] The number of peak values ​​N_peaks is counted and compared with the reference value of 8 for the standard number of character rows in a license plate to calculate the character integrity index C_int.

[0164] C_int=min(N_peaks / 8,1)

[0165] The closer C_int is to 1, the more complete the character arrangement.

[0166] Local contrast calculation involves calculating the global grayscale standard deviation σ_gray of the preprocessed license plate image I_denoised and comparing it with the reference value 128 to obtain the local contrast index C_contrast.

[0167] C_contrast=min(σ_gray / 128,1)

[0168] The higher this index, the greater the grayscale difference between the characters and the background in the image, and the better the contrast.

[0169] High-frequency energy calculation involves applying a Laplacian filter to I_denoised to obtain the high-frequency response map L. The global standard deviation σ_lap of L is calculated and compared with a reference value of 100 to obtain the high-frequency energy index H_freq.

[0170] H_freq=min(σ_lap / 100,1)

[0171] High-frequency energy reflects the sharpness of edges and the richness of detail in an image; sharp images typically have a higher Laplacian standard deviation.

[0172] 4. Multi-feature fusion and sharpness score quantification: Multi-index weighted fusion is performed, and the five sharpness indices obtained in step S12—edge density D, connectivity C_conn, character integrity C_int, local contrast C_contrast, and high-frequency energy H_freq—are linearly weighted according to a preset weight vector to obtain the original sharpness comprehensive value S_raw.

[0173] S_raw=w1×D+w2×C_conn+w3×C_int+w4×C_contrast+w5×H_freq

[0174] The weights wi satisfy Σwi=1, and the typical weight configuration is w1=0.20, w2=0.30, w3=0.25, w4=0.15, w5=0.10. The weights can be adjusted through experimental calibration according to the actual application scenario.

[0175] Sigmoid nonlinear mapping enhancement uses the Sigmoid function to perform a nonlinear transformation on S_raw to enhance the discrimination in the medium resolution range, resulting in a resolution score S_map after mapping.

[0176] S_map=1 / (1+exp(-k×(S_raw-0.5)))

[0177] The kurtosis coefficient k is set to 4.0. This mapping smoothly compresses the score to the (0,1) interval and provides higher sensitivity near S_raw=0.5, which is more in line with the human eye's perception of changes in sharpness.

[0178] The output (optionally time-smoothed), S_map, is the sharpness score of the license plate image in the current frame, with a value range of [0,1]. In practical applications, S_map can be further filtered using a time-moving average filter similar to glare intensity quantization to obtain a smooth score S_final, thus avoiding abrupt changes in single-frame fluctuations that could affect subsequent control decisions. S_final will be output to the collaborative decision-making module for dynamically adjusting control parameters such as supplementary lighting brightness.

[0179] Furthermore, based on the above embodiments, the steps of determining the wiper speed control command and the supplementary light dimming command according to the preset expert rule base and image feature set, and performing drive control, can be refined as follows:

[0180] The image feature set is fused to determine the fuzzy logic input parameters; based on the preset expert rule base, multi-objective optimization decision is made on the fuzzy logic input parameters to determine the wiper speed control command and the supplementary light dimming command; the wiper speed control command is converted into a wiper frequency signal to determine the wiper drive signal; the supplementary light dimming command is converted into a supplementary light brightness adjustment signal to determine the supplementary light drive signal; and drive control is performed through the wiper drive signal and the supplementary light drive signal.

[0181] In this embodiment, the fuzzy logic input parameters refer to the standardized parameters obtained after normalizing and weighted fusion of the image feature set, adapting to the needs of fuzzy inference calculation. The wiper frequency signal is the electrical signal converted from the speed control command, matching the wiper motor drive requirements to achieve continuous stepless speed regulation. The wiper drive signal is the final control signal that drives the wiper motor, directly controlling the wiper's start / stop and wiping frequency. The supplementary light brightness adjustment signal is the electrical signal converted from the dimming command, matching the supplementary light LED drive requirements to achieve continuous stepless dimming. The supplementary light drive signal is the control signal that directly controls the supplementary light switch and brightness level.

