Vehicle lamp control system and method based on pedestrian perception and dynamic shielding

Abstract

The invention discloses a vehicle lamp control system and method based on pedestrian perception and dynamic shielding, and relates to the field of vehicle control, the vehicle lamp control system comprises a perception module, a control processing module and a light beam projection module; the sensing module is used for identifying a pedestrian target and accurately calculating the position, the movement track and the eye space coordinates of the pedestrian relative to the vehicle; the control processing module is used for identifying pedestrians based on the data acquired by the sensing module, positioning eye areas of the pedestrians and generating a pixel-level control instruction; and the light beam projection module is used for receiving the pixel-level control instruction, projecting a basic illumination light beam and dynamically forming a dynamic shielding area corresponding to the eye area in the light beam. Through pixel-level dynamic shielding of the scheme of the invention, the core contradiction that anti-glare and visibility protection cannot be achieved at the same time in the prior art is solved, pedestrian glare blindness is avoided, it is ensured that a driver can clearly recognize pedestrians, and the pedestrian accident rate at night is remarkably reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of vehicle control, in particular to a vehicle light control system based on pedestrian perception and dynamic shielding, a vehicle light control method based on pedestrian perception and dynamic shielding, an electronic device, a storage medium and a vehicle platform. BACKGROUND

[0002] There are certain deficiencies in the existing vehicle light application scenarios: for example, when meeting pedestrians at night, the glare caused by high beams of cars is a major safety hazard. For the above application scenarios, the existing technical solutions have obvious deficiencies:

[0003] 1) Basic automatic vehicle light (such as CN223297744U): only can automatically turn on or off the vehicle light according to the intensity of the ambient light, cannot identify specific targets (such as pedestrians), and cannot adjust the beam shape.

[0004] 2) Static or simple dynamic shielding vehicle light (such as AFS, Adaptive Front-lighting System): only can perform basic beam angle switching (such as curve lighting) according to the steering wheel angle and vehicle speed, lacks pixel-level precise shielding capability for specific obstacles on the road, especially pedestrians.

[0005] 3) General adaptive projection vehicle light (such as CN120656030A): although it uses projection technology and adaptive algorithms, its design goal is usually general road condition adaptation (such as lane line projection), and it does not specifically target the special safety problem of "pedestrian glare", and lacks an active interaction mechanism with traffic participants (such as pedestrians).

[0006] The core defect of the existing technical solutions is that they cannot avoid direct light on the eyes of pedestrians while ensuring sufficient illumination of the body of the pedestrian and the surrounding area, i.e. the contradiction between "avoiding glare" and "ensuring visibility" has not been effectively solved.

[0007] Therefore, a vehicle light control strategy based on pedestrian perception and dynamic shielding is needed, which can accurately perceive the position and posture of pedestrians through multi-sensor fusion technology, and control a high-resolution beam projection unit to generate a dynamic shielding area in the illumination beam that matches the eye position of the pedestrian, thereby eliminating the glare of the pedestrian while ensuring the overall visibility of the pedestrian, and further improving road traffic safety. SUMMARY

[0008] The purpose of the present application is to provide a vehicle light control system based on pedestrian perception and dynamic shielding, a vehicle light control method based on pedestrian perception and dynamic shielding, an electronic device, a storage medium and a vehicle platform, which at least solve one of the technical problems.

[0009] The problem that the pedestrian's body and the surrounding area cannot be ensured to obtain sufficient illumination while avoiding direct sunlight from shining on the pedestrian's eyes, that is, the contradiction between "avoiding glare" and "ensuring visibility" cannot be effectively solved.

[0010] The present application provides the following solutions:

[0011] According to a first aspect of the present application, an intelligent vehicle lamp system based on pedestrian perception and dynamic shielding is provided, comprising:

[0012] a perception module, a control processing module, and a light beam projection module;

[0013] The perception module is configured to identify a pedestrian target and accurately calculate the position, motion trajectory, and eye spatial coordinates of the pedestrian relative to the vehicle.

[0014] The control processing module is configured to identify the pedestrian and locate the eye region based on the data collected by the perception module, and generate pixel-level control instructions.

[0015] The light beam projection module is configured to receive the pixel-level control instructions, project a basic illumination light beam, and dynamically form a dynamic shielding area corresponding to the eye region within the light beam.

[0016] Further, the perception module comprises:

[0017] an image acquisition unit, a ranging radar unit, and an inertial navigation system;

[0018] The image acquisition unit identifies the pedestrian contour through a pre-set target detection model.

[0019] The ranging radar unit provides pedestrian distance, speed, and azimuth angle data.

[0020] The inertial navigation system compensates for the impact of vehicle motion on positioning.

[0021] The perception module realizes multi-sensor data fusion through Kalman filtering, and dynamically adjusts the weight of each sensor data according to environmental conditions.

[0022] The image acquisition unit includes a camera, the ranging radar unit includes a millimeter wave radar, and the inertial navigation system includes an IMU.

[0023] According to a second aspect of the present application, a vehicle lamp control method based on pedestrian perception and dynamic shielding is provided, comprising a perception step, a pedestrian eye positioning step, and a dynamic shielding light beam step.

[0024] The perception step: through multi-sensor cooperative data acquisition, identify the pedestrian target and obtain the pedestrian position, motion trajectory, and related feature data.

