Intelligent vehicle lamp dynamic light interaction control method based on scene recognition

Abstract

The present application belongs to the technical field of intelligent car light control, and particularly relates to a dynamic light interaction control method for intelligent car light based on scene recognition, which comprises the following steps: multi-source traffic scene fusion perception collection, scene grading determination and interaction mode matching, integrated pixel dynamic light language real-time generation, partition anti-dazzle synchronous driving regulation, ambient light adaptive brightness correction, external subject bidirectional interaction feedback response, dynamic light shadow projection output and state feedback, and dynamic interaction light language online learning iteration self-updating, which solves the defects of traditional car light control logic static solidification, one-way interaction, easy to produce glare and poor scene recognition accuracy, realizes the upgrading of car light from a single static light source to a global dynamic interaction system from the control algorithm level, has low response delay and full-scene self-adaptation, and is suitable for high-order automatic driving new energy vehicles equipped with pixel matrix car lights.

Description

Technical Field

[0001] This invention belongs to the field of intelligent vehicle lighting control technology, and in particular relates to an intelligent vehicle lighting dynamic light interaction control method based on scene recognition. Background Technology

[0002] Traditional vehicle headlight control logic only performs fixed-level switching for high and low beams and parking lights. The control logic is simplistic, only providing basic static lighting output and lacking dynamic interactive control capabilities. Existing control methods for headlights with projection capabilities suffer from the following core defects: 1. Low scene recognition and fusion: Relying on a single camera for image acquisition, scene recognition accuracy is low and latency is high, failing to accurately distinguish between pedestrians, oncoming vehicles, intersections, highways, and other differentiated traffic scenarios; 2. Static and fixed light language output: The control logic has built-in fixed text / pattern projection templates, only able to statically project preset signs, lacking a continuous dynamic light language generation mechanism; 3. One-way interaction without a closed loop: Only supports vehicles actively projecting light signals outwards, unable to support reverse interaction recognition and feedback control processes from external entities. Pedestrian waving, oncoming vehicle flashing lights, and roadside V2X warning commands cannot trigger corresponding dynamic light and shadow feedback from the headlights; 4. The lighting and interactive dimming are separated, and the control algorithm schedules the lighting pixels and projection pixels separately with two independent logics. It is impossible to use the pixel array to synchronously complete the driving lighting and interactive projection, which wastes hardware resources. In addition, there is no partitioned anti-glare control process, and the interactive light and shadow can easily cause glare to oncoming drivers. 5. There is no adaptive brightness level logic, and the output power of the interactive light and shadow is not automatically adjusted according to day and night, tunnels, and strong sunlight on sunny days. Publication No. CN 119291715A discloses a method for road anomaly detection and vehicle headlight interaction based on vehicle-mounted LiDAR. This solution relies on a single vehicle-mounted LiDAR to collect environmental point clouds, identifies abnormal road areas such as road damage and obstacles through point cloud clustering, and then maps the abnormal areas to LED headlight pixels using multi-point perspective imaging. Only the corresponding pixels are subjected to fixed brightness enhancement, thereby improving the driver's observation clarity of road obstacles. However, the applicable conditions are narrow, only targeting the illumination enhancement of road obstacles at night. It cannot adapt to all road conditions such as smooth highways, meeting at urban intersections, transition between light and dark in tunnels, and parking self-iteration, making it difficult to meet the diverse needs of vehicle external visualization and safety interaction in the era of intelligent driving.

[0003] Therefore, existing vehicle lighting control methods have not yet formed a complete dynamic interactive control link, and cannot support the upgrading of vehicle lights from "static lighting sources" to a full-domain human-vehicle-road interactive system. Summary of the Invention

[0004] To address the aforementioned technical issues, this invention provides an intelligent vehicle headlight dynamic light interaction control method based on scene recognition. This method aims to solve the problems of existing vehicle headlight control logic, such as single scene recognition, static and fixed light language, unidirectional output, lack of bidirectional interactive closed loop, and failure to consider anti-glare and adaptive dimming. Through multi-source perception fusion, scene hierarchical matching, real-time generation of dynamic light language, external interactive feedback response, and pixel-partitioned anti-glare synchronous control, it achieves parallel output of basic lighting and dynamic light and shadow interaction, constructing a bidirectional dynamic interactive closed loop between people and vehicles, vehicles and vehicles, and vehicles and roads. This completes the intelligent upgrade of vehicle headlights from static light sources to a multi-subject dynamic interactive system.