[0182] Specifically, the controller normalizes rainfall intensity, glare intensity, and license plate clarity, and integrates multi-dimensional features through weighted fusion to obtain fuzzy logic input parameters suitable for fuzzy inference. The controller performs fuzzy inference on these input parameters based on an expert rule base, defuzzifies the output with a weighted average, and then outputs an initial command. After time-smoothing filtering to suppress signal jitter, it finally outputs continuous, stepless wiper speed control and supplementary light dimming commands within the 0-1 range, balancing multiple objectives such as rain intensity adaptation, glare suppression, and license plate clarity. The normalized speed control command is mapped to a frequency control electrical signal recognizable by the wiper motor, matching the motor voltage and current drive parameters to output a stable wiper drive signal, achieving continuous, stepless adjustment of the wiper frequency. The normalized dimming command is mapped to a PWM dimming signal recognizable by the supplementary light LED, matching the LED driver chip parameters to output a stable supplementary light drive signal, achieving continuous, stepless adjustment of the supplementary light brightness. Both drive signals are sent to the wiper motor and supplementary light drive module via the vehicle's CAN bus to adjust the wiper frequency and supplementary light brightness in real time.

[0183] Based on the above embodiments, the steps of fusing image feature sets to determine fuzzy logic input parameters can be refined as follows:

[0184] The rainfall intensity, glare intensity, and license plate area clarity are normalized to determine normalized feature values. The normalized feature values ​​are then mapped based on a preset triangular membership function to determine the fuzzy membership degree of each input quantity. The fuzzy membership degrees are then combined to obtain the fuzzy logic input parameters.

[0185] In this embodiment, a preset triangular membership function is used to map precise numerical values ​​to a piecewise linear function of fuzzy sets (low / medium / high). The fuzzy membership degree is the degree of membership of the precise value to the "low / medium / high" fuzzy set, and takes a value from 0 to 1, with a higher value indicating a higher degree of matching.

[0186] Specifically, the controller can uniformly map the quantified rainfall intensity, glare intensity, and license plate area clarity to a standard range of 0-1, eliminating dimensional differences between different physical quantities, ensuring consistent numerical ranges and unified calculation benchmarks for the three, and obtaining three normalized feature values ​​to provide standard input for subsequent fuzzy mapping. For each normalized feature value, a preset "low, medium, and high" triangular membership function is matched to calculate its membership degree in the three fuzzy sets; through piecewise linear calculation, the degree to which the feature value belongs to each fuzzy level is accurately characterized, outputting three sets of fuzzy membership degrees corresponding to rainfall, glare, and clarity. The controller can concatenate the three fuzzy membership degrees of rainfall, glare, and clarity according to their dimensions to form a structured fuzzy input vector, i.e., fuzzy logic input parameters, providing a complete and standardized input basis for subsequent fuzzy rule reasoning.

[0187] Based on the above embodiments, the steps for determining the wiper speed control command and the supplementary light dimming command by performing multi-objective optimization decision-making on fuzzy logic input parameters based on a preset expert rule base can be refined as follows:

[0188] Based on the preset expert rule base, the activation intensity of each fuzzy rule is determined; based on the weighted average result of the activation intensity, the initial control command is determined, which includes the initial wiper command and the initial fill light command; the initial control command is smoothed, filtered and limited to determine the wiper speed adjustment command and the fill light dimming command.

[0189] In this embodiment, fuzzy rules refer to control rules described in fuzzy language in the rule base, in the form of "if rainfall is A, glare is B, and clarity is C, then the wiper command is w and the fill light command is l". Activation intensity can be understood as the degree to which the preconditions of a single fuzzy rule are satisfied, ranging from 0 to 1, with a larger value indicating a more sufficient rule triggering condition. Weighted average can be understood as using the activation intensity of each rule as a weight to perform a weighted summation of the rule output commands, obtaining a comprehensive control value, and realizing multi-rule fusion decision-making. Initial control commands can be understood as the original commands after fuzzy inference without smoothing or amplitude limiting processing, including initial wiper commands and initial fill light commands, with values ​​ranging from 0 to 1.