[0025] Pedestrian eye localization steps: Based on multi-sensor fusion data and deep learning algorithms, locate the pedestrian's eye area and calculate its spatial coordinates, while predicting the pedestrian's future movement trajectory;

[0026] Dynamic beam shading steps: Based on eye coordinates and predicted trajectory, control commands are generated to control the beam projection unit to form a dynamic shading zone while maintaining effective illumination of the pedestrian's body and surrounding area.

[0027] Furthermore, the sensing step includes data preprocessing:

[0028] Data preprocessing includes denoising / enhancing, point cloud filtering / segmentation, and dynamically adjusting the detection confidence threshold based on ambient light and weather conditions for the acquired image data.

[0029] Furthermore, the pedestrian eye localization process includes the transformation between the image coordinate system and three-dimensional spatial coordinates:

[0030] Among them, beam response delay compensation is performed by combining vehicle speed and steering angle.

[0031] Furthermore, in the dynamic beam occlusion step, the beam shape is adaptively adjusted according to the scene:

[0032] In the pedestrian lateral movement scenario, an elliptical beam is used, including a beam width that is dynamically adjusted as the pedestrian distance changes;

[0033] The intersection scene uses a fan-shaped beam, including a avoidance path covering a 15° range to the left / right of pedestrians.

[0034] Furthermore, it also includes steps to eliminate secondary risks:

[0035] Among them, the predicted reflection area is analyzed by combining the intensity of millimeter-wave radar reflection signal and camera texture;

[0036] The light intensity in the predicted reflection area will be attenuated to <5%;

[0037] When an obstacle or fall hazard appears behind or to the side of a pedestrian, a warning light spot with a projection intensity of 50% will be projected onto the ground corresponding to the obstacle or fall hazard.

[0038] According to a third aspect of the present invention, an electronic device is provided, comprising: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus;

[0039] The memory stores a computer program, which, when executed by the processor, causes the processor to perform steps such as a vehicle lighting control method based on pedestrian perception and dynamic occlusion.

[0040] According to a fourth aspect of the present invention, a computer-readable storage medium is provided storing a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform steps such as those of a vehicle lighting control method based on pedestrian perception and dynamic occlusion.

[0041] According to a fifth aspect of the present invention, a vehicle platform is provided, comprising:

[0042] Electronic devices for implementing steps such as vehicle lighting control methods based on pedestrian perception and dynamic occlusion;

[0043] The processor runs a program that, when running, executes steps such as a vehicle lighting control method based on pedestrian perception and dynamic occlusion, based on data output from the electronic device.

[0044] Storage medium for storing programs that, when running, execute steps such as vehicle lighting control methods based on pedestrian perception and dynamic occlusion in response to data output from electronic devices.

[0045] The above solution achieves the following beneficial technical effects:

[0046] This application solves the core contradiction of the prior art that "anti-glare" and "visibility" cannot be achieved simultaneously by using pixel-level dynamic masking. It avoids pedestrians being blinded by glare and ensures that drivers can clearly identify pedestrians, thus significantly reducing the pedestrian accident rate at night.

[0047] This application provides more targeted protection for high-risk traffic participants by adopting a tiered priority strategy for special groups such as children and cyclists, combined with risk weight adjustments for scenarios such as school zones and intersections. This further reduces the risk of accidents in extreme scenarios and improves the adaptability and safety of high-risk scenarios.

[0048] This application enables the system to maintain stable performance in complex environments such as rain, fog, and nighttime by using multi-sensor dynamic weight allocation, adaptive beam shape adjustment, and secondary risk elimination. It also eliminates additional risks caused by water accumulation and reflection, vehicle movement deviation, etc., and enhances environmental and motion state adaptability.

[0049] This application utilizes a two-way interactive design combining V2X communication and road safety guidance lines to enable pedestrians to quickly receive warnings and clearly define their avoidance paths, significantly improving avoidance efficiency and effectively reducing the risk of traffic congestion or collisions caused by pedestrian hesitation or misjudgment, thus optimizing the efficiency of human-vehicle collaboration.

[0050] This application reduces hardware costs by leveraging mature technologies such as DLP, and, in conjunction with a closed-loop learning module, not only meets the conditions for mass production but also continuously improves performance as usage scenarios accumulate, achieving the long-term value of "becoming smarter with longer use" and ensuring the long-term optimization and mass production feasibility of the system.

[0051] This application utilizes a fully automated perception-decision-adjustment process that eliminates the need for manual driver intervention. It employs gradual light intensity attenuation to avoid driver discomfort caused by sudden beam changes, while also eliminating secondary problems such as reflected light and beam lag. This further enhances the vehicle's driving safety and reduces the driver's workload and secondary risks. Attached Figure Description

[0052] Figure 1 This is a structural diagram of a vehicle lighting control system based on pedestrian perception and dynamic occlusion provided by one or more embodiments of the present invention.

[0053] Figure 2 This is a flowchart of a vehicle headlight control method based on pedestrian perception and dynamic occlusion provided by one or more embodiments of the present invention.

[0054] Figure 3 This is a schematic diagram of the hardware components and signal connection framework provided in a specific embodiment of the present invention.

[0055] Figure 4 This is a schematic diagram of the data processing flow of a pedestrian perception and positioning algorithm provided in a specific embodiment of the present invention.

[0056] Figure 5 This is a schematic diagram of the dynamic beam blocking principle provided in a specific embodiment of the present invention.