[0005] This invention provides a method for dynamic light interaction control of intelligent vehicle lights based on scene recognition, specifically including the following steps: S1: Multi-source traffic scene fusion perception acquisition steps, linking vehicle ADAS system, V2X communication unit, and ambient light sensor to synchronously acquire multi-dimensional raw perception data, and perform spatiotemporal alignment fusion processing on the raw perception data to output standardized scene feature map; S2: Scene classification and interaction mode matching steps. A hierarchical dynamic light language strategy library is pre-stored. The library is divided into multiple levels of dynamic interaction templates according to traffic risk level. Based on the scene feature map output in step S1, scene semantics are determined and the corresponding level of dynamic interaction strategy is matched. Simultaneously output lighting partition parameters, interactive projection partition parameters, light and shadow timing control parameters, and target anti-glare shielding pixel coordinates. S3: The integrated pixel dynamic light language real-time generation step, based on the completed interaction strategy, performs unified scheduling of the pixel matrix LED array, and uses the same set of pixel hardware to synchronously generate basic driving static lighting beams and dynamic interactive light and shadow. S4: Zoned anti-glare synchronous drive control step. Based on the anti-glare shielding pixel coordinates output by step S2, the pixel matrix is ​​divided into independent low beam lighting zone, high beam lighting zone, and dynamic interactive projection zone. At the same time, the pixel points directly facing the driver's field of vision are turned off separately, and only the interactive light and shadow output is retained in the gap between two vehicles and the road surface area in front of pedestrians. S5. Ambient light adaptive brightness correction step: Real-time reading of ambient light intensity to establish a brightness mapping curve; S6: External subject two-way interactive feedback response steps, continuously monitor the internal and external interactive trigger signals of the scene feature map, capture the effective trigger signal, and immediately switch the dynamic light language timing and pattern to output feedback dynamic light and shadow, forming a two-way interactive closed loop of "external traffic subject perception signal input → vehicle light dynamic light and shadow feedback output". S7: Dynamic light and shadow projection output and status feedback steps, all pixel driving instructions are sent to the vehicle light driving module, dynamic interactive light and shadow are projected through the optical components, and the current lighting status, interactive light language type and scene recognition results are simultaneously fed back to the vehicle domain controller and V2X platform for subsequent interactive strategy iteration and optimization. S8: Dynamic interactive light language online learning iterative self-updating steps, continuously collect full-cycle scene interaction data, evaluate the light language effect through quantitative models, automatically adjust light and shadow parameters and update the local light language strategy library.

[0006] A further improvement of the present invention is that, in step S1, the original perception data includes pedestrian / non-motorized vehicle identification data from camera images, target distance and speed coordinates from millimeter-wave radar, oncoming vehicle light signals, roadside V2X-issued road condition warning commands, ambient light intensity, and the vehicle speed and road type.

[0007] A further improvement of the present invention is that: in step S2, the multi-level dynamic interactive template includes a pedestrian yielding gradient light carpet, a vehicle meeting anti-glare stream light, a speed limit dynamic digital projection at intersections, and a high-speed collaborative interactive light strip; the scene classification judgment standard is divided into three levels, namely, low-risk high-speed traffic scene, medium-risk urban intersection vehicle meeting scene, and high-risk zebra crossing pedestrian crossing scene, and each level of scene corresponds to a set of independent dynamic light and shadow time sequence gradient logic.