[0190] Specifically, the controller can match the preconditions of each fuzzy rule in the preset expert rule base based on the membership degrees of rainfall, glare, and sharpness in the fuzzy logic input parameters. Using the Mamdani minimum inference method, the minimum value of each membership degree in the rule precondition is taken as the activation strength of the rule, quantifying the triggering effectiveness of each rule. Using the activation strength of each rule as a weight, the wiper command value and the supplementary light command value in the rule conclusion are weighted, summed, and normalized to obtain the initial wiper command and initial supplementary light command, achieving multi-rule decision fusion and outputting the original control target value. A first-order low-pass filter is used to smooth the initial command in the time domain, suppressing inter-frame abrupt changes and jitter. The smoothed command is then limited to the 0-1 range, finally outputting continuous stepless wiper speed adjustment command and supplementary light dimming command, ensuring stable control and conforming to the actuator's operating range.

[0191] For example, a specific example can be used to demonstrate how the instruction is determined. Figure 5 A flowchart for determining instructions in a vehicle control method is provided in Embodiment 1 of the present invention, as follows: Figure 5 As shown, the steps for determining the fuzzy logic input parameters may include: 1. Input acquisition and fuzzification: Input quantity acquisition and normalization. The three continuous input quantities—rain intensity RR, glare intensity GG, and license plate area clarity CC—are acquired in real-time from each perception module for the current frame. Their values ​​have been normalized to the [0,1] interval. 0 and 1 represent the lowest and highest levels of each physical quantity, respectively. Fuzzy sets and triangular membership functions are defined, setting three fuzzy linguistic variables: "Low", "Medium", and "High". For any input quantity x∈{RR,GG,CC}, the triangular membership function maps the precise value to the membership degree μ(x) of each fuzzy set. The triangular membership function is completely determined by the vertex parameters a, b, and c, and is defined as follows:

[0192] μ(x;a,b,c)=0,x≤a

[0193] (xa) / (ba),a <x<b

[0194] 1, x=b

[0195] (cx) / (cb),b <x<c

[0196] 0, x≥c

[0197] For rainfall intensity membership calculation, three membership function parameters are defined for rainfall intensity RR:

[0198] Low membership degree: μL_R(RR)=triMF(RR,0.0,0.0,0.4); Medium membership degree: μM_R(RR)=triMF(RR,0.3,0.5,0.7); High membership degree: μH_R(RR)=triMF(RR,0.6,1.0,1.0), thus obtaining the membership vector of the current RR to the three fuzzy rainfall sets.

[0199] Glare intensity value membership calculation: For the glare intensity value GG, define three membership function parameters: Low membership: μL_G(GG)=triMF(GG,0.0,0.0,0.4); Medium membership: μM_G(GG)=triMF(GG,0.3,0.5,0.7); High membership: μH_G(GG)=triMF(GG,0.6,1.0,1.0), thus obtaining the membership vector of the current GG to the three glare fuzzy sets.

[0200] License plate area sharpness membership calculation: For sharpness CC, define three membership function parameters: Low membership: μL_C(CC)=triMF(CC,0.0,0.0,0.4); Medium membership: μM_C(CC)=triMF(CC,0.3,0.5,0.7); High membership: μH_C(CC)=triMF(CC,0.6,1.0,1.0), thus obtaining the membership vector of the current CC to the three sharpness fuzzy sets.

[0201] The membership vectors of the three input quantities are combined to form the fuzzy input state vector of the current scene, which can be used for subsequent fuzzy inference.

[0202] A predefined fuzzy rule base containing M rules, each rule having the following form:

[0203] "If (rainfall is A_r) and (glare is A_g) and (clarity is A_c), then (wiper speed control command is w, and fill light dimming command is l)", where A_r, A_g, A_c ∈ {low(0), medium(1), high(2)}, and w and l are recommended output values ​​within [0,1]. The rule set used in this embodiment is as follows: Rule 1: IF High Rainfall AND Low Glare THEN High Wiper (0.85) AND Medium Fill Light (0.50); ​​Rule 2: IF High Rainfall AND High Glare THEN Medium Wiper (0.50) AND Low Fill Light (0.15); Rule 3: IF Low Sharpness AND Low Glare THEN Low Wiper (0.15) AND High Fill Light (0.85); Rule 4: IF Low Sharpness AND High Glare THEN Medium Wiper (0.50) AND Medium Fill Light (0.50); ​​Rule 5: Default rule (unconditionally activated): Medium Wiper (0.50) AND Medium Fill Light (0.50).