[0057] Figure 6 This is a schematic diagram of a vehicle headlight-pedestrian V2X interaction scenario provided in a specific embodiment of the present invention.

[0058] Figure 7 This is a schematic diagram of the principle framework of a closed-loop optimization system provided in a specific embodiment of the present invention.

[0059] Figure 8 This is a block diagram of an electronic device structure for a vehicle headlight control method based on pedestrian perception and dynamic occlusion provided in one or more embodiments of the present invention. Detailed Implementation

[0060] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] Figure 1 This is a structural diagram of a vehicle lighting control system based on pedestrian perception and dynamic occlusion provided by one or more embodiments of the present invention.

[0062] like Figure 1The intelligent vehicle lighting system shown, based on pedestrian perception and dynamic occlusion, includes:

[0063] The module consists of a sensing module, a control processing module, and a beam projection module.

[0064] The perception module is used to identify pedestrian targets and accurately calculate the pedestrian's position, movement trajectory, and eye spatial coordinates relative to the vehicle;

[0065] The control processing module is used to identify pedestrians and locate their eye areas based on the data collected by the perception module, and generate pixel-level control commands.

[0066] The beam projection module receives pixel-level control commands, projects a basic illumination beam, and dynamically forms a dynamic occlusion area within the beam that corresponds to the eye region.

[0067] In this embodiment, the sensing module includes:

[0068] Image acquisition unit, ranging radar unit, and inertial navigation system;

[0069] The image acquisition unit identifies pedestrian outlines using a pre-set target detection model;

[0070] The ranging radar unit provides pedestrian distance, speed, and azimuth data;

[0071] Inertial navigation systems compensate for the impact of vehicle motion on positioning;

[0072] The sensing module achieves multi-sensor data fusion through Kalman filtering and dynamically adjusts the weights of each sensor's data according to environmental conditions.

[0073] The image acquisition unit includes a camera, the ranging radar unit includes a millimeter-wave radar, and the inertial navigation system includes an IMU.

[0074] In this embodiment, the control processing module includes:

[0075] Built-in multi-task deep learning recognition algorithm;

[0076] The multi-task deep learning recognition algorithm includes a YOLOv5+Transformer hybrid architecture, which is used to implement multi-task joint training for pedestrian detection, pose recognition and motion feature extraction;

[0077] Among them, the detection speed of YOLOv5 is controlled to be ≥30FPS, and the Transformer is used to extract pedestrian pose and motion features.

[0078] In this embodiment, the control processing module further includes:

[0079] Built-in classification algorithm;

[0080] The built-in classification algorithm is used to classify pedestrians into categories such as people, children, and cyclists based on their posture and motion feature regions, and to allocate different beams to adjust priorities according to the category;

[0081] Among them, children and pedestrians trigger the highest priority beam adjustment;

[0082] Cyclists should extend the duration of the beam obstruction.

[0083] In this embodiment, the control processing module further includes:

[0084] Built-in beam control decision algorithm;

[0085] The beam control decision algorithm includes generating control commands based on the PPO reinforcement learning algorithm and the results of pedestrian motion state modeling.

[0086] Pedestrian motion state modeling includes using the OpenPose pose recognition algorithm and the LSTM time series prediction model to predict the pedestrian's position and direction of movement within a preset time interval in the future.

[0087] Specifically, PPO stands for Proximal Policy Optimization, a mainstream approach to "policy gradient" algorithms in the field of reinforcement learning. It was proposed by OpenAI in 2017 and is one of the most widely implemented reinforcement learning algorithms in the industry.

[0088] 1. A simple explanation of the core logic

[0089] The PPO algorithm can be likened to a "self-correcting headlight strategy tuner":

[0090] Its goal is to find the optimal headlight control strategy (such as how to adjust the shading area and light intensity) through "trial and error".

[0091] The core principle is "small-step iteration and steady progress": each time the strategy is adjusted, the difference between the new strategy and the old strategy is limited (achieved by "Clip" operation) to avoid making too extreme changes all at once and causing the system to go out of control (such as the beam suddenly going completely dark / all bright).

[0092] Compared to traditional reinforcement learning algorithms (such as PG and TRPO), PPO solves the problems of unstable training and easy crashes of the PG algorithm, and is simpler to implement and has lower computational cost than TRPO, making it very suitable for computing scenarios in automotive embedded systems.

[0093] 2. Core Mathematical Logic (Simplified Version)

[0094] The core of PPO is the "editing objective function," which essentially adds a "brake" to strategy updates.

[0095] LCLIP(θ)=E^t[min(rt(θ)A^t,clip(rt(θ),1−ϵ,1+ϵ)A^t)]

[0096] rt(θ): The ratio of the probability of the new strategy to that of the old strategy (measures the magnitude of the strategy change);

[0097] ϵ: Editing factor (usually 0.2), limiting strategy variation to no more than ±20%;

[0098] A^t: Advantage function (measures the "goodness" of the current action, such as blocking a pedestrian's eyes but preserving the body's visibility is a "good action" and will be given a positive reward).

[0099] Simply put: if the difference between the new strategy and the old strategy exceeds 20%, it will be "cut" back to a safe range to ensure the stability of the training process.

