Self-adaptive illumination control system and method based on basic lamp control function and ADAS bidirectional sensing cooperation
By establishing a two-way perception and collaboration framework between basic lighting control functions and ADAS systems, two-way transmission of environmental information and fine-grained data mapping are realized. This solves the problems of wasted perception resources and insufficient control precision caused by data fragmentation in existing technologies, and improves the target detection rate of ADAS systems and the light shape control precision of ADB functions in strong light scenarios.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-14
- Publication Date
- 2026-04-07
AI Technical Summary
The existing basic lighting control functions and ADAS unidirectional collaborative architecture suffer from systemic defects due to data fragmentation, insufficient information granularity, and asynchronous states, resulting in wasted sensing resources and insufficient control precision.
By constructing a bidirectional deep integration framework between basic lighting control functions and ADAS, bidirectional channel transmission of environmental information and fine-grained data mapping are realized. Combined with the environmental perception fusion module and feature mapping and decision module, the target detection rate of ADAS system and the light shape control accuracy of ADB function in strong light scenarios are improved.
This has improved the perception reliability of ADAS systems in strong light scenarios, increased the target detection rate, achieved a breakthrough in the light shape control accuracy of ADB function, reduced the matching error between light shape and target contour, and improved the utilization rate and accuracy of LED lighting.
Smart Images
Figure CN121815496A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of vehicle electronic control and intelligent lighting intersection technology, specifically to an adaptive lighting control system and method that coordinates basic lighting control functions with ADAS bidirectional perception. Background Technology
[0002] With the widespread adoption of matrix LED headlights, basic lighting control functions can achieve vehicle attitude calculation (ALS function) and adaptive high beam shading or low beam enhancement (ADB function). In the current mainstream solutions involving the fusion of basic lighting control and ADAS, the ADAS relies on unidirectional transmission of structured target data (such as vehicle coordinate rectangles) to the basic lighting control module to execute basic lighting control. However, crucial raw information captured by ALS sensors (such as ambient light sensors monitoring ambient light intensity) (e.g., glare intensity at tunnel exits) cannot be fed back to the ADAS system. This causes ADAS to continuously output incorrect target data in camera-limited scenarios (strong light blindness). Furthermore, the rectangular target data transmitted by ADAS loses some perceptual details (such as obstacle edge contours and background brightness gradients), resulting in the ADB function only being able to output a relatively coarse rectangular shading light shape, unable to adapt to the target contour.
[0003] Therefore, the current architecture is prone to problems such as waste of perceived resources, insufficient control precision, and failure of system coordination. Summary of the Invention
[0004] The technical problem this invention aims to solve is the systemic defects in the existing basic lighting control function and ADAS unidirectional collaborative architecture caused by data fragmentation, insufficient information granularity, and asynchronous state.
[0005] To this end, the present invention provides an adaptive lighting control system and method that coordinates basic lighting control functions with ADAS bidirectional perception, achieving the following breakthroughs by constructing a bidirectional deep integration framework: Establish a two-way data channel between the underlying sensors of basic lighting control and ADAS, so that environmental information can directly participate in ADAS perception and decision-making; Establish a fine-grained data mapping mechanism from ADAS to ADB functions, transforming the coordinates of a rectangular target into gradient feature parameters that can drive precise light shape control.
[0006] The technical solution adopted by this invention to solve its technical problem is: An adaptive lighting control system that coordinates basic lighting control functions with ADAS bidirectional sensing includes, An environmental perception fusion module integrates data collected by the basic lighting sensor (ALS) and the raw data from the ADA system to adjust the brightness data injected into the ADAS system. Meanwhile, the environmental perception fusion module is also used to obtain high-dimensional features (target semantic segmentation, edge gradient enhancement) of the ADAS system perception output and fuse them.
[0007] The feature mapping and decision module injects the adjusted brightness data into the ADAS system perception algorithm (such as the CNN feature extraction layer) in real time, so as to realize the focus of computing resources in the dark area and improve the target detection rate of the ADAS system in overexposed scenes (such as strong light and backlight). The feature mapping and decision module is also used to convert the fused data mapping into a light shape control matrix executable by the vehicle's ADB system. This enables an accurate description of the target contour intensity distribution, thereby reducing the matching error between the light shape and the target contour.