[0204] Rule activation strength calculation (Mamdani minimum inference): For each rule j, based on the fuzzy set indices corresponding to rainfall, glare, and sharpness in its preconditions, the corresponding membership values ​​are taken from the obtained membership vector, and the activation strength α_j of the rule is calculated using the minimum value operation:

[0205] α_j=min(μ_r,j,μ_g,j,μ_c,j)

[0206] Where μ_r,j is the membership degree corresponding to the rainfall condition in rule j, and μ_g,j and μ_c,j are similar. If the rule is an unconditional default rule, then α_j is directly set to 1.

[0207] The initial control commands were obtained by defuzzifying the wiper speed control command and the supplementary light dimming command using a weighted average method (center of gravity method):

[0208] W_raw=(Σ_{j=1}^{M}α_j×w_j) / (Σ_{j=1}^{M}α_j)

[0209] L_raw=(Σ_{j=1}^{M}α_j×l_j) / (Σ_{j=1}^{M}α_j)

[0210] Where w_j and l_j are the recommended output values ​​for the wiper and fill light given in the consequent of rule j, respectively. If the sum of the activation intensities of all rules is 0, the default value of 0.5 is output.

[0211] To avoid drastic jitter in control commands between different frames, a first-order low-pass filter (exponential smoothing) is applied to W_raw and L_raw respectively to obtain the smoothed output commands W(t) and L(t) for the current frame:

[0212] W(t) = α × W_raw + (1 - α) × W(t-1)

[0213] L(t) = α × L_raw + (1 - α) × L(t-1)

[0214] Where α is the smoothing coefficient, typically 0.7; W(t-1) and L(t-1) are the smoothing instruction values ​​of the previous frame, initially set to 0.5.

[0215] Finally, the smoothed instruction values ​​are limited to the range [0,1]:

[0216] W_final=clamp(W(t),0,1)

[0217] L_final=clamp(L(t),0,1)

[0218] The final output W_final is used to drive the wiper frequency (0 for stop, 1 for maximum speed), and L_final is used to drive the brightness of the fill light (0 for off, 1 for maximum brightness). The two work together to achieve adaptive optimization of visual clarity and image quality in rainy night scenes.

[0219] In terms of perception accuracy, this invention utilizes the Canny algorithm's high sensitivity to raindrop edges. Through dynamic ROI delineation and multi-dimensional feature filtering, it can accurately distinguish continuous rainfall changes from light rain to heavy rain. Compared to the limitations of traditional optical sensors that can only output a limited range of speeds, this significantly improves the precision of rainfall detection. Simultaneously, through brightness histogram and edge feature fusion analysis, it can quantify glare interference levels in real time and assess license plate clarity based on character edge integrity, providing multi-dimensional and accurate perception input for collaborative control. Regarding collaborative control, this invention breaks through the technical barrier of independent operation of traditional wiper and supplementary lighting systems. Through a fuzzy logic decision-maker, it fuses and infers the three perception signals—rainfall, glare, and license plate clarity—to dynamically generate wiper speed adjustment and supplementary lighting dimming commands. This mechanism can automatically adjust the wiping frequency according to the real-time rain intensity to avoid visual interference caused by the wipers being too fast or too slow; at the same time, it dynamically adjusts the brightness of the supplementary light according to the glare level and the clarity of the license plate, actively reducing light and suppressing scattering in strong glare, and moderately increasing light to enhance contrast when the license plate is blurry, thereby simultaneously optimizing the driver's visual comfort and the monitoring and evidence collection effect of the vehicle DVR in rainy night environments.

[0220] Example 2

[0221] Figure 6 This is a schematic diagram of a vehicle control device provided in Embodiment 2 of the present invention. Figure 6 As shown, the device includes:

[0222] The parameter acquisition module 61 is used to acquire the vehicle's windshield wiper-related parameters, glare-related parameters, and real-time video stream from the vehicle-mounted DVR camera;

[0223] The feature determination module 62 is used to determine an image feature set based on the real-time video stream, the wiper-related parameters, and the glare-related parameters. The image feature set includes rainfall intensity value, glare intensity value, and license plate area clarity.

