Intelligent dimming method based on environmental data driving

Abstract

The invention belongs to the technical field of light regulation and control, and particularly relates to an intelligent dimming method based on environmental data driving, which comprises the following steps of: acquiring external and internal multi-source environmental data of a vehicle in real time through a vehicle-mounted environmental sensing system, performing data fusion on the acquired multi-source environmental data by using a multi-sensor fusion algorithm, and obtaining a dimming result; constructing an environment state feature vector based on the fused data, and constructing an input feature vector of subsequent intelligent dimming by introducing historical data features; inputting the constructed input features of intelligent dimming into an intelligent dimming decision model, and outputting an intelligent dimming decision according to the intelligent dimming decision model; according to the invention, by adopting the dynamic weight distribution and weight fusion modeling technology based on the environment change rate and the driving state, the technical effect of comprehensively reflecting the external illumination change, the weather complexity and the driver state by the environment state feature vector is realized.

Description

Technical Field

[0001] This invention belongs to the technical field of lighting control, specifically a smart dimming method driven by environmental data. Background Technology

[0002] With the rapid development of science and technology, intelligent control technology is becoming increasingly important, especially intelligent dimming technology driven by environmental data.

[0003] However, existing technologies are mainly limited in achieving adaptive adjustment of in-vehicle lighting and display brightness based on environmental changes due to the lack of deep integration between environmental perception data and dimming control, the single dimension of multi-source environmental data utilization, and dimming strategies based on static thresholds or manual experience rules. This makes it difficult to balance visual safety under complex lighting and changing road conditions with comfort experience under different driver states and personalized needs in practical applications, thus affecting the overall performance of the intelligent dimming system in terms of safety, adaptability, and user experience. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an intelligent dimming method driven by environmental data, which solves the problems in the prior art such as the lack of deep integration between environmental perception data and dimming control, the single dimension of multi-source environmental data utilization, and dimming strategies mostly based on static thresholds or manual experience rules.

[0005] An intelligent dimming method based on environmental data includes the following steps: real-time acquisition and fusion of multi-source environmental data, specifically:

[0006] Through the vehicle-mounted environmental perception system, multi-source environmental data from both the exterior and interior of the vehicle are collected in real time. Multi-sensor fusion algorithms are used to synchronize the collected multi-source environmental data in time and align its features. After time synchronization and feature alignment are completed, based on the environmental complexity and data reliability of the current driving scenario, corresponding fusion weights are assigned to various environmental features. Weighted fusion calculations are performed based on the adjusted weight coefficients, and the calculation results are used to construct an environmental state feature vector. Furthermore, historical data features are introduced to construct the input feature vector for subsequent intelligent dimming.

[0007] Furthermore, adaptive dimming based on the constructed intelligent dimming input features is specifically as follows:

[0008] The input features of the constructed intelligent dimming are input into the intelligent dimming decision model. Based on the intelligent dimming decision output by the intelligent dimming decision model, the brightness, color temperature and display parameters of the in-vehicle lighting are adjusted in a coordinated manner. During the dimming process, dimming effect data is continuously collected to update the environmental state feature vector.

[0009] Preferably, the real-time acquisition of multi-source environmental data from both the exterior and interior of the vehicle is specifically as follows:

[0010] By deploying in-vehicle environmental perception systems at different locations on the vehicle, real-time data on the external environment and the internal environment of the vehicle can be collected.

[0011] External environment data is obtained by using an ambient light sensor to capture real-time light intensity values ​​outside the vehicle, and combined with road image data collected by a camera to identify the current road type, weather conditions, and the distribution of surrounding obstacles.

[0012] Internal environmental data is collected by obtaining the current interior light level through the vehicle's illuminance sensor, obtaining information on the brightness, on / off status, and usage duration of the display screen through the display system interface, and obtaining driver status data, including driving time, steering wheel operation frequency, and changes in gaze, to construct a multi-dimensional raw environmental data set.

[0013] Preferably, the time synchronization is as follows:

[0014] For the collected multi-source environmental data, a unified timestamp is added to different types of environmental data, and a time sliding window is set to aggregate multi-source environmental data within the same time window. Specifically:

[0015] Set a unified time base, add a unified timestamp to different types of environmental data, and set a time sliding window to aggregate environmental data from multiple sources within the same time window.

[0016] Preferably, the feature alignment is as follows:

[0017] The aggregated environmental data undergoes feature alignment to map different types of environmental data to a unified feature representation space. This includes normalization and encoding transformation, specifically:

[0018] Normalization is a process that compares the original feature values ​​with their corresponding historical ranges, transforming the magnitude of feature changes into a relative proportional relationship.