[0008] Differentiated dynamic interaction strategies are matched according to risk level, and an ambient light brightness correction mechanism is superimposed to balance daytime interaction recognition and nighttime anti-glare requirements, making it suitable for all scenarios of high-level autonomous driving pixel matrix vehicle lights. A further improvement of the present invention is that: in step S1, the spatiotemporal alignment and fusion processing method is to unify the timestamps of radar, camera, and V2X data acquisition, complete the multi-sensor spatial coordinate transformation based on the vehicle chassis coordinates, eliminate duplicate invalid targets, and output a standardized scene map containing target category, relative distance, motion trend, and road events.

[0009] By integrating multi-dimensional data from visual radar, V2X, and ambient light, the system achieves accurate scene determination and low latency, solving the problems of large recognition errors and inability to adapt to complex mixed traffic scenarios involving people, vehicles, and roads caused by single sensors, thus providing a reliable decision-making basis for dynamic interaction.

[0010] A further improvement of the present invention is that, in step S3, the dynamic light language generation timing logic is that the interactive light and shadow adopts continuous gradual output, the transition time of a single set of light language patterns is controlled within 10-50ms, and there are no sudden bright or dark changes, so as to avoid visual stimulation to surrounding traffic participants.

[0011] Dynamic light language generation abandons the traditional fixed static projection logic and can continuously adjust the shape, brightness, and flicker rhythm of light and shadow in real time according to the target position, vehicle speed, and lighting conditions. It realizes the core upgrade of vehicle lights from "static light source" to dynamic interactive system from the underlying control logic.

[0012] A further improvement of the present invention is that, in step S6, the trigger signal includes a pedestrian waving action, an oncoming vehicle flashing its high and low beams alternately, and a V2X roadside equipment collaborative interaction command.

[0013] The new steps for external subject signal recognition and light and shadow feedback are complete, no longer limited to one-way vehicle light projection. The algorithm level opens up a two-way interactive link between people and vehicles, vehicles and vehicles and roads. A further improvement of the present invention is that, in step S6, the bidirectional interactive feedback logic is as follows: when a pedestrian's waving action is detected, a flowing soft light carpet is projected from far to near; when an oncoming vehicle's flashing lights are detected, a bidirectional synchronous gradient warning light is switched; and when a V2X road warning command is received, a dynamic warning digital pattern is projected onto the road surface.

[0014] A further improvement of the present invention is that: the S3 step includes a lightweight rendering compression sub-process, a pre-built vector light language template library, real-time rendering based on vector units, frame differential transmission of light and shadow data, and dynamic allocation of domain control rendering computing power according to scene risks.

[0015] A further improvement of the present invention is that, in step S5, the overall grayscale brightness of dynamic interactive light and shadow is increased in high-light daytime scenes, while the peak brightness of light and shadow is linearly reduced in tunnel and low-light nighttime scenes, thus balancing interactive recognition and nighttime driving safety.

[0016] A further improvement of the present invention is that: in step S8, the fine-tuning of the optical language model parameters is only performed when the vehicle is parked or in a low-speed safety scenario, and the iterative calculation is paused in the high-speed driving state.