[0224] The instruction control module 63 is used to determine the wiper speed adjustment instruction and the fill light dimming instruction according to the preset expert rule base and the image feature set, and to perform drive control.

[0225] The technical solution of this invention achieves multi-dimensional and precise quantification of rainfall, glare, and license plate clarity based on real-time video streams from an in-vehicle DVR, wiper parameters, and glare-related parameters. It integrates multi-dimensional features and uses a pre-set expert rule base to achieve multi-objective optimization decisions, dynamically generating wiper speed adjustment and supplementary lighting dimming commands. This effectively suppresses glare in rainy nights and avoids overexposure of license plates on vehicles ahead, achieving dynamic and adaptive control of wiper frequency and supplementary lighting brightness. In rainy night environments, it simultaneously optimizes driver visibility and the effectiveness of in-vehicle DVR monitoring and evidence collection. No additional rain or light sensors are required, resulting in low hardware costs, strong compatibility, and suitability for various in-vehicle scenarios.

[0226] Furthermore, the feature determination module 62 includes:

[0227] The first determining unit is used to determine the rainfall intensity value based on the real-time video stream and the wiper-related parameters;

[0228] The second determining unit is used to determine the glare intensity value based on the real-time video stream and the glare-related parameters;

[0229] The third determining unit is used to analyze the license plate area of ​​the real-time video stream and determine the clarity of the license plate area.

[0230] The fourth determining unit is used to use the rainfall intensity value, the glare intensity value, and the license plate area clarity as an image feature set.

[0231] Specifically, the first determining unit is used for:

[0232] Based on the physical coverage range of the wipers in the wiper-related parameters, determine the dynamic region of interest in the real-time video stream and extract sub-images;

[0233] The sub-images are preprocessed to determine the preprocessed image;

[0234] An adaptive edge detection operator is used to extract edges from the preprocessed image to determine the raindrop edge contours in the preprocessed image;

[0235] The edge contours of the raindrops are filtered based on their geometric features and inter-frame motion consistency to determine the effective raindrop features;

[0236] The rainfall intensity value is determined based on the spatiotemporal distribution of the effective raindrop characteristics.

[0237] Specifically, the second determining unit is used for:

[0238] Based on the driver's field of vision in the glare-related parameters, the region of interest for glare assessment in the real-time video stream is determined;

[0239] The region of interest for glare assessment is converted to the HSV color space and the luminance channel is extracted to determine the luminance distribution image;

[0240] The brightness glare index is determined based on the high brightness distribution image and the proportion of high brightness.

[0241] Edge feature analysis is performed on the brightness distribution image to determine the edge glare index;

[0242] The glare intensity parameters are determined based on the brightness glare index and the edge glare index.

[0243] Specifically, the third determining unit is used for:

[0244] Based on the detection and tracking strategy, the license plate area of ​​the vehicle in front is analyzed in the real-time video stream to determine the license plate area of ​​the vehicle in front.

[0245] The license plate area is preprocessed to determine the preprocessed license plate image;

[0246] The character edges in the preprocessed license plate image are extracted and a multi-dimensional sharpness index is calculated to determine the sharpness of the license plate area.

[0247] Furthermore, the instruction control module 63 includes:

[0248] The fifth determining unit is used to fuse the image feature set to determine the fuzzy logic input parameters;

[0249] The sixth determining unit is used to perform multi-objective optimization decision-making on the fuzzy logic input parameters based on a preset expert rule base, and to determine the wiper speed adjustment command and the supplementary light dimming command.

[0250] The seventh determining unit is used to convert the wiper speed control command into a wiper wiping frequency signal and determine the wiper drive signal;

[0251] The eighth determining unit is used to convert the supplementary light dimming command into a supplementary light brightness adjustment signal and determine the supplementary light driving signal;

[0252] The ninth determining unit is used to perform drive control through the wiper drive signal and the supplementary light drive signal.