[0019] Encoding transformation is the process of converting categorical environmental features into numerical representations that do not introduce sequential semantics by using predefined mapping rules.

[0020] Preferably, the allocation of corresponding fusion weights to various environmental features is as follows:

[0021] By introducing weighted features into both the external and internal environment feature vectors, and dynamically adjusting them based on the rate of environmental change and driving state, we have:

[0022]

[0023] in, This indicates the magnitude of change in external light intensity per unit time. This indicates the set maximum light reference value. Represents the weather complexity factor. , This represents the weighting adjustment coefficient. This represents the weight of external environmental features.

[0024] Preferably, the intelligent dimming decision model is as follows:

[0025] It is constructed using a multi-level structure, including an input layer, a feature extraction layer, a feature fusion layer, and a dimming decision output layer. Feature information is sequentially passed between each layer via forward connections, specifically:

[0026] The input layer is used to receive intelligent dimming input features and divides them into multiple feature channels according to the feature source;

[0027] The feature extraction layer is a convolutional neural network sub-model built for road image data, used to identify road signs, pedestrians and other obstacles;

[0028] The feature fusion layer jointly models environmental state features, visual semantic features, device state features, and human-computer interaction features, and establishes the correlation between various features through weighted fusion.

[0029] The dimming decision output layer determines intelligent dimming decisions based on the fused feature results using an adaptive dimming algorithm.

[0030] Preferably, the input layer is as follows:

[0031] Based on the source of features, multiple feature channels are divided, including environmental feature channels, visual semantic feature channels, device status feature channels, and human-computer interaction feature channels. These include:

[0032]

[0033] in, Indicates environmental characteristic channels, Represents visual semantic feature channels. This indicates the device status characteristic channel. Indicates the human-computer interaction feature channel. This indicates a new intelligent dimming input feature.

[0034] Preferably, the feature extraction layer is as follows:

[0035] By performing multi-layer convolution and nonlinear transformation on the input image, high-level semantic features related to road safety and visual load are extracted, and a visual semantic feature vector representing the complexity and risk level of the current driving scenario is output. The output visual semantic feature vector is used as an independent feature channel input to the feature fusion layer.

[0036] Preferably, the dimming decision output layer is as follows:

[0037] Using an adaptive dimming algorithm to determine intelligent dimming decisions, including target adjustment values ​​for interior lighting brightness, color temperature, and display parameters, we have:

[0038]

[0039] in, This indicates the result of the target dimming decision. Indicates the target brightness. Indicates the target color temperature. Indicates the target display parameters. This represents the adaptive dimming mapping function.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. This invention achieves a unified and accurate representation of the complex lighting environment outside the vehicle and the usage status inside the vehicle by adopting real-time acquisition, time synchronization and feature alignment fusion technology of multi-source environmental data in the vehicle. This effectively solves the technical problem in the prior art that the multi-source environmental data is scattered, the dimensions are inconsistent and the timing is not synchronized, resulting in fragmented environmental perception results that are difficult to use directly for dimming control.

[0042] 2. This invention achieves a comprehensive reflection of external illumination changes, weather complexity, and driver status by adopting dynamic weight allocation and weighted fusion modeling technology based on environmental change rate and driving status. This solves the problems of single dimension of multi-source environmental data utilization and difficulty in quantifying the impact of different environmental factors on dimming decisions in the existing technology, and improves the data credibility and scene adaptability of dimming decisions.

[0043] 3. By adopting the technology of introducing historical environmental features and constructing intelligent dimming input feature vectors, this invention achieves the technical effect of sensing and utilizing environmental change trends, thereby avoiding the problem of dimming response lag or frequent fluctuations caused by existing technologies that only make dimming judgments based on instantaneous environmental data, and improving the stability and foresight of dimming behavior.

[0044] 4. This invention adopts a multi-level intelligent dimming decision model and introduces a convolutional neural network in the feature extraction layer to perform visual semantic recognition of road signs, pedestrians and obstacles. This achieves the technical effect of adaptive response of dimming decision to road risk level and driver visual load, thereby overcoming the technical bottleneck of traditional dimming methods based on static thresholds or empirical rules that are difficult to balance driving safety and visual comfort.

[0045] 5. This invention achieves a synergistic balance between safety priority, comfort constraints, and personalized preferences in dimming strategies by jointly modeling and weighting environmental state features, visual semantic features, equipment operation features, and human-computer interaction features. This solves the problem that dimming strategies in the prior art cannot simultaneously meet the needs of safety and personalization.