[0017] Compared with existing technologies, this invention can achieve the following technical effects: 1. This invention generates standardized scene maps by fusing multi-sensor and V2X multi-source perception data, eliminating blind spots and data conflicts of single sensors, and comprehensively collecting multi-dimensional information on people, vehicles, and roads. It has strong scene recognition and provides accurate decision-making basis for lighting interaction; 2. It matches exclusive dynamic light language templates according to traffic risk levels, and differentiates the output lighting and interaction parameters for highways, intersections, and zebra crossings, taking into account both traffic efficiency and road safety; 3. Low hardware cost and lightweight computing power. This invention uses a single set of pixel LEDs to output lighting and interactive light and shadow, saving hardware costs. Combined with lightweight strategies such as vector rendering, frame differential transmission, and high-speed pause iteration, it significantly reduces the onboard computing power occupation; 4. It is equipped with pixel partition anti-glare control and ambient light adaptive brightness correction processes, which independently control the pixel positions of each partition, accurately shielding the vision of oncoming drivers and retaining only the human-vehicle interaction light and shadow area, thus eliminating glare interference while retaining visual interactive prompts on the road surface; 5. The system features full-environment brightness adaptation, dynamically calibrating light and shadow brightness based on ambient light levels. It enhances visibility during the day and reduces peak brightness in low-light scenarios, ensuring clear and soft lighting in various day / night and tunnel environments. 6. This invention overcomes the shortcomings of traditional vehicle lighting control logic, such as static and fixed logic, one-way interaction, glare, and poor scene recognition accuracy. It upgrades vehicle lighting from a single static light source to a full-domain dynamic interactive system at the control algorithm level, resulting in low response latency and full-scene adaptation. It is suitable for high-level autonomous driving new energy vehicles equipped with pixel matrix headlights, adapting to all road conditions, improving the intelligent vehicle's external interaction capabilities, and reducing accident risks. Attached Figure Description

[0018] The illustrative embodiments and descriptions of the present invention are used to explain the present invention and do not constitute an improper limitation of the present invention. Detailed Implementation

[0019] The following will describe the implementation of the present invention in detail with reference to the embodiments, so that the process of how the present invention uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly. Example

[0020] This embodiment provides a scene recognition-based intelligent vehicle headlight dynamic light interaction control method, which runs on an intelligent vehicle domain controller equipped with pixel matrix LED headlights, and is executed in conjunction with millimeter-wave radar, a forward-facing camera, a V2X receiving unit, and an ambient light sensor. The complete loop process is as follows: Example 1: High-speed, low-risk traffic scenario S1: Multi-source traffic scene fusion perception and acquisition, camera identifies no pedestrians or non-motorized vehicles ahead, millimeter-wave radar detects the distance between vehicles in front and behind >80m and the relative speed difference <20km / h, V2X has no road warning instructions, ambient light sensor collects daytime light intensity of 60000 lux and vehicle speed of 95km / h; unify the timestamps of the data collected by each sensor, transform the spatial coordinates based on the chassis coordinate system, remove duplicate road signs and invalid targets in the distance, and output a standardized scene map of high-speed and smooth traffic. S2: Scene classification and interaction mode matching. If the scene is classified as a low-risk high-speed traffic scene, the corresponding dynamic interaction template is matched: basic high beam full-area lighting + minimalist narrow dynamic collaborative interactive light strip; synchronous output lighting zones are fully turned on, interactive projection zones are only narrow road surface areas, and there are no anti-glare shielding pixel coordinates. S3: Integrated pixel dynamic light language real-time generation steps, vector unit calls high-speed collaborative light strip template, frame differential transmission of light and shadow data, dynamic allocation of low rendering computing power; basic far beam and horizontal narrow gradient interactive light strip are generated synchronously, and the light strip single frame switching time is 30ms continuous gradient without flicker. S4: Zoned anti-glare synchronous drive control, no oncoming vehicles at close range, all pixels are lit normally, retaining the full-area road lighting and interactive light strip; S5. Ambient light adaptive brightness correction: In daylight, it enhances the overall grayscale brightness of the interactive light strip and ensures long-distance visibility. S6: External subject two-way interactive feedback response, no pedestrian waving, oncoming vehicle flashing lights, V2X warning trigger signal, maintain basic high-speed light signal; S7: Dynamic light and shadow projection output and status feedback; pixel-driven commands are sent to the headlight module to project high beams and horizontal collaborative light strips; high-speed scene recognition results, collaborative light language types, and normal headlight operation status are fed back to the vehicle domain controller and highway roadside V2X platform. S8: Dynamic interactive light language online learning iterative self-update. The current vehicle speed of 95km / h is considered high speed. Light language model parameter optimization calculation is paused to ensure priority for driving computing power.

[0021] Implementation results: No glare interference during high-speed driving; front and rear vehicles complete collaborative interactive reminders through narrow light strips; the interactive light strips are clearly identifiable under strong daylight; it does not consume the main control computing power of the vehicle, thus improving the safety of high-speed driving.