[0253] Specifically, the fifth determining unit is used for:

[0254] The rainfall intensity value, the glare intensity value, and the license plate area clarity are normalized to determine the normalized feature value;

[0255] The normalized feature values ​​are mapped based on a preset triangular membership function to determine the fuzzy membership degree of each input quantity.

[0256] The fuzzy membership degrees are combined to obtain the fuzzy logic input parameters.

[0257] Specifically, the sixth determining unit is used for:

[0258] The activation strength of each fuzzy rule is determined based on a pre-set expert rule base.

[0259] Based on the weighted average of the activation intensity, an initial control command is determined, which includes an initial wiper command and an initial fill light command.

[0260] The initial control command is smoothed, filtered, and limited to determine the wiper speed control command and the supplementary light dimming command.

[0261] The vehicle control device provided in the embodiments of the present invention can execute the vehicle control method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0262] Example 3

[0263] Figure 7 This is a structural schematic diagram of a vehicle provided in Embodiment 3 of the present invention, as shown below. Figure 7 As shown, the vehicle includes a controller 71, a memory 72, an input device 73, an output device 74, and an onboard DVR camera 75; the number of controllers 71 in the vehicle can be one or more. Figure 7 Taking a controller 71 as an example, the number of vehicle-mounted DVR cameras 75 can be one or more. Figure 7 Taking a vehicle-mounted DVR camera 75 as an example; the controller 71, memory 72, input device 73, output device 74, and vehicle-mounted DVR camera 75 in the vehicle can be connected via a bus or other means. Figure 7 Taking the example of a connection between China and Israel via a bus.

[0264] The memory 72, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the vehicle control method in this embodiment of the invention (e.g., the parameter acquisition module 61, feature determination module 62, and instruction control module 63 in the vehicle control device). The controller 61 executes various vehicle functions and data processing by running the software programs, instructions, and modules stored in the memory 62, thereby realizing the aforementioned vehicle control method.

[0265] The memory 62 may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function; the data storage area may store data created based on terminal usage. Furthermore, the memory 62 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory, or other non-volatile solid-state storage device. In some instances, the memory 62 may further include memory remotely configured relative to the controller 61, which can be connected to the vehicle via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0266] Input device 63 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the cloud platform. Output device 64 may include display devices such as a display screen.

[0267] Example 4

[0268] Embodiment 4 of the present invention also provides a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform a vehicle control method, including:

[0269] Acquire vehicle wiper parameters, glare parameters, and real-time video stream from the vehicle's DVR camera;

[0270] Based on the real-time video stream, the wiper-related parameters, and the glare-related parameters, an image feature set is determined, which includes rainfall intensity value, glare intensity value, and license plate area clarity.

[0271] Based on the preset expert rule base and the image feature set, the wiper speed adjustment command and the supplementary light dimming command are determined and driven.

[0272] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0273] It is worth noting that in the above embodiments of the vehicle control device, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0274] In one embodiment, the present invention further includes a computer program product, which includes a computer program that, when executed by a processor, implements the transaction rate limiting method of any embodiment of the present invention.

[0275] In implementing the computer program product, computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0276] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A vehicle control method, characterized in that, include: Acquire vehicle wiper parameters, glare parameters, and real-time video stream from the vehicle's DVR camera; Based on the real-time video stream, the wiper-related parameters, and the glare-related parameters, an image feature set is determined, which includes rainfall intensity value, glare intensity value, and license plate area clarity. Based on the preset expert rule base and the image feature set, wiper speed adjustment command and supplementary light dimming command are determined and driven.

2. The method according to claim 1, characterized in that, The step of determining the image feature set based on the real-time video stream, the wiper-related parameters, and the glare-related parameters includes: The rainfall intensity value is determined based on the real-time video stream and the wiper parameters. The glare intensity value is determined based on the real-time video stream and the glare-related parameters. The real-time video stream is analyzed to determine the clarity of the license plate area. The rainfall intensity value, the glare intensity value, and the license plate area clarity are used as the image feature set.