[0046] 6. By adopting a technology that coordinates the adjustment of the in-vehicle lighting system and the display system, the present invention achieves a coordinated and consistent brightness relationship between the overall in-vehicle light environment and the display of key information, thereby avoiding the problems of local glare or decreased readability caused by adjusting only a single lighting or display device in the prior art. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the overall method steps of an intelligent dimming method based on environmental data driven by the present invention. Detailed Implementation

[0048] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.

[0049] Example 1

[0050] Reference Figure 1 As an embodiment of the present invention, an intelligent dimming method based on environmental data is provided, comprising the following steps:

[0051] S1: Real-time acquisition and fusion of multi-source environmental data.

[0052] Specifically, the real-time acquisition and fusion of multi-source environmental data is achieved through an onboard environmental perception system that collects multi-source environmental data from both the exterior and interior of the vehicle in real time. External environmental data includes ambient light intensity, weather conditions, road type, and the distribution of surrounding obstacles. Internal environmental data includes in-vehicle illuminance levels, display screen usage status, driving duration, and driver behavior and state characteristics. A multi-sensor fusion algorithm is used to synchronize and align the collected multi-source environmental data, constructing an environmental state feature vector to reflect the comprehensive light environment characteristics of the current driving scenario. This allows for accurate characterization of the complex and dynamic driving environment, providing a precise data foundation for subsequent dimming decisions. The specific implementation is as follows:

[0053] Real-time acquisition of multi-source environmental data from both the exterior and interior of the vehicle is achieved through onboard environmental perception systems deployed at different locations within the vehicle. Specifically, this involves:

[0054] During vehicle operation, onboard environmental perception systems located at different positions within the vehicle continuously collect driving-related environmental data. These systems include ambient light sensors, cameras, millimeter-wave radar or lidar, and in-vehicle illuminance sensors.

[0055] Set at The collection of multi-source environmental data acquired in real time is as follows:

[0056]

[0057] in, express A collection of multi-source environmental data collected in real time. This represents the collected data on the vehicle's external environment. This represents the collected data on the vehicle's interior environment.

[0058] External environment data is obtained through ambient light sensors, capturing real-time light intensity values ​​outside the vehicle to reflect trends in natural light. Combined with road image data from cameras, it identifies the current road type, weather conditions, and the distribution of surrounding obstacles. Therefore:

[0059]

[0060] in, This indicates the ambient light intensity value outside the vehicle. Indicates weather conditions (such as sunny, cloudy, rainy or snowy). Indicates road type characteristics, Indicates the distribution characteristics of surrounding obstacles;

[0061] Internal environmental data is collected by using in-vehicle illuminance sensors to obtain the current interior light level, the display system interface to obtain information on screen brightness, on / off status, and usage duration, and driver status data, including driving duration, steering wheel operation frequency, and changes in gaze. This multi-dimensional raw environmental data set is used to simultaneously reflect both the external light environment and the in-vehicle usage status. Therefore:

[0062]

[0063] in, This indicates the interior light intensity as measured by the in-vehicle illuminance sensor. This represents the display system status vector, including display brightness, on / off status, and cumulative usage time. Indicates the driver's driving time. The driving behavior state feature vector includes driving duration, steering wheel operation frequency, and line of sight changes.

[0064] Multi-sensor fusion algorithms are used to synchronize and align the collected multi-source environmental data in time, as detailed below:

[0065] Time synchronization involves adding a unified timestamp to different types of environmental data collected from multiple sources, and setting a time sliding window to aggregate environmental data from multiple sources within the same time window. Specifically:

[0066] Set a unified time base Adding a uniform timestamp to different types of environmental data results in:

[0067]

[0068] in, Indicates the first The initial sampling time of each sensor, This indicates a sliding window for setting the time;

[0069] By setting a time-sliding window to aggregate multi-source environmental data within the same time window, we have:

[0070]

[0071] in, This refers to the aggregated multi-source environmental data within the same time window. This indicates the time-sliding window that is set. Indicates the first The initial sampling time of each sensor.

[0072] Feature alignment is an operation performed on aggregated environmental data to map different types of environmental data to a unified feature representation space. This includes normalization and encoding transformation.

[0073] Normalization is a process that transforms the scale of continuous environmental characteristics, mapping their values ​​to a unified numerical range. This makes the characteristics of different physical quantities comparable at the numerical level, thus avoiding the characteristic imbalance caused by differences in dimensions. Specifically:

[0074] By comparing the original feature values ​​with their historical ranges, the magnitude of feature changes is transformed into a relative proportion, making the contribution of features in the fusion model depend more on the trend of change and relative differences, rather than the absolute numerical value.