[0022] Example 2: Medium-risk oncoming traffic scenario at an urban intersection, nighttime environment, with vehicles approaching from both directions. S1: Multi-source traffic scene fusion perception and acquisition, camera identifies oncoming small cars, millimeter-wave radar measures distance of 45m and relative speed of 30km / h, V2X roadside issues intersection meeting warning, ambient light of 120lux, vehicle speed of 22km / h; spatiotemporal alignment and coordinate transformation, outputting urban intersection meeting scene map. S2: Scene classification and interaction mode matching; determine medium-risk intersection meeting scenarios and match the meeting anti-glare streaming template; output low beam full-area illumination, interactive projection zone is the road surface between the two vehicles, and the coordinates of all pixels in the field of vision of the oncoming driver are masked. S3: Integrated pixel dynamic light language real-time generation steps, vector rendering of bidirectional gradient flowing light pattern, switching time 20ms, and synchronous output of low beam illumination beam and bidirectional gradient interactive flowing light. S4: Zoned anti-glare synchronous drive control, based on the shielding coordinates, individually turns off all pixels facing the eyes of the oncoming driver, and only retains the gradient flowing light and shadow interaction on the road surface between the two vehicles; S5. Ambient light adaptive brightness correction: In low light conditions at night, linearly reduce the peak brightness of the trailing light to avoid stimulating the vision of oncoming drivers. S6: External subject two-way interactive feedback response, when the oncoming vehicle briefly flashes its high and low beams alternately (trigger signal), immediately switches to two-way synchronous gradual warning stream light, and after 2 seconds, it returns to the normal oncoming stream light, forming a two-way interactive closed loop; S7: Dynamic light and shadow projection output and status feedback, outputting low beam + middle road warning stream light, uploading the intersection meeting scene and warning light status to the domain controller and intersection V2X; S8: Dynamic interactive light language online learning iteration self-updates. In a low-speed safety scenario with a vehicle speed of 22km / h, the background synchronously collects the interaction data of this vehicle meeting, quantitatively evaluates the light flow recognition effect, fine-tunes the light flow width parameter and stores it in the local strategy library to complete the iteration.

[0023] Implementation results: Completely shields glare from oncoming drivers; the gradual flow of light on the road between the two vehicles clearly conveys the intention to avoid a collision; oncoming vehicles flash their lights and receive a synchronized warning feedback; the brightness at night is soft and not dazzling; the light signal parameters are automatically optimized at low speeds; and the recognition of oncoming traffic interaction at intersections continues to improve.

[0024] Example 3: High-risk pedestrian crossing scenario at a zebra crossing, at dusk at a tunnel exit zebra crossing, a pedestrian prepares to cross the road and raises their hand to signal for yielding. S1: Multi-source traffic scene fusion perception and collection, camera recognizes pedestrian waving action at zebra crossing, millimeter-wave radar pedestrian distance is 12m, V2X roadside zebra crossing pedestrian warning, ambient light sensor illumination of 800 lux (transition between light and dark at tunnel exit), vehicle speed of 10km / h; fusion processing outputs zebra crossing pedestrian crossing high-risk scene map. S2: Scene classification and interaction mode matching; identify high-risk pedestrian scenes and match pedestrian yield gradient light carpet template; output low beam lighting zone and interactive projection zone as zebra crossing area in front of pedestrian, no need to block pixels when there are no oncoming vehicles. S3: Integrated pixel dynamic light language real-time generation steps, retrieve vector light carpet template, frame differential transmission of light and shadow, allocate medium rendering computing power, light carpet gradient switching time 40ms. S4: Zoned anti-glare synchronous drive control, no oncoming vehicles, all interactive pixels are lit normally, only the zebra crossing in front of pedestrians is projected with a flowing light carpet. S5. Ambient light adaptive brightness correction, tunnel exit transition lighting, moderately increase the gray brightness of the light blanket, and take into account the visibility of the dark area of ​​the tunnel and the external twilight light. S6: External subject two-way interactive feedback response, captures pedestrian waving trigger signal, immediately projects a flowing courtesy light carpet from far to near, the light carpet extends synchronously with the pedestrian's movement, conveying the vehicle courtesy signal to the pedestrian; if the roadside V2X pushes pedestrian warning at the same time, superimposes dynamic road warning digital pattern. S7: Dynamic light and shadow projection output and status feedback, low beam + pedestrian flow yielding light carpet projection, upload the zebra crossing high-risk scene and yielding light message status to the vehicle domain control and roadside platform. S8: Dynamic interactive light language online learning iteration self-updates. In low-speed safety scenarios with a vehicle speed of 10km / h, it collects pedestrian interaction feedback data, optimizes the light carpet extension speed parameters, and updates the local light language strategy library.