[0008] An adaptive lighting control method includes the following steps: S1, Ambient brightness distribution transmission and recognition control: The ambient brightness distribution acquired by ALS is mapped to a region weight mask for ADAS image processing; the weights are injected into the feature extraction layer of the functional image processing network of the ADAS system to reduce the data credibility and target detection weight of overexposed areas.
[0009] By transmitting the brightness distribution information of the ALS light sensor to the image processing algorithm of ADAS, the problem of ADAS perception degradation in strong light scenes is solved, the computing resources are focused on the effective dark area, and the target detection rate is improved under strong exposure.
[0010] S2, Target Gradient Transmission and Light Shape Control: ADAS image processing obtains a semantic segmentation branch and a gradient generation branch through dual-branch feature extraction. The binary mask of the target object in the semantic segmentation branch and the dynamic gradient map in the gradient generation branch are fused to obtain the light shape control matrix. The light shape control matrix is sent to the ADB system, and the ADB system controls the corresponding LED brightness according to the light shape control matrix.
[0011] By transmitting the gradient information of ADAS to the ADB module, the intensity distribution of the target contour can be accurately described, thereby reducing the matching error between the light pattern and the target contour, improving the effective lighting utilization rate, reducing glare pollution, and providing real-time performance.
[0012] Further, step S1 specifically includes: The S11 uses the standard ALS light sensor to collect ambient brightness distribution data. And mapped to the region weight mask for image processing in ADAS systems.
[0013] It should be noted that the weight mask includes the weights of all pixels. Two-dimensional matrix / image carrier, weights It is the core component of the weight mask.
[0014] S12 uses the pixel-level brightness matrix output by the light sensor. Re-weighting calculate:
[0015] in, As weight; The attenuation coefficient is... ; The larger the value, the stronger the light intensity, indicating an overexposed area, and thus a higher weight. The smaller it is.
[0016] S13 calibration in different scenarios Values, calculate the corresponding weights Weights are injected into the feature extraction layer of the ADAS functional image processing network to reduce the data reliability of overexposed areas and the weight of target detection. This allows computing resources to be focused on the effective dark area.
[0017] Further, in step S13, the attenuation coefficient The calibration method is as follows: A biological model is introduced, and non-uniform calibration is performed based on the visual characteristics of the human eye. The human eye's light adaptation curve function is:
[0018] in, It refers to ambient brightness; This is the initial sensitivity coefficient, with a value ranging from 0.15 to 0.2; It is the adaptation rate coefficient, with a value ranging from 0.002 to 0.004; It is a fundamental condition constant, with a value ranging from 0.01 to 0.03.
[0019] Further, step S2 specifically includes: Images acquired by the S21 ADAS system are processed by the semantic segmentation branch of a constructed gradient-aware neural network to output a binary mask. ;
[0020] The gradient prediction branch in the S22 gradient-aware neural network generates a dynamic gradient map using the posterior probability distribution of semantic edges in the image. :
[0021] in, This indicates a meaningful object edge event for that pixel; Input image; These are network parameters used to calculate dynamic gradients.
[0022] S23 achieves gradient mask fusion through a gradient-guided attention fusion module, resulting in an enhanced mask.
[0023] S24 abstracts the enhancement mask into probability values. The LED pixel array of the ADB system headlights is spatially mapped one-to-one, establishing a nonlinear mapping function from probability to target brightness, thereby obtaining a dynamic light shape control matrix for controlling the ADB system headlights.
[0024] The beneficial effects of this invention are: 1. Improved perception reliability: By establishing a two-way data channel between the underlying sensors of basic lighting control (light sensor) and ADAS, environmental information (such as glare intensity) can directly participate in ADAS perception decision-making, solving the problem of ADAS perception degradation in strong light scenarios and improving the target detection rate under strong exposure. 2. Breakthrough in beam shape control precision: Establish a fine-grained data mapping mechanism from ADAS to ADB functions, transform the coordinates of a rectangular target into gradient feature parameters that can drive precise beam shape control, drive the LED matrix to generate adaptive light spots that match the contours of obstacles, and improve the utilization rate and accuracy of LED lighting. Attached Figure Description
[0025] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0026] Figure 1 This is a structural block diagram of the control system of the present invention; Figure 2 This is a flowchart of the environmental brightness distribution transmission and recognition control in the control method of the present invention; Figure 3 This is a flowchart of the target gradient transmission and optical shape control in the control method of the present invention. Detailed Implementation
[0027] The invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention and therefore showing only the components relevant to the invention. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values of the components and steps set forth in these embodiments do not limit the scope of this disclosure.