[0075] Encoding transformation involves converting categorical environmental features into numerical representations that do not introduce size or order semantics through predefined mapping rules. This allows them to participate as independent dimensions in feature fusion modeling. Specifically:

[0076] For features with a limited number of value categories and no order relationship, a "category-vector" mapping table is established. Each category corresponds to a unique fixed-length vector. In practice, the system maintains an encoding dictionary. When discrete data such as "weather status: rainy day" is collected, the dictionary is queried to obtain the predefined vector encoding. For example, "rainy day" corresponds to the vector [0,0,1,0,0]. The dimension of this vector is equal to the total number of categories. It is 1 only at the corresponding category position and 0 at the rest. This mathematically achieves category distinction without introducing false size relationships.

[0077] Constructing the environmental state feature vector involves assigning corresponding fusion weights to various environmental features based on the environmental complexity and data reliability of the current driving scenario, after completing time synchronization and feature alignment. Specifically, the following steps are taken:

[0078] Based on the variation in external light intensity, weather complexity, and driving stability, the weights of external and internal environmental data features are dynamically adjusted. Specifically:

[0079] When external ambient light changes, increasing the weight of external environmental features in the fusion model; during long-term driving, increasing the weight of internal environment and driving state features, then:

[0080] Introduce weight vectors,

[0081]

[0082] in, Represents the weights of external environmental features. This represents the weights of internal environmental features, and the weighted features satisfy the formula... , This represents the introduced weight vector, whose weights are dynamically adjusted based on the rate of change of the environment and the driving state, specifically:

[0083]

[0084]

[0085] in, This indicates the magnitude of change in external light intensity per unit time. This indicates the maximum reference value for illumination, which is set by the implementer based on the actual application scenario. Represents the weather complexity factor. , This represents the weighting adjustment coefficient, which is set by the implementers based on the actual application scenario. Represents the weights of external environmental features. This represents the weight of internal environmental characteristics.

[0086] It should be noted that the weather complexity factor is not determined by a single factor, but is calculated by combining the following three quantifiable sub-indicators: visibility impact, glare risk, and road surface reflectivity variation. These are weighted and summed according to preset weights, and the result is used as the weather complexity factor. Specifically:

[0087] Visibility impact is calculated by obtaining the current visibility value through vehicle-mounted vision sensors or weather data interfaces, comparing it with the standard clear visibility (such as 1000 meters), and determining the percentage decrease in visibility as the impact.

[0088] Glare risk level is determined by combining weather type (rain, snow, fog) and external light intensity to assess the likelihood of glare caused by weather-related reflection and scattering. For example, in rainy weather with strong light, road surface reflection is severe, and this value increases.

[0089] The road surface reflectance variation is an empirical value directly assigned based on weather type, reflecting the degree of change in light reflection characteristics of the road surface due to wetness, snow accumulation, etc. For example, a low value is assigned to a dry road surface and a high value is assigned to a waterlogged road surface.

[0090] The weather complexity factor is obtained by weighting and summing three sub-indicators according to preset weights (such as visibility 40%, glare 30%, and reflection 30%).

[0091] After dynamic weight adjustment, a weighted fusion calculation is performed based on the adjusted weight coefficients, and the calculation result is constructed as an environmental state feature vector to characterize the light environment characteristics of the current driving scenario. This environmental state feature vector comprehensively reflects external lighting conditions, road environment, and in-vehicle usage and driving state information, resulting in:

[0092]

[0093] in, Represents the feature vectors of the external environment. This represents a feature vector representing the internal environment, including a feature vector representing the driving state. Represents the weights of external environmental features. Indicates the weight of internal environmental characteristics. This represents the constructed environmental state feature vector.

[0094] It should be noted that, for the constructed environmental state feature vector, in order to enhance the ability of subsequent dimming decisions to perceive dynamic environmental changes, historical data features are introduced to construct the input feature vector for subsequent intelligent dimming, resulting in:

[0095] Introducing historical data features, we have:

[0096]

[0097] in, Indicates the length of the historical sampling window. This indicates the trend of changes in environmental state characteristics;

[0098] Based on the determined environmental state characteristics and trends, the input features for intelligent dimming are constructed as follows:

[0099]

[0100] in, This indicates the trend of changes in environmental state characteristics. This represents the constructed environmental state feature vector. The input characteristics of the constructed intelligent dimming are represented and output to the intelligent dimming decision module as a unified format data interface for subsequent coordinated adjustment and control of brightness, color temperature and display parameters.

[0101] S2: Adaptive dimming based on input features of the constructed intelligent dimming system.