[0100] In this application, PPO is the core of the "beam control decision module," responsible for converting the input from the sensing module into optimal beam control commands. The specific workflow is as follows:

[0101] 1. First step: Define the "reward function" (the optimization objective of the algorithm)

[0102] The reward function is the "behavioral guideline" of PPO, directly determining the direction in which the algorithm will optimize. The core reward rule of this system is as follows:

[0103] Behavior / State Reward / Punishment Purpose Pedestrian eye light intensity < 10% +10 (positive reward) Core anti-glare Pedestrian body light intensity > 50% +8 (positive reward) Guarantee driver visibility Beam adjustment lag < 0.1 seconds +5 (positive reward) Real-time response to pedestrian movement Reflection zone (water accumulation / reflection object) light intensity > 5% -15 (negative punishment) Eliminate secondary glare Children / cyclists not triggering high priority -20 (negative punishment) Strengthen protection for high-risk groups

[0104] 2. Second step: Define the input / output (the algorithm's "perception" and "action")

[0105] Input (state space): All data collected by the sensing module, including:

[0106] Pedestrian 3D coordinates, movement speed / trajectory (camera + radar);

[0107] Pedestrian type (adult / child / cyclist);

[0108] Environmental parameters (rain / fog / night / light intensity);

[0109] Scene parameters (school area / intersection / ordinary road);

[0110] Vehicle status (vehicle speed, steering angle, IMU compensation data).

[0111] Output (Action Space): Control commands for the beam projection module, including:

[0112] The pixel coordinates and size of the dynamic occlusion area;

[0113] Light intensity parameters for each area (eye / body / guide line);

[0114] Beam shape (elliptical / 15° fan-shaped);

[0115] Light intensity attenuation method (exponential progressive).

[0116] 3. The third step: Training and iteration (the "learning" process of the algorithm)

[0117] Initialization strategy: First, set a basic set of beam control rules (such as default eye occlusion and body retaining 50% light intensity).

[0118] Interactive trial and error: The system runs in real road conditions and collects "status-action-reward" data;

[0119] Strategy Update: PPO updates the strategy using the clipping objective function, making minor adjustments to control instructions to ensure that each update is within a safe range;

[0120] Closed-loop optimization: By combining feedback from the closed-loop learning module (driver satisfaction, pedestrian behavior data), the reward function and strategy are continuously iterated to achieve "the more it is used, the smarter it becomes".

[0121] Specifically, 1. OpenPose: Extracts the "dynamic pose features" of pedestrians.

[0122] OpenPose is a classic multi-person pose estimation algorithm in the field of computer vision. Its core capability is to accurately identify the two-dimensional / three-dimensional coordinates of pedestrian body key points (such as head, eyes, shoulders, limbs, and feet) from video frames.

[0123] In layman's terms: it can "see" whether a pedestrian is "walking, running, stopping, or turning around" at any given moment, and can even pinpoint the eye coordinates that we care about most.

[0124] Core value: Compared to simple pedestrian bounding box detection (such as YOLO which only outlines the pedestrian's range), OpenPose can extract more fine-grained pose features (such as a pedestrian stepping with their left leg or leaning forward), which are key to predicting the direction of movement.

[0125] Output: Coordinates of 18 key points of the pedestrian in each frame (e.g., eyes: (x1, y1), head: (x2, y2), right foot: (x3, y3)...).

[0126] 2. LSTM: Processes time-series data to predict "future motion trajectories".

[0127] LSTM (Long Short-Term Memory) is a deep learning model specifically designed for time series prediction. It can remember the "state of a past period of time" and predict the future based on this.

[0128] In layman's terms: It's like a "memory master" that can remember the movement trajectory of a pedestrian 10 frames (e.g., 0.3 seconds) before determining whether the pedestrian is "going straight, turning left, or crossing the road," and then predicting where they will be in the next 3 seconds.

[0129] Core value: Pedestrian movement is a continuous temporal process. Ordinary CNN models can only process single frame static data, while LSTM can capture "movement trends" and avoid prediction lag.

[0130] Suitable scenarios: Perfectly solves the need for "adjusting the beam obscuration area in advance" in vehicle headlight systems (for example, when a pedestrian is crossing the road, adjust the position of the obscuration area in advance to avoid losing the beam).

[0131] 3. The necessity of combining the two

[0132] OpenPose alone can only tell you "the pedestrian's current posture", and LSTM alone can only predict "simple position sequences". When combined, it can predict based on multi-dimensional features of "posture + position", with an accuracy improvement of more than 30% compared to simply predicting position. It is especially suitable for complex motion scenarios such as pedestrians suddenly changing direction or turning around.

[0133] OpenPose is responsible for extracting fine-grained pose keypoints of pedestrians (especially the eyes), providing core features for trajectory prediction;

[0134] LSTM utilizes temporal modeling capabilities to predict a pedestrian's position and direction of motion over the next 3 seconds based on a sequence of poses / positions from consecutive frames.

[0135] The core value of combining the two is to predict pedestrian trajectories in advance, accurately and stably, so that the dynamic masking area of ​​the headlights can "predict pedestrian movement" rather than passively follow it, thus greatly improving the anti-glare effect.

[0136] In this embodiment, the beam projection module includes:

[0137] Based on DLP digital light processing projection technology, a non-uniform light intensity distribution is implemented through a micromirror array;

[0138] In the dynamic shading area, the light intensity is reduced to <10%, while the light intensity in other areas of the pedestrian's body is retained at 50%, and the light intensity decreases exponentially from the center to the edge.