[0028] It should also be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn to actual scale. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit this disclosure or its application or use.
[0029] Techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and apparatus should be considered part of the specification. In all examples shown and discussed herein, any specific values should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.
[0030] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, features defined with "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.
[0031] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.
[0032] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.
[0033] Reference Figure 1 An adaptive lighting control system that coordinates basic lighting control functions with ADAS bidirectional perception includes an environmental perception fusion module and a feature mapping and decision module, with the output of the environmental perception fusion module serving as the input of the feature mapping and decision module.
[0034] The environmental perception fusion module maps the ambient brightness distribution data collected by the basic lighting control sensor (ALS) into a region weight mask for ADAS image processing. By introducing a weight calculation formula, it readjusts the weight values in the weight mask, reducing the weight of overexposed areas. The module also performs target semantic segmentation and gradient prediction on the original image collected by the ADAS system to obtain a binary mask and dynamic gradient map of the target object, and fuses these two to obtain an enhanced mask. The feature mapping and decision module re-injects the adjusted weight mask into the feature extraction layer of the ADAS system, enabling the ADAS system to focus its computational resources on effectively dark areas and improve the target detection rate in overexposed scenes (such as strong light and backlight). The feature mapping and decision module obtains the probability values of the enhanced mask and obtains a light shape control matrix through the mapping relationship between the probability value matrix and the LED target brightness. The ADB system controls the corresponding LED brightness of the vehicle lights according to the light shape control matrix.
[0035] Reference Figure 2 , Figure 3 An adaptive lighting control method that coordinates basic lighting control functions with ADAS bidirectional sensing includes the following steps: S1, Ambient brightness distribution transmission and recognition control: The ambient brightness distribution acquired by ALS is mapped to a region weight mask for ADAS image processing; the weights are injected into the feature extraction layer of the functional image processing network of the ADAS system to reduce the data credibility and target detection weight of overexposed areas.
[0036] Specifically: The S11 uses the standard ALS light sensor to collect ambient brightness distribution data. And mapped to the region weight mask for image processing in ADAS systems.
[0037] It should be noted that the weight mask includes the weights of all pixels. Two-dimensional matrix / image carrier, weights It is the core component of the weight mask.
[0038] S12 uses the pixel-level brightness matrix output by the light sensor. Re-weighting calculate:
[0039] in, As weight; The attenuation coefficient is... ; The larger the value, the stronger the light intensity, indicating an overexposed area, and thus a higher weight. The smaller it is.
[0040] S13 calibration in different scenarios Values (e.g., in a tunnel) Normal dusk ), calculate the corresponding weights (weight) As a core component of the weighted mask, weights are injected into the feature extraction layer of the ADAS functional image processing network to reduce the data credibility and target detection weights in overexposed areas, thereby focusing computational resources on effective dark areas.
[0041] It should be noted that, as can be seen from the above, The calibration of the values has a significant impact on the weight calculation of strong light regions, therefore it is necessary to clarify them. Methods for calibrating values.
[0042] In this scheme, the attenuation coefficient The calibration method is as follows: A biological model is introduced, and non-uniform calibration is performed based on the visual characteristics of the human eye. The human eye's light adaptation curve function is:
[0043] in, It is the distribution of ambient brightness. ; This is the initial sensitivity coefficient, with a value ranging from 0.15 to 0.2; It is the adaptation rate coefficient, with a value ranging from 0.002 to 0.004; It is a fundamental condition constant, with a value ranging from 0.01 to 0.03.
[0044] By introducing biological models to obtain K values, we can simulate the photosensitivity of the real human eye, protect sensitivity in low light, and control the speed of strong light suppression. In one embodiment, common sense can be used to calibrate. , , The values are shown in the table below.