[0102] Specifically, adaptive dimming based on the constructed intelligent dimming input features involves inputting these features into an intelligent dimming decision model. Based on the intelligent dimming decision model's output, the brightness, color temperature, and display parameters of the in-vehicle lighting are adjusted collaboratively. Furthermore, dimming effect data is continuously collected during the dimming process to update the environmental state feature vector. The specific implementation is as follows:

[0103] Based on the constructed intelligent dimming input features, in order to provide comprehensive intelligent dimming decision data, image recognition data and human-computer interaction data are collected simultaneously. The newly collected data is then uniformly organized with the constructed intelligent dimming input features to form new input features. The new input features include environmental state feature vectors, which are used to characterize external lighting conditions, weather conditions, and in-vehicle illuminance; environmental semantic features, which are used to characterize the distribution of road signs, pedestrians, and obstacles; equipment operation features; driver preference parameters; and real-time interaction command features. Through structured encapsulation, a unified set of input features for the intelligent dimming decision model is formed.

[0104] The intelligent dimming decision model is constructed using a multi-level structure, including an input layer, a feature extraction layer, a feature fusion layer, and a dimming decision output layer. Feature information is sequentially passed between each layer via a forward connection. Specifically:

[0105] The input layer receives intelligent dimming input features and divides them into multiple feature channels according to their source, including environmental feature channels, visual semantic feature channels, device status feature channels, and human-computer interaction feature channels.

[0106]

[0107] in, Indicates environmental characteristic channels, Represents visual semantic feature channels. This indicates the device status characteristic channel. Indicates the human-computer interaction feature channel. This indicates a new intelligent dimming input feature.

[0108] The feature extraction layer is a convolutional neural network sub-model built for road image data, used to identify road signs, pedestrians, and other obstacles. Specifically:

[0109] By performing multi-layer convolution and nonlinear transformations on the input image, high-level semantic features related to road safety and visual load are extracted, and a visual semantic feature vector representing the complexity and risk level of the current driving scenario is output. The output visual semantic feature vector is then used as an independent feature channel input to the feature fusion layer, resulting in:

[0110] Based on the extraction of high-order features from visual semantic channels using convolutional neural networks, and the unification of scale for the remaining channels through nonlinear mapping, we have:

[0111]

[0112] in, This represents the extracted high-level visual semantic features. , These represent the weighting coefficient and the bias term, which are set by the implementers based on the actual application scenario. This represents a non-linear activation function, which is set by the implementer according to the actual application scenario.

[0113] It should be noted that the convolutional neural network adopts an optimized layer structure design to ensure real-time performance on the vehicle computing unit. The network input receives standardized images from the surround-view camera. The network core consists of multiple "convolution-activation-pooling" modules connected in series. Each module gradually extracts hierarchical features of the image from edges and textures to objects and scenes. The last fully connected layer of the network outputs a fused feature vector.

[0114] Convolutional neural networks are trained through a "multi-task learning" approach, simultaneously learning to recognize road signs, detect pedestrians, estimate obstacle density, and assess scene complexity. This ensures that the feature vectors extracted by the network can comprehensively represent the complexity, risk level, and information load of the current visual environment, providing a direct basis for dimming decisions. In practice, the network can be pre-trained using publicly available large-scale driving datasets and then fine-tuned using data collected from specific vehicles to ensure its adaptability and accuracy. This invention does not limit the specific implementation scenario.

[0115] The feature fusion layer jointly models environmental state features, visual semantic features, device state features, and human-computer interaction features. Through weighted fusion, it establishes the correlation between various features. Ambient lighting change trends and visual semantic risk features are used to jointly determine the safety priority of the dimming strategy. Device state features are used to constrain the dimming amplitude range, and human-computer interaction features are used to personalize the dimming results. Therefore:

[0116]

[0117] in, This represents the combined feature vector after fusion. , , , This represents the dynamic weighting coefficient, which is set by the implementers according to the actual application scenario.

[0118] The dimming decision output layer, based on the fused feature results, uses an adaptive dimming algorithm to determine intelligent dimming decisions, including target adjustment values ​​for in-vehicle lighting brightness, color temperature, and display parameters. Specifically:

[0119] The weights of each dimming parameter are dynamically adjusted based on the magnitude of environmental changes, road risk levels, and driver visual needs. This ensures that the dimming response can adapt to environmental changes promptly while avoiding excessive and frequent adjustments that could interfere with the driver.

[0120]

[0121] in, This indicates the result of the target dimming decision. Indicates the target brightness. Indicates the target color temperature. Indicates the target display parameters. This represents an adaptive dimming mapping function used to balance safety, comfort, and visual load.