[0025] Implementation results: Pedestrians can intuitively recognize vehicles' intention to yield, the moving light carpet is soft and not dazzling, and the pattern is clear in the light and dark transition environment of the tunnel; the timing logic of the light carpet is automatically optimized in low-speed scenarios, and the subsequent zebra crossing interaction is more adaptable, which greatly reduces the risk of collision between people and vehicles.

[0026] Example 4: Full Iterative Optimization of Light Language in Parking Scene When a vehicle is parked at a red light at an intersection at a speed of 0 km / h, it continuously collects full-cycle interactive data on passing pedestrians, oncoming vehicles, and roadside warnings. The onboard controller quantitatively evaluates the pedestrian recognition rate and oncoming vehicle response time for different light signals, automatically adjusts parameters such as light and shadow switching duration, brightness, and projection width, and fully updates the multi-level dynamic light signal strategy library. The entire process does not affect the vehicle's basic power consumption while parked, and all iterative calculations are directly paused when driving at high speeds to ensure the priority of driving control computing power.

[0027] The foregoing description illustrates and describes several preferred embodiments of the present invention. However, as previously stated, it should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the inventive concept described herein through the foregoing teachings or techniques or knowledge in related fields. Any modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A method for dynamic light interaction control of intelligent vehicle lights based on scene recognition, characterized in that: Specifically, it includes the following steps: S1: Multi-source traffic scene fusion perception acquisition steps, linking vehicle ADAS system, V2X communication unit, and ambient light sensor to synchronously acquire multi-dimensional raw perception data, and perform spatiotemporal alignment fusion processing on the raw perception data to output standardized scene feature map; S2: Scene classification and interaction mode matching steps. A hierarchical dynamic light language strategy library is pre-stored. The library is divided into multiple levels of dynamic interaction templates according to traffic risk level. Based on the scene feature map output in step S1, scene semantics are determined and the corresponding level of dynamic interaction strategy is matched. Simultaneously output lighting partition parameters, interactive projection partition parameters, light and shadow timing control parameters, and target anti-glare shielding pixel coordinates. S3: The integrated pixel dynamic light language real-time generation step, based on the completed interaction strategy, performs unified scheduling of the pixel matrix LED array, and uses the same set of pixel hardware to synchronously generate basic driving static lighting beams and dynamic interactive light and shadow. S4: Zoned anti-glare synchronous drive control step. Based on the anti-glare shielding pixel coordinates output by step S2, the pixel matrix is ​​divided into independent low beam lighting zone, high beam lighting zone, and dynamic interactive projection zone. At the same time, the pixel points directly facing the driver's field of vision are turned off separately, and only the interactive light and shadow output is retained in the gap between two vehicles and the road surface area in front of pedestrians. S5. Ambient light adaptive brightness correction step: Real-time reading of ambient light intensity to establish a brightness mapping curve; S6: External subject two-way interactive feedback response steps, continuously monitor the internal and external interactive trigger signals of the scene feature map, capture the effective trigger signal, and immediately switch the dynamic light language timing and pattern to output feedback dynamic light and shadow, forming a two-way interactive closed loop of "external traffic subject perception signal input → vehicle light dynamic light and shadow feedback output". S7: Dynamic light and shadow projection output and status feedback steps, all pixel driving instructions are sent to the vehicle light driving module, dynamic interactive light and shadow are projected through the optical components, and the current lighting status, interactive light language type, and scene recognition results are simultaneously fed back to the vehicle domain controller and V2X platform. S8: Dynamic interactive light language online learning iterative self-updating steps, continuously collect full-cycle scene interaction data, evaluate the light language effect through quantitative models, automatically adjust light and shadow parameters and update the local light language strategy library.