[0045] Table 1 , , Example table of parameter numerical calibration
[0046] S2, Target Gradient Transmission and Light Shape Control: The image acquired by the ADAS system is processed through dual-branch feature extraction to obtain a semantic segmentation branch and a gradient generation branch. The binary mask output by the semantic segmentation branch and the dynamic gradient map output by the gradient prediction branch are fused to generate a light shape control matrix. The light shape control matrix is sent to the ADB system, and the ADB system controls the corresponding LED brightness according to the light shape control matrix.
[0047] Specifically: S21 constructs a two-branch gradient-aware neural network, which includes a shared feature encoder, a semantic segmentation branch, and a gradient prediction branch. The shared feature encoder is used to extract general features from the input image. This general feature map is then fed in parallel to the semantic segmentation branch and the gradient prediction branch. The semantic segmentation branch calculates a binary mask of the target object through the ADAS system's semantic segmentation network. ;
[0048] S22 generates a dynamic gradient map. ; The S221 gradient prediction branch is used to output a dynamic gradient map with the same resolution as the input image. :
[0049] in, This indicates a meaningful object edge event for that pixel; Input image; These are network parameters, which are the "transformation rules" for generating enhanced edge features from general feature maps, enabling the calculation of dynamic gradients.
[0050] It should be noted that the core of the gradient prediction branch includes a gradient direction pooling layer; this layer computes the spatial gradients of the input general feature map in parallel along eight preset directions (0°, 45°, 90°, 135°, 180°, 225°, 270°, 315°), and performs max pooling on the channel dimension of these eight gradient maps, outputting enhanced edge features (enhanced edge features are...). (intermediate feature carrier), which maps enhanced edge features to... Dynamic gradient map at the same resolution.
[0051] S222 employs a collaborative training strategy to optimize the gradient-aware neural network; the loss function used during training... It consists of three parts: semantic segmentation loss Gradient reconstruction loss And the unique collaborative consistency loss Among them, the collaborative consistency loss is used to constrain the dynamic gradient graph. The high-response regions in the semantic segmentation binary mask must fall within the semantic segmentation binary mask. Near the edge, ensure that the gradient is semantically aligned with the target contour.
[0052] After the S223 gradient perception neural network is trained, it is deployed in an automotive-grade computing unit. During forward inference, it is input with real-time images from the vehicle's onboard camera in the ADAS system, and the gradient prediction branch outputs a dynamic gradient map. In this image, each pixel value is a continuous probability value, representing the likelihood that the location belongs to a salient edge.
[0053] S23 will use the dynamic gradient map Binary masks for semantic segmentation The fusion process generates an enhanced mask for beam shape control, specifically including: S231: Design a gradient-guided attention fusion module; this module uses a semantic segmentation binary mask. and dynamic gradient plot As input, the dynamic gradient map is first thresholded and enhanced using a gating function to generate an edge attention weight map. The calculation formula is:
[0054] Where σ is the Sigmoid function; k is an enhancement coefficient greater than 5; This is the preset gradient threshold (usually set to 0.5).
[0055] S232 will use the edge attention weight map Binary masks for semantic segmentation Element-wise multiplication is performed to obtain the enhancement weights for the target edges; then, these enhancement weights are added to the base weights (set to 1) to generate the final pixel-level attention map. The calculation formula is: .
[0056] S233 will use pixel-level attention maps Compared with the original semantic segmentation binary mask Element-wise multiplication yields the enhanced target mask. The calculation formula is:
[0057] This step enhances the pixel weights located on the target contour (high gradient region) by 1 to 2 times, while keeping the weights of flat regions inside the target unchanged, thereby sharpening the edge contours of the binary mask.
[0058] S234 for enhancing mask Subpixel-level edge contour extraction is performed; steep probability faults formed near the target contour by an enhanced mask are used to accurately calculate the position of the contour line with a probability value of 0.5 using a linear interpolation algorithm. This position is then resolved as the physical contour boundary of the target, with a positioning accuracy of centimeter level (better than ±2cm). This accurate contour serves as the direct basis for the subsequent adaptive lighting system to modulate the light shape.
[0059] S24 is based on the probability value of the enhanced mask. Designing light patterns involves the following steps: S241: The enhanced mask output from step S23 Defined as probability value ,in Represents image coordinates The confidence level that a pixel belongs to the target area to be masked; this probability distribution is spatially mapped one-to-one with the LED pixel array of the ADB system headlights.