[0122] It should be noted that the adaptive dimming mapping function is not a single mathematical formula, but a decision-making process involving multiple steps, including safety rule filtering, initial neural network mapping, personalization and context fine-tuning, and smooth output processing. The specific implementation is as follows:

[0123] Safety rule filtering is based on national or industry standards and ergonomics research results, setting upper and lower safety limits for various dimming parameters (such as brightness and color temperature) to ensure that the output value falls within the range of these hard constraints.

[0124] The initial mapping of the neural network is to use a trained small neural network to map the comprehensive feature vector of the input to the initial dimming parameter values. The trained small neural network captures complex nonlinear relationships by learning a large amount of "ideal environment - comfortable dimming" pairing data.

[0125] Personalization and contextual fine-tuning involves fine-tuning parameters based on the neural network output, combined with real-time human-machine interaction commands (such as driver manual adjustment preferences) and vehicle status information (such as a preference for energy-saving mode when the battery is low).

[0126] Smoothing output processing involves smoothing the final output with a filter to ensure that changes in brightness and color temperature are continuous and gradual, avoiding excessive jumps in dimming commands between adjacent moments that could cause driver discomfort.

[0127] Furthermore, based on the dimming decision output by the intelligent dimming decision model, the in-vehicle lighting system and display system are controlled in a coordinated manner. The overall brightness and color temperature of the in-vehicle interior are adjusted through the lighting control module, and the brightness, contrast and display mode of the display screen are adjusted through the display control module.

[0128] In addition, during dimming, the brightness relationship between the lighting system and the display system is kept in harmony to avoid visual discomfort caused by excessive brightness or darkness in certain areas. Through coordinated dimming, the interior lighting environment ensures overall comfort while prioritizing the clear readability of key driving information.

[0129] It should be noted that after dimming is executed, dimming effect data is continuously collected, including changes in in-vehicle illuminance, display readability, and driver behavior. The environmental state feature vector is updated based on the collected dimming effect data and used as subsequent input to the intelligent dimming decision model. Therefore:

[0130] according to Coordinated control of lighting and display systems;

[0131] If dimming effect data is continuously collected, then:

[0132]

[0133] in, This represents the dimming effect feedback vector. This indicates the interior illuminance after dimming. Indicates the readability evaluation index of the display screen. Parameters representing changes in driver's operational behavior;

[0134] Based on the collected dimming effect data, the environmental state feature vector is updated, then:

[0135]

[0136] in, This represents the updated environmental state feature vector. This indicates the feature update step size coefficient, which can be set by the implementer according to the actual application scenario.

[0137] It should be noted that the display screen readability evaluation index is obtained by comprehensively calculating multiple indicators, including physical contrast ratio, visual load assessment, and content clarity simulation. The constructed multiple indicators are then weighted and combined, specifically as follows:

[0138] The brightness of the display screen is measured by an in-vehicle illuminance sensor and the brightness of its surrounding background environment. The ratio between the two is calculated. The higher the ratio, the better the basic readability and the higher the physical contrast.

[0139] Using small cameras or infrared sensors deployed near the display screen, monitor the driver's eye movement characteristics when looking at the display screen, such as gaze stability and changes in blink frequency. When visual load increases, the readability score decreases accordingly, and the visual load assessment decreases.

[0140] The system has a built-in image processing algorithm that simulates the display effect of a standard test image (including text and icons) on the screen under the current lighting conditions, and calculates its sharpness score. The higher the sharpness score, the better the content sharpness simulation effect.

[0141] The constructed multiple indicators are weighted and calculated to obtain the display readability evaluation index.

[0142] Driver behavior change parameters are determined by continuously monitoring driver behavior data during driving, including steering wheel operation signals, pedal operation signals, and vehicle trajectory signals. These parameters are specifically defined based on the detected behavior data.

[0143] High-frequency data of steering wheel angle and torque are acquired via CAN bus, and their smoothness and correction frequency are analyzed to obtain steering wheel operation signals.

[0144] By monitoring the opening curves of the accelerator and brake pedals, the smoothness and timeliness of the operation are evaluated, and pedal operation signals are obtained.

[0145] By using GPS and lane line recognition, the vehicle's ability to maintain its lateral position relative to the lane centerline is calculated, and vehicle trajectory signals are obtained.

[0146] The acquired behavioral data signals are used to construct driver operation behavior curves. At the same time, an operation behavior baseline is constructed based on the driver's behavior data under ideal conditions (such as sufficient rest and good daytime weather). The cosine similarity between the driver's operation behavior curve and the operation behavior baseline is calculated, and the parameters for driver operation behavior changes are determined based on the calculation results.