2. The intelligent vehicle headlight dynamic light interaction control method based on scene recognition according to claim 1, characterized in that: In step S1, the raw perception data includes pedestrian / non-motorized vehicle identification data from camera images, target distance and speed coordinates from millimeter-wave radar, oncoming vehicle light signals, road condition warning commands issued by roadside V2X, ambient light intensity, and the vehicle's speed and road type.

3. The intelligent vehicle headlight dynamic light interaction control method based on scene recognition according to claim 1, characterized in that: In step S2, the multi-level dynamic interactive template includes pedestrian yielding gradient light carpet, oncoming vehicle anti-glare stream light, intersection speed limit dynamic digital projection, and high-speed collaborative interactive light strip; the scene classification judgment standard is divided into three levels, namely low-risk high-speed traffic scene, medium-risk urban intersection oncoming vehicle scene, and high-risk zebra crossing pedestrian crossing scene, each level of scene corresponds to a set of independent dynamic light and shadow time sequence gradient logic.

4. The intelligent vehicle headlight dynamic light interaction control method based on scene recognition according to claim 1, characterized in that: In step S1, the spatiotemporal alignment and fusion processing method is to unify the timestamps of radar, camera, and V2X data acquisition, complete the multi-sensor spatial coordinate transformation based on the vehicle chassis coordinates, remove duplicate and invalid targets, and output a standardized scene map containing target category, relative distance, motion trend, and road events.

5. The intelligent vehicle headlight dynamic light interaction control method based on scene recognition according to claim 1, characterized in that: In step S3, the dynamic light language generation timing logic is that the interactive light and shadow adopts continuous gradual output, and the transition time of a single light language pattern is controlled within 10-50ms.

6. The intelligent vehicle headlight dynamic light interaction control method based on scene recognition according to claim 1, characterized in that: In step S6, the triggering signals include pedestrian waving gestures, oncoming vehicles flashing their lights alternately with high and low beams, and V2X roadside equipment collaborative interaction commands.

7. The intelligent vehicle headlight dynamic light interaction control method based on scene recognition according to claim 1, characterized in that: In step S6, the two-way interactive feedback logic is as follows: when a pedestrian's waving action is detected, a flowing soft light carpet is projected from far to near; when an oncoming vehicle flashes its lights to warn the pedestrian, a two-way synchronous gradient warning light is switched. When receiving a V2X road warning command, a dynamic warning digital pattern is projected onto the road surface.

8. The intelligent vehicle headlight dynamic light interaction control method based on scene recognition according to claim 1, characterized in that: The S3 step includes a lightweight rendering compression sub-process, a pre-built vector light language template library, real-time rendering based on vector units, frame differential transmission of light and shadow data, and dynamic allocation of domain control rendering computing power according to scene risks.

9. The intelligent vehicle headlight dynamic light interaction control method based on scene recognition according to claim 1, characterized in that: In step S5, the overall grayscale brightness of dynamic interactive light and shadow is increased in high-light daytime scenes, while the peak brightness of light and shadow is linearly reduced in tunnel and low-light nighttime scenes.

10. The intelligent vehicle headlight dynamic light interaction control method based on scene recognition according to claim 1, characterized in that: In step S8, the fine-tuning of the optical language model parameters is only performed when the vehicle is parked or in a low-speed safe scenario; the iterative calculation is paused when the vehicle is in a high-speed driving state.