3. The method according to claim 2, characterized in that, The step of determining the rainfall intensity value based on the real-time video stream and the wiper-related parameters includes: Based on the physical coverage range of the wipers in the wiper-related parameters, determine the dynamic region of interest in the real-time video stream and extract sub-images; The sub-images are preprocessed to determine the preprocessed image; An adaptive edge detection operator is used to extract edges from the preprocessed image to determine the raindrop edge contours in the preprocessed image; The raindrop edge contours are filtered based on raindrop geometric features and inter-frame motion consistency to determine valid raindrop features; The rainfall intensity value is determined based on the spatiotemporal distribution of the effective raindrop characteristics.

4. The method according to claim 2, characterized in that, The step of determining the glare intensity value based on the real-time video stream and the glare-related parameters includes: Based on the driver's field of vision in the glare-related parameters, the region of interest for glare assessment in the real-time video stream is determined; The region of interest for glare assessment is converted to the HSV color space and the luminance channel is extracted to determine the luminance distribution image; The brightness glare index is determined based on the high brightness distribution image and the proportion of high brightness. Edge feature analysis is performed on the brightness distribution image to determine the edge glare index; The glare intensity parameters are determined based on the brightness glare index and the edge glare index.

5. The method according to claim 2, characterized in that, The step of analyzing the license plate area of ​​the real-time video stream to determine the clarity of the license plate area includes: Based on the detection and tracking strategy, the license plate area of ​​the vehicle in front is analyzed in the real-time video stream to determine the license plate area of ​​the vehicle in front. The license plate area is preprocessed to determine the preprocessed license plate image; The character edges in the preprocessed license plate image are extracted and a multi-dimensional sharpness index is calculated to determine the sharpness of the license plate area.

6. The method according to claim 1, characterized in that, The step of determining the wiper speed control command and the supplementary light dimming command based on the preset expert rule base and the image feature set, and performing drive control, includes: The image feature set is fused to determine the fuzzy logic input parameters; Based on a preset expert rule base, multi-objective optimization decision-making is performed on the fuzzy logic input parameters to determine the wiper speed adjustment command and the supplementary light dimming command. The wiper speed control command is converted into a wiper wiping frequency signal to determine the wiper drive signal; The supplementary light dimming command is converted into a supplementary light brightness adjustment signal to determine the supplementary light drive signal; Drive control is performed using the wiper drive signal and the fill light drive signal.

7. The method according to claim 6, characterized in that, The process of fusing the image feature set to determine the fuzzy logic input parameters includes: The rainfall intensity value, the glare intensity value, and the license plate area clarity are normalized to determine the normalized feature value; The normalized feature values ​​are mapped based on a preset triangular membership function to determine the fuzzy membership degree of each input quantity. The fuzzy membership degrees are combined to obtain the fuzzy logic input parameters.

8. The method according to claim 6, characterized in that, The process of performing multi-objective optimization decision-making on the fuzzy logic input parameters based on a preset expert rule base to determine the wiper speed adjustment command and the supplementary light dimming command includes: The activation strength of each fuzzy rule is determined based on a pre-set expert rule base. Based on the weighted average of the activation intensity, an initial control command is determined, which includes an initial wiper command and an initial fill light command. The initial control command is smoothed, filtered, and limited to determine the wiper speed control command and the supplementary light dimming command.

9. A vehicle control device, characterized in that, include: The parameter acquisition module is used to acquire vehicle wiper-related parameters, glare-related parameters, and real-time video stream from the vehicle-mounted DVR camera; The feature determination module is used to determine an image feature set based on the real-time video stream, the wiper-related parameters, and the glare-related parameters. The image feature set includes rainfall intensity value, glare intensity value, and license plate area clarity. The instruction control module is used to determine the wiper speed adjustment instruction and the fill light dimming instruction based on the preset expert rule base and the image feature set, and to perform drive control.

10. A vehicle, characterized in that, The vehicles include: At least one controller; The vehicle-mounted DVR camera is communicatively connected to the at least one controller; and a memory that is communicatively connected to the at least one controller; The memory stores a computer program that can be executed by the at least one controller, which is then executed by the at least one controller to enable the at least one controller to perform the vehicle control method according to any one of claims 1-8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause the controller to perform the vehicle control method according to any one of claims 1-8.

Images