[0139] In this embodiment, a vehicle-to-everything (V2X) communication module is also included:

[0140] The vehicle-to-everything (V2X) communication module is used to communicate with pedestrian smart terminals and send safety warnings, while simultaneously commanding the beam projection module to project text or graphic prompts onto the road surface in front of the pedestrian.

[0141] In this embodiment, a closed-loop learning module is also included:

[0142] The closed-loop learning module is used to record the beam adjustment effect, including glare complaint rate, driver intervention, and pedestrian avoidance behavior data, and continuously optimize the beam control decision algorithm parameters.

[0143] In this embodiment, the beam projection module further includes:

[0144] Replace it with a high-resolution LED array or laser scanning LBS strategy;

[0145] The ranging radar unit has been replaced with a LiDAR (Light Detection and Ranging) strategy.

[0146] In this embodiment, the beam strategy of the beam projection module includes regional dynamic occlusion:

[0147] The safety zone of this vehicle retains full-intensity lighting;

[0148] Within the pedestrian safety zone, the light intensity in the area corresponding to the eye height range of 80-170cm is reduced to <10%;

[0149] The pedestrian guidance area features a safety guide line with a projection width of 50cm and a length of ≥5 meters, with a light intensity of 20%.

[0150] Figure 2 This is a flowchart of a vehicle headlight control method based on pedestrian perception and dynamic occlusion provided by one or more embodiments of the present invention.

[0151] like Figure 2 The vehicle headlight control method based on pedestrian perception and dynamic occlusion shown includes: a perception step, a pedestrian eye localization step, and a dynamic occlusion beam step;

[0152] Perception step S1: Collect data through multiple sensors to identify pedestrian targets and obtain pedestrian location, movement trajectory and related feature data;

[0153] Pedestrian eye localization step S2: Based on multi-sensor fusion data and deep learning algorithms, locate the pedestrian's eye area and calculate its spatial coordinates, while predicting the pedestrian's future movement trajectory;

[0154] Step S3 of dynamic beam masking: Generate control commands based on eye coordinates and predicted trajectory to control the beam projection unit to form a dynamic masking zone while maintaining effective illumination of the pedestrian's body and surrounding area.

[0155] In this embodiment, the sensing step includes data preprocessing:

[0156] Data preprocessing includes denoising / enhancing, point cloud filtering / segmentation, and dynamically adjusting the detection confidence threshold based on ambient light and weather conditions for the acquired image data.

[0157] In this embodiment, the pedestrian eye localization step includes the transformation between image coordinate system and three-dimensional spatial coordinate system:

[0158] Among them, beam response delay compensation is performed by combining vehicle speed and steering angle.

[0159] In this embodiment, during the dynamic beam occlusion step, the beam shape is adaptively adjusted according to the scene:

[0160] In the pedestrian lateral movement scenario, an elliptical beam is used, including a beam width that is dynamically adjusted as the pedestrian distance changes;

[0161] The intersection scene uses a fan-shaped beam, including a avoidance path covering a 15° range to the left / right of pedestrians.

[0162] In this embodiment, a secondary risk elimination step is also included:

[0163] Among them, the predicted reflection area is analyzed by combining the intensity of millimeter-wave radar reflection signal and camera texture;

[0164] The light intensity in the predicted reflection area will be attenuated to <5%;

[0165] When an obstacle or fall hazard appears behind or to the side of a pedestrian, a warning light spot with a projection intensity of 50% will be projected onto the ground corresponding to the obstacle or fall hazard.

[0166] In this embodiment, a human-vehicle interaction step is also included:

[0167] Among them, safety warnings are sent to pedestrian smart terminals through the V2X communication module, while text or graphic prompts are projected on the road surface.

[0168] In this embodiment, during the dynamic beam shielding step, the beam strategy priority is dynamically adjusted based on the road environment;

[0169] Among them, the beam strategy for road environments in school zones and intersections has a higher priority than the beam strategy for general road environments.

[0170] In this embodiment, during the sensing step, the data processing priority is dynamically adjusted based on the weather environment;

[0171] Among these measures, the weight of camera data is increased under low light conditions, and the weight of radar data is enhanced under rainy and foggy weather conditions.

[0172] In this embodiment, a closed-loop optimization step is also included:

[0173] Continuously record driver satisfaction, pedestrian avoidance behavior data, and environmental condition parameters;

[0174] Based on continuously recorded data on driver satisfaction, pedestrian avoidance behavior, and environmental conditions, the beam control algorithm and shading strategy are continuously optimized.

[0175] In this embodiment, the deep learning algorithm also includes YOLOv8 or other deep learning models with equivalent detection accuracy and speed.

[0176] It is worth noting that although this system / device only discloses the above-mentioned modules / units, it does not mean that this system / device is limited to the above-mentioned basic functional modules. On the contrary, what this invention intends to express is that, based on the above-mentioned basic functional modules, those skilled in the art can add one or more functional modules in combination with the prior art to form an infinite number of embodiments or technical solutions. That is to say, this system is open rather than closed. It cannot be assumed that the scope of protection of the claims of this invention is limited to the above-disclosed basic functional modules just because this embodiment only discloses a few basic functional modules.