[0060] S242 establishes a probability value LED target brightness The nonlinear mapping function has the following core formula: in: Corresponding to image coordinates LED pixels The driving brightness value; It is the maximum drivable brightness of the LED pixel; It is to enhance the mask in coordinates The probability value at that location; It is the dynamic edge steepness factor, a function of vehicle speed v. The higher the speed, The larger the value, the more accurate the mapping function becomes in terms of probability value. The steeper the slope nearby, the clearer and sharper the edge of the generated light spot, ensuring the timeliness and accuracy of shading at high speeds; at low speeds, the edge is softer, improving visual comfort. It is the background light intensity attenuation factor. , With the target area (The area of all object regions in the current frame image that are identified by the ADAS semantic segmentation network and enhanced by the gradient-guided attention fusion module, and that need to be selectively occluded (or dimmed) by the ADB system headlights) accounts for the total field of view area. The proportion is directly proportional. The coefficient is used when the target area occupies a relatively small area. Approaching 0, the background area remains close. The brightness is sufficient to provide adequate basic lighting; when the target area is large (such as in congested traffic), Increase the brightness of non-target areas and automatically reduce the overall brightness to save energy and reduce glare and interference to other vehicles. It is a hyperbolic tangent function used to generate a smooth, continuous S-shaped brightness transition curve, ensuring that the light spot has no abrupt changes.
[0061] S243 will calculate all The final light shape control matrix is formed, which precisely describes the desired brightness of each controllable pixel of the ADB system LED headlight, thereby generating a light spot in physical space that matches the target contour, has smooth edges, and adaptive brightness distribution.
[0062] This concludes the detailed description of an adaptive lighting control system and method that coordinates basic lighting control functions with ADAS bidirectional sensing according to this disclosure. To avoid obscuring the concept of this disclosure, some details known in the art have not been described. Those skilled in the art will fully understand how to implement the technical solutions disclosed herein based on the above description.
[0063] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined by the scope of the claims.
Claims
1. An adaptive lighting control system that coordinates basic lighting control functions with ADAS bidirectional sensing, characterized in that, include, An environmental perception fusion module adjusts the data collected by the basic lighting control sensors in real time to focus computing resources on the effective dark area. The environment perception fusion module is also used to fuse the binary mask and dynamic gradient map based on the image output by the ADAS system; The feature mapping and decision module injects the adjusted data into the feature extraction layer of the ADAS system in real time to improve the detection rate of the ADAS system in overexposure scenarios; the feature mapping and decision module transforms the fused data into a light shape control matrix that can be executed by the vehicle ADB system.
2. The adaptive lighting control system with basic lighting control function and ADAS bidirectional perception coordination as described in claim 1, characterized in that, The environmental perception fusion module integrates the ambient brightness distribution data collected by the basic lighting control sensor. The region weight mask is mapped to the image processing of the ADAS system, and the weights in the weight mask are readjusted through the weight calculation formula to reduce the weight values of overexposed areas.
3. The adaptive lighting control system with basic lighting control function and ADAS bidirectional perception coordination as described in claim 2, characterized in that, The weight calculation formula is as follows: in, As weight; The attenuation coefficient is... .
4. The adaptive lighting control system with basic lighting control function and ADAS bidirectional perception coordination as described in claim 3, characterized in that, The attenuation coefficient The calibration is calculated using the human eye adaptation curve function: in, It refers to ambient brightness; This is the initial sensitivity coefficient, with a value ranging from 0.15 to 0.2; It is the adaptation rate coefficient, with a value ranging from 0.002 to 0.004; It is a fundamental condition constant, with a value ranging from 0.01 to 0.
03.
5. The adaptive lighting control system with basic lighting control function and ADAS bidirectional perception coordination as described in claim 1, characterized in that, The feature mapping and decision module completes gradient mask fusion through the attention fusion module and outputs an enhanced mask; The light pattern is designed based on the enhanced mask values.