[0147] To further verify the beneficial effects of the present invention, an experimental verification scheme for an intelligent dimming system based on the method of the present invention is provided, as follows:

[0148] In typical scenarios such as urban roads, highways and tunnels, an ambient light sensor is used to collect external light intensity data with a sampling frequency of 10Hz and a light intensity range of 50lx to 120000lx; at the same time, an in-vehicle illuminance sensor is used to collect the initial illuminance level inside the vehicle with a sampling accuracy of ±2lx.

[0149] Road image data was acquired using a forward-facing camera (1920×1080 resolution, 30fps), and then analyzed by a convolutional neural network model to identify the distribution of road signs, pedestrians, and obstacles, generating visual semantic recognition results. The accuracy rate for road sign recognition was 96.2%, and the accuracy rate for pedestrian recognition was 94.7%.

[0150] The driver's operation behavior data and preference parameters are collected through the in-vehicle human-machine interaction module, including the display brightness preference range (40%~70%), color temperature preference range (4500K~6500K) and real-time interaction command data, to form a personalized input dataset for the driver;

[0151] The collected external lighting data, in-vehicle illuminance data, visual semantic recognition results, equipment operation status data, and human-computer interaction data are processed for time synchronization, with a unified timestamp accuracy of 10ms, and feature aggregation is performed based on a sliding time window (window width 1s).

[0152] Continuous features are normalized, and discrete visual semantic features are encoded and transformed to construct environmental state feature vectors, environmental semantic feature vectors, equipment operation feature vectors, and human-computer interaction feature vectors. These are then encapsulated in a structured manner to form a unified input feature set for the intelligent dimming decision model, with a feature dimension of 128.

[0153] When the rate of change in external illumination is greater than 30% / s or the visual semantic risk level is higher than the preset threshold, the weight of environmental features and visual semantic features is increased to 0.6; when the driver fatigue characteristics are significant, the weight of human-computer interaction features is increased to 0.4, and corresponding brightness, color temperature and display parameter adjustment values ​​are generated.

[0154] Based on the dimming decision results, the brightness and color temperature of the interior lighting are adjusted through the lighting control module, and the brightness, contrast and display mode of the display screen are adjusted synchronously through the display control module, with the dimming response delay controlled within 200ms.

[0155] During the dimming process, dimming effect data is continuously collected, including changes in in-vehicle illuminance after dimming, display contrast improvement rate, and driver operation behavior changes. The dimming effect data is then fed back to the environmental state feature vector update module for the next round of dimming decision-making.

[0156] Based on the output dimming decision and existing technology, the dimming performance and effect are evaluated. The specific evaluation results are shown in the table below:

[0157] Performance indicators This invention Existing technology Evaluation results Average dimming response delay (ms) 120 250 Response speed improved by approximately 52%. In-vehicle illuminance fluctuation range (lx) ±15 ±50 Stability improved by approximately 70%. Display readability metrics ≥0.9 0.6~0.8 Readability is significantly improved and it is more stable. Increased similarity in driver operating behavior +12% constant The driving operation is closer to the ideal state Visual comfort rating (out of 5) 4.5 3.2 Subjective comfort significantly improved Environmental adaptability score (out of 5) 4.6 3.0 More adaptable to different environments Multi-scene dimming consistency 95% 65% High policy consistency across multiple scenarios

[0158] As can be easily seen from the comparison table above, the technical solution adopted by the present invention has good real-time performance, adaptability and user experience consistency in terms of multi-source environment perception, dynamic weight fusion and intelligent decision output. All key performance indicators are significantly better than the traditional fixed threshold dimming method.

[0159] Furthermore, if the aforementioned function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0160] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0161] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0162] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A smart dimming method based on environmental data, characterized in that: Includes the following steps, Real-time acquisition and fusion of multi-source environmental data, specifically: Through the vehicle-mounted environmental perception system, multi-source environmental data from both the exterior and interior of the vehicle are collected in real time. Multi-sensor fusion algorithms are used to synchronize the collected multi-source environmental data in time and align its features. After time synchronization and feature alignment are completed, based on the environmental complexity and data reliability of the current driving scenario, corresponding fusion weights are assigned to various environmental features. Weighted fusion calculations are performed based on the adjusted weight coefficients, and the calculation results are used to construct an environmental state feature vector. Furthermore, historical data features are introduced to construct the input feature vector for subsequent intelligent dimming. And, based on the constructed intelligent dimming input features, adaptive dimming is specifically as follows: The input features of the constructed intelligent dimming are input into the intelligent dimming decision model. Based on the intelligent dimming decision output by the intelligent dimming decision model, the brightness, color temperature and display parameters of the in-vehicle lighting are adjusted in a coordinated manner. During the dimming process, dimming effect data is continuously collected to update the environmental state feature vector.