[0177] In one specific embodiment, an intelligent vehicle lighting system based on pedestrian perception and dynamic occlusion is disclosed, such as... Figure 3 As shown, its core objective is to accurately perceive the location and posture of pedestrians through multi-sensor fusion technology, and control the high-resolution beam projection unit to generate a dynamic, real-time shading area within the illumination beam that matches the pedestrian's eye position. This eliminates pedestrian glare while ensuring overall visibility, further enhancing road traffic safety. This includes:

[0178] The perception module includes an image acquisition unit (camera) and a ranging radar unit (millimeter-wave radar), which work together to identify pedestrian targets and accurately calculate their position relative to vehicles, movement trajectory, and spatial coordinates of key parts (especially the eyes). For example... Figure 4 As shown.

[0179] Control and processing module: Connected to the perception module, it incorporates a pedestrian recognition and localization algorithm (based on the YOLOv5 model) and a beam control decision algorithm (based on the PPO reinforcement learning algorithm). It determines the pedestrian's eye region in real time based on perception data and generates pixel-level control commands.

[0180] Beam projection module: It adopts digital light processing (DLP) projection technology, and after receiving control commands, it projects a basic illumination beam and dynamically forms a shielding area within the beam that coincides with the position of the pedestrian's eyes.

[0181] Vehicle-to-Everything (V2X) communication module: Used to communicate with pedestrian smart terminals, send safety warnings to pedestrians, and instruct the beam projection module to project text or graphic prompts onto the road surface in front of pedestrians. For example... Figure 6 As shown.

[0182] Closed-loop learning module: Records the beam adjustment effect (such as glare complaint rate and driver intervention), and continuously optimizes the beam control decision algorithm parameters.

[0183] This also includes multi-sensor fusion positioning algorithms:

[0184] A multi-sensor fusion positioning algorithm combining millimeter-wave radar, a front-facing camera, and an inertial navigation system (IMU) is employed, and real-time dynamic positioning of pedestrians is achieved through Kalman filtering.

[0185] Millimeter-wave radar: provides high-precision data on pedestrian distance, speed, and azimuth.

[0186] Front-facing camera: It uses the YOLOv5 object detection model to identify pedestrian outlines and converts them into three-dimensional spatial coordinates by combining the image coordinate system.

[0187] IMU: Compensates for the impact of vehicle motion on positioning, eliminating positioning deviations caused by vehicle acceleration and steering.

[0188] Dynamic weight allocation: The weights of each sensor data are dynamically adjusted according to environmental conditions (such as rainy / foggy weather, light intensity). For example, the weight of the camera is increased under low light conditions, and the weight of the radar is increased under rainy / foggy weather.

[0189] Pedestrian motion state modeling: By using pedestrian pose recognition (OpenPose algorithm) and motion trajectory analysis (LSTM time series prediction model), the position and direction of movement of pedestrians within the next 3 seconds are predicted.

[0190] Beam response delay compensation: By combining vehicle speed and steering angle, the timing of beam adjustment is predicted to avoid positioning lag caused by system response delay.

[0191] Dynamic occlusion region generation: Dynamic occlusion regions (such as the path areas that pedestrians may cross) are generated based on the predicted pedestrian trajectory, rather than fixed occlusion regions.

[0192] This also includes multi-task deep learning recognition algorithms:

[0193] A hybrid architecture of YOLOv5 and Transformer is adopted to achieve multi-task joint training of pedestrian detection, pose recognition and motion feature extraction.

[0194] YOLOv5: Responsible for quickly detecting pedestrian targets (detection speed ≥30FPS).

[0195] Transformer: Extracts pedestrian pose and motion features (such as OpenPose keypoint coordinates + motion vectors), improving the recognition ability of small targets (such as children) and occluded scenes.

[0196] Adaptive threshold adjustment: The detection confidence threshold is dynamically adjusted based on ambient light (ambient light sensor data) and weather conditions (millimeter-wave radar reflection signal intensity) to avoid false detections and missed detections.

[0197] Classification algorithm: Based on pedestrian posture and motion characteristics (such as stride length and speed), it distinguishes different categories such as adults, children, and cyclists.

[0198] Children and pedestrians: Trigger the highest priority beam adjustment (such as immediate occlusion + enhanced display of safety guide lines).

[0199] Cyclists: Extend the duration of beam blocking (e.g., dynamically adjust according to bicycle speed).

[0200] Risk weight allocation: Dynamically adjust the priority of beam strategy based on pedestrian type and road environment (such as school zones and intersections).

[0201] This also includes beam strategies:

[0202] 1) Regional dynamic occlusion strategy

[0203] Safety zone of this vehicle: Maintain full-intensity illumination within the driver's field of vision (dynamically adjusting distance based on vehicle speed).

[0204] Pedestrian safety zone:

[0205] Area aimed at pedestrians' eyes (height range 80-170cm): light intensity attenuated to <10% to avoid glare.

[0206] Other areas of the pedestrian's body: retain 50% of the light intensity to ensure visibility.

[0207] Pedestrian guidance area: Project safety guidance lines (50cm wide, ≥5m long) with a light intensity of 20% to avoid interfering with pedestrians' vision.

[0208] Non-uniform light intensity distribution: Light intensity differences in different regions are achieved through micromirror array control of DLP projection.

[0209] 2) Adaptive beam shape optimization

[0210] Elliptical beam: For scenarios where pedestrians are moving laterally, the beam width is dynamically adjusted (e.g., the width decreases as the pedestrian approaches).