6. An adaptive lighting control method that coordinates basic lighting control functions with ADAS bidirectional sensing according to any one of claims 1-5, characterized in that, Includes the following steps: S1, Ambient brightness distribution transmission and recognition control: Map the ambient brightness distribution acquired by ALS to a region weight mask for ADAS image processing; inject the weights into the feature extraction layer of the functional image processing network of the ADAS system; S2, Target Gradient Transmission and Light Shape Control: The image acquired by ADAS is processed by dual-branch feature extraction to obtain a semantic segmentation branch and a gradient generation branch. The binary mask of the target object in the semantic segmentation branch and the dynamic gradient map in the gradient generation branch are fused to obtain the light shape control matrix. The light pattern control matrix is sent to the ADB system, which then controls the corresponding LED brightness based on the light pattern control matrix.
7. The adaptive lighting control method based on the basic lighting control function and ADAS bidirectional perception coordination as described in claim 6, characterized in that, Step S1 specifically includes: The S11 uses the standard ALS light sensor to collect ambient brightness distribution data. And mapped to a region weight mask for image processing in ADAS systems; S12 uses the pixel-level brightness matrix output by the light sensor. Re-weighting calculate; S13 calibration in different scenarios Values, calculate the corresponding weights Weights are injected into the feature extraction layer of the ADAS functional image processing network.
8. The adaptive lighting control method based on the basic lighting control function and ADAS bidirectional sensing coordination as described in claim 6, characterized in that, Step S2 specifically includes: S21 Constructs a two-branch gradient-aware neural network, which includes a shared feature encoder, a semantic segmentation branch, and a gradient prediction branch; the shared feature encoder is used to extract general features of the input image; the semantic segmentation branch is used to output a binary mask of the target object. ; S22 describes a gradient prediction branch that generates a dynamic gradient graph. ; S23 will use the dynamic gradient graph Binary mask for semantic segmentation The mixture is then fused to generate an enhanced mask for beam pattern control. ; S24 is based on the probability value of the enhanced mask. Designing light patterns involves the following steps: S241: The enhanced mask output from step S23 Defined as probability value ,in Represents image coordinates The confidence level that a pixel belongs to the target area to be masked; this probability distribution is spatially mapped one-to-one with the LED pixel array of the ADB system headlights; S242 establishes a probability value To LED target brightness Nonlinear mapping function: in: Corresponding to image coordinates LED pixels The driving brightness value; It is the maximum drivable brightness of the LED pixel; It is to enhance the mask in coordinates The probability value at that location; It is the dynamic edge steepness factor, a function of vehicle speed v. The higher the vehicle speed, the more pronounced the steepness factor. The larger; The background light intensity attenuation factor is directly proportional to the ratio of the target area to the total field of view. It is the hyperbolic tangent function; S243 will calculate all The final light shape control matrix is formed, which precisely describes the desired brightness of each controllable pixel of the ADB system LED headlight, thereby generating a light spot in physical space that matches the target contour, has smooth edges, and adaptive brightness distribution.
9. The adaptive lighting control method based on the basic lighting control function and ADAS bidirectional sensing coordination as described in claim 8, characterized in that, The gradient prediction branch in the gradient-aware neural network generates a dynamic gradient map using the posterior probability distribution of semantic edges in the image. : in, This indicates a meaningful object edge event for that pixel; Input image; Using network parameters, we can calculate the dynamic gradient, thus obtaining the dynamic gradient map. ; The dynamic gradient map Each pixel value is a continuous probability value, representing the likelihood that the location belongs to a salient edge.
10. The adaptive lighting control method based on the basic lighting control function and ADAS bidirectional perception coordination as described in claim 8, characterized in that, Step S23 specifically includes: S231: Design a gradient-guided attention fusion module; this module uses a semantic segmentation binary mask. and dynamic gradient plot As input, the dynamic gradient map is first thresholded and enhanced using a gating function to generate an edge attention weight map. : Where σ is the Sigmoid function; k is an enhancement coefficient greater than 5; The preset gradient threshold; S232 will use the edge attention weight map Binary mask for semantic segmentation Element-wise multiplication is performed to obtain the enhancement weights of the target edges; then, these enhancement weights are added to the base weight 1 to generate the final pixel-level attention map. : S233 will use the pixel-level attention map Compared with the original semantic segmentation binary mask Element-wise multiplication yields the enhanced target mask. : S234 for enhancing mask Perform subpixel-level edge contour extraction.