Intelligent dimming method and system

Abstract

The invention discloses an intelligent dimming method and system, and relates to the technical field of dimming technology, and the method comprises the steps: collecting the brightness, color temperature and reflection distribution data of an exhibit region; performing feature extraction and light field reconstruction on the acquired data to generate a three-dimensional illumination field; generating a light parameter optimization scheme through a deep learning model and a reinforcement learning dimming engine; and the light control table is used for adaptively adjusting the angle, the brightness and the color temperature of the lamp. According to the invention, three-dimensional modeling and intelligent analysis are carried out on exhibit area illumination based on multi-mode sensing data, an optimal dimming strategy is automatically generated through deep learning and reinforcement learning, and adaptive adjustment of lamp brightness, color temperature and angle is realized, so that the dimming efficiency is remarkably improved, instability of manual debugging is avoided, and the dimming effect is improved. The illumination uniformity and the color temperature consistency are enhanced, real-time optimization of the illumination effect can be kept when the environment changes, the displayed lighting effect is more standardized and can be reproduced, and high stability and visual comfort are achieved.

Description

Technical Field

[0001] This invention relates to the field of dimming technology, and in particular to an intelligent dimming method and system. Background Technology

[0002] As exhibitions, museum displays, and commercial window displays increasingly demand higher quality lighting, traditional lighting adjustment methods are no longer sufficient to meet the needs of precise and intelligent displays. Currently, exhibit lighting largely relies on manual adjustment of the angle, brightness, and color temperature of the lamps, with professional lighting technicians adjusting each lamp individually.

[0003] However, exhibition areas typically have a large number of lighting fixtures, and the process of manually adjusting each light individually is tedious and makes it difficult to achieve high-quality lighting in a short time. The judgment of lighting adjustment often relies on the subjective experience of the staff, making it difficult to form a standardized lighting effect. Moreover, the adjustment results of different personnel will vary. Furthermore, when faced with complex exhibit materials (such as glass, metal, and fabric) and multi-angle reflection characteristics, manual adjustment is difficult to accurately handle uneven lighting issues such as shadows, reflected highlights, and dark areas. Therefore, we propose an intelligent dimming method and system. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent dimming method and system to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent dimming method, comprising the following steps: Step 1: Collect data on brightness, color temperature, and reflectance distribution of the exhibit area using a magnetic or suction cup illuminance meter. Step 2: Based on the multimodal artificial intelligence model, feature extraction and light field reconstruction are performed on the collected data to generate a three-dimensional illumination field; Step 3: Generate lighting parameter optimization schemes through deep learning models and reinforcement learning dimming engines. The optimization targets include brightness uniformity, color temperature stability, and visual comfort. Step 4: The lighting control console adaptively adjusts the angle, brightness, and color temperature of the lights, forming a closed-loop optimization to ensure the display effect of the exhibits.

[0006] Preferably, step 1 includes: Multiple sensor nodes were deployed around the exhibits to collect data on brightness L(t), color temperature C(t), reflectivity R(t), and background illumination B(t). Preprocess the time-series data collected by the sensor to eliminate noise and interference and extract key illumination features; We use deep convolutional neural networks to extract features from the data, generate a light field model in three-dimensional space, and calculate the uniformity of illumination and the rate of color temperature change.

[0007] Preferably, the uniformity evaluation formula for the three-dimensional light field