[0211] Fan-shaped beam: In intersection scenarios, it covers possible pedestrian avoidance paths (such as within a 15° range to the left / right).

[0212] Progressive shading: Light intensity decreases exponentially from the center to the edge (to avoid driver discomfort caused by sudden changes).

[0213] 3) Secondary risk elimination strategy

[0214] Local light intensity suppression: In areas where pedestrians may reflect light (such as waterlogged roads or metal signs), the light intensity is locally reduced (attenuated to <5%) by DLP projection.

[0215] Dynamic obstacle avoidance projection: When an obstacle (such as a curb or billboard) appears behind a pedestrian, a warning light spot (50% light intensity) is projected in advance to prompt the pedestrian to avoid it.

[0216] Environmental reflection prediction: By combining the intensity of millimeter-wave radar reflection signals and camera texture analysis, the reflection area is predicted and the light intensity is actively suppressed.

[0217] This also includes a closed-loop optimization control system:

[0218] Closed-loop optimization control system includes,

[0219] Vehicle safety feedback: The system records the driver's satisfaction with the vehicle's safety and continuously optimizes the beam strategy.

[0220] Pedestrian behavior feedback: Optimize the display effect of safety guide lines by using pedestrian avoidance behavior data.

[0221] Environmental adaptability optimization: Automatically adjusts beam parameters according to environmental conditions (rain, fog, night, etc.).

[0222] This also includes the mechanism by which the technical effect of this embodiment is generated:

[0223] The perception module accurately locates pedestrian eye coordinates by fusing data from cameras and radar; the control processing module identifies pedestrians and calculates eye positions based on the YOLOv5 algorithm, generating occlusion commands through the PPO algorithm; the beam projection module uses DLP technology to precisely project the occlusion area into the illumination beam, effectively "erasing" glare. V2X interaction enhances human-vehicle communication, and the closed-loop learning module continuously optimizes the system, improving long-term performance. Figure 7 As shown.

[0224] like Figure 5 As shown, the key advantages of the technical solution in this embodiment compared with the prior art are:

[0225] 1) Enhanced vehicle safety: Ensures clear visibility for the driver and reduces driving risks.

[0226] 2) Improved pedestrian guidance efficiency: Safety guidance lines improve pedestrian avoidance efficiency and reduce traffic congestion.

[0227] 3) Elimination of secondary risks: Eliminate secondary risks such as reflected light and reduce the risk of accidents to pedestrians.

[0228] 4) Pedestrian visibility protection: Avoid direct sunlight on pedestrians' eyes while ensuring pedestrian visibility and reducing pedestrian accident rate.

[0229] 5) Environmental adaptability: In complex environments such as rain, fog, and night, the system automatically optimizes beam parameters to ensure the best lighting effect.

[0230] 6) Cost optimization: Based on existing DLP projection technology, reduce system hardware cost investment and make mass production feasible.

[0231] 7) Human-vehicle collaboration optimization: Improve overall traffic efficiency and safety through two-way interaction between vehicle lights and pedestrians.

[0232] Figure 8 This is a block diagram of an electronic device structure for a vehicle headlight control method based on pedestrian perception and dynamic occlusion provided in one or more embodiments of the present invention.

[0233] like Figure 8 As shown, this application provides an electronic device, including: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0234] The memory stores a computer program that, when executed by the processor, causes the processor to perform steps of a vehicle lighting control method based on pedestrian perception and dynamic occlusion.

[0235] This application also provides a computer-readable storage medium storing a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the steps of a vehicle lighting control method based on pedestrian perception and dynamic occlusion.

[0236] This application also provides a vehicle platform, including:

[0237] Electronic equipment for implementing a vehicle lighting control method based on pedestrian perception and dynamic occlusion;

[0238] The processor runs a program that, when running, executes steps of a vehicle lighting control method based on pedestrian perception and dynamic occlusion based on data output from electronic devices.

[0239] Storage medium for storing programs that, when running, execute steps of a vehicle lighting control method based on pedestrian perception and dynamic occlusion in response to data output from electronic devices.

[0240] The communication bus mentioned in the above electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not indicate that there is only one bus or one type of bus.

[0241] The electronic device comprises a hardware layer, an operating system layer running on top of the hardware layer, and an application layer running on the operating system. The hardware layer includes hardware such as a central processing unit (CPU), a memory management unit (MMU), and memory. The operating system can be any one or more computer operating systems that control the electronic device through processes, such as Linux, Unix, Android, iOS, or Windows. Furthermore, in this embodiment of the invention, the electronic device can be a smartphone, tablet computer, or other handheld device, or a desktop computer, portable computer, or other electronic device; there is no particular limitation in this embodiment.

[0242] In this embodiment of the invention, the executing entity for electronic device control can be an electronic device itself, or a functional module within an electronic device capable of calling and executing a program. The electronic device can obtain the firmware corresponding to the storage medium. This firmware is provided by the supplier, and different storage media may have the same or different firmware; no limitation is made here. After obtaining the firmware corresponding to the storage medium, the electronic device can write this firmware into the storage medium; specifically, it burns the firmware corresponding to the storage medium into the storage medium. The process of burning the firmware into the storage medium can be implemented using existing technology, and will not be elaborated upon in this embodiment of the invention.