2. The intelligent dimming method based on environmental data as described in claim 1, characterized in that: The real-time acquisition of multi-source environmental data from both the exterior and interior of the vehicle is as follows: By deploying in-vehicle environmental perception systems at different locations on the vehicle, real-time data on the external environment and the internal environment of the vehicle can be collected. External environment data is obtained by using an ambient light sensor to capture real-time light intensity values ​​outside the vehicle, and combined with road image data collected by a camera to identify the current road type, weather conditions, and the distribution of surrounding obstacles. Internal environmental data is collected through in-vehicle illuminance sensors to obtain the current interior light level, and through the display system interface to obtain information on display brightness, on / off status, and usage duration, as well as driver status data. This includes driving time, steering wheel operation frequency, and changes in line of sight, constructing a multi-dimensional raw environmental data set.

3. The intelligent dimming method based on environmental data as described in claim 2, characterized in that: The time synchronization is as follows: For the collected multi-source environmental data, a unified timestamp is added to different types of environmental data, and a time sliding window is set to aggregate multi-source environmental data within the same time window. Specifically: Set a unified time base, add a unified timestamp to different types of environmental data, and set a time sliding window to aggregate environmental data from multiple sources within the same time window.

4. The intelligent dimming method based on environmental data as described in claim 3, characterized in that: The feature alignment is as follows: The aggregated environmental data undergoes feature alignment to map different types of environmental data to a unified feature representation space. This includes normalization and encoding transformation, specifically: Normalization is a process that compares the original feature values ​​with their corresponding historical ranges, transforming the magnitude of feature changes into a relative proportional relationship. Encoding transformation is the process of converting categorical environmental features into numerical representations that do not introduce sequential semantics by using predefined mapping rules.

5. The intelligent dimming method based on environmental data as described in claim 4, characterized in that: The specific fusion weights assigned to various environmental features are as follows: By introducing weighted features into both the external and internal environment feature vectors, and dynamically adjusting them based on the rate of environmental change and driving state, we have: in, This indicates the magnitude of change in external light intensity per unit time. This indicates the set maximum light reference value. Represents the weather complexity factor. , This represents the weighting adjustment coefficient. This represents the weight of external environmental features.

6. The intelligent dimming method based on environmental data as described in claim 5, characterized in that: The intelligent dimming decision model is as follows: It is constructed using a multi-level structure, including an input layer, a feature extraction layer, a feature fusion layer, and a dimming decision output layer. Feature information is sequentially passed between each layer via forward connections, specifically: The input layer is used to receive intelligent dimming input features and divides them into multiple feature channels according to the feature source; The feature extraction layer is a convolutional neural network sub-model built for road image data, used to identify road signs, pedestrians and other obstacles; The feature fusion layer jointly models environmental state features, visual semantic features, device state features, and human-computer interaction features, and establishes the correlation between various features through weighted fusion. The dimming decision output layer determines intelligent dimming decisions based on the fused feature results using an adaptive dimming algorithm.

7. The intelligent dimming method based on environmental data as described in claim 6, characterized in that: The input layer is specifically as follows: Based on the source of features, multiple feature channels are divided, including environmental feature channels, visual semantic feature channels, device status feature channels, and human-computer interaction feature channels. These include: in, Indicates environmental characteristic channels, Represents visual semantic feature channels. This indicates the device status characteristic channel. Indicates the human-computer interaction feature channel. This indicates a new intelligent dimming input feature.

8. The intelligent dimming method based on environmental data as described in claim 7, characterized in that: The feature extraction layer is specifically as follows: By performing multi-layer convolution and nonlinear transformation on the input image, high-level semantic features related to road safety and visual load are extracted, and a visual semantic feature vector representing the complexity and risk level of the current driving scenario is output. The output visual semantic feature vector is used as an independent feature channel input to the feature fusion layer.

9. The intelligent dimming method based on environmental data as described in claim 8, characterized in that: The dimming decision output layer is specifically as follows: Using an adaptive dimming algorithm to determine intelligent dimming decisions, including target adjustment values ​​for interior lighting brightness, color temperature, and display parameters, we have: in, This indicates the result of the target dimming decision. Indicates the target brightness. Indicates the target color temperature. Indicates the target display parameters. This represents the adaptive dimming mapping function.

Images