[0243] Electronic devices can also obtain reset commands corresponding to the storage media. The reset commands corresponding to the storage media are provided by the supplier. The reset commands corresponding to different storage media can be the same or different, and no restrictions are imposed here.

[0244] At this time, the storage medium of the electronic device is a storage medium on which the corresponding firmware has been written. The electronic device can respond to the reset command corresponding to the storage medium on which the corresponding firmware has been written, thereby resetting the storage medium on which the corresponding firmware has been written according to the reset command. The process of resetting the storage medium according to the reset command can be implemented by existing technology and will not be described in detail in this embodiment of the invention.

[0245] For ease of description, the above devices are described separately by function as various units and modules. Of course, in implementing this application, the functions of each unit and module can be implemented in one or more software and / or hardware.

[0246] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the meaning consistent with their meaning in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined.

[0247] For the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0248] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, 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 storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0249] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent vehicle light system based on pedestrian perception and dynamic occlusion, characterized in that, Comprising: a perception module, a control processing module and a light beam projection module; the perception module is used for identifying a pedestrian target and accurately calculating the position, motion trajectory and eye spatial coordinates of the pedestrian relative to the vehicle; the control processing module is used for identifying the pedestrian and positioning the eye region of the pedestrian based on the data collected by the perception module, and generating pixel-level control instructions; the light beam projection module is used for receiving the pixel-level control instructions, projecting a basic lighting light beam and dynamically forming a dynamic shielding area corresponding to the eye region in the light beam.

2. The intelligent vehicle light system based on pedestrian perception and dynamic occlusion of claim 1, wherein, The perception module comprises: an image acquisition unit, a ranging radar unit and an inertial navigation system; the image acquisition unit identifies the pedestrian contour through a preset target detection model; the ranging radar unit provides pedestrian distance, speed and azimuth angle data; the inertial navigation system compensates for the influence of vehicle motion on positioning; the perception module realizes multi-sensor data fusion through Kalman filtering, and dynamically adjusts the weight of each sensor data according to environmental conditions; wherein the image acquisition unit comprises a camera, the ranging radar unit comprises a millimeter wave radar, and the inertial navigation system comprises an IMU.

3. A vehicle headlight control method based on pedestrian perception and dynamic occlusion, characterized in that, Comprising: a perception step, a pedestrian eye positioning step and a dynamic shielding light beam step; the perception step: through multi-sensor cooperative data collection, identifying a pedestrian target and obtaining pedestrian position, motion trajectory and related feature data; the pedestrian eye positioning step: based on multi-sensor fusion data and deep learning algorithm, positioning the eye region of the pedestrian and calculating its spatial coordinates, while predicting the future motion trajectory of the pedestrian; the dynamic shielding light beam step: generating control instructions according to the eye coordinates and the predicted trajectory, controlling the light beam projection unit to form a dynamic shielding area, while maintaining the effective illumination of the pedestrian body and the surrounding area.

4. The method of controlling vehicle lights based on pedestrian perception and dynamic occlusion of claim 3, wherein, The perception step includes data preprocessing: The data preprocessing includes, for the collected image data, denoising / enhancement, point cloud filtering / segmentation, and dynamically adjusting the detection confidence threshold according to the ambient light and weather conditions.

5. The method for car light control based on pedestrian perception and dynamic occlusion of claim 3, wherein, The pedestrian eye positioning step includes conversion of image coordinate system and three-dimensional space coordinates: wherein the light beam response delay compensation is combined with the vehicle speed and steering angle.

6. The method of controlling vehicle lights based on pedestrian perception and dynamic occlusion of claim 3, wherein, In the dynamic shielding light beam step, the light beam shape is adjusted adaptively according to the scene: wherein the pedestrian lateral movement scene adopts an elliptical light beam, including dynamic adjustment of the light beam width with the pedestrian distance; the intersection scene adopts a fan-shaped light beam, including covering the avoidance path within the 15° range on the left / right side of the pedestrian.

7. The method for car light control based on pedestrian perception and dynamic occlusion of claim 3, wherein, It also includes a secondary risk elimination step: wherein the predicted reflection area is analyzed by combining the millimeter wave radar reflection signal strength and the camera texture; the light intensity of the predicted reflection area is attenuated to <5%; when an obstacle or a falling risk object appears behind or on the side of the pedestrian, a warning light spot with a light intensity of 50% is projected on the ground corresponding to the obstacle or the falling risk object.

8. An electronic device, comprising: Comprising: a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory complete mutual communication through the communication bus; A computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the steps of the vehicle light control method based on pedestrian perception and dynamic occlusion according to any one of claims 3 to 7.

9. A computer-readable storage medium, characterized in that, A computer program executable by the electronic device is stored, and when the computer program runs on the electronic device, the electronic device executes the steps of the vehicle light control method based on pedestrian perception and dynamic occlusion according to any one of claims 3 to 7.

10. A vehicle platform characterized by, Comprise: An electronic device for implementing the steps of the vehicle light control method based on pedestrian perception and dynamic occlusion according to any one of claims 3 to 7; A processor, the processor running the program, when the program runs, executes the steps of the vehicle light control method based on pedestrian perception and dynamic occlusion according to any one of claims 3 to 7 from the data output from the electronic device; A storage medium for storing the program, which executes the steps of the vehicle light control method based on pedestrian perception and dynamic occlusion according to any one of claims 3 to 7 when running for data output from the electronic device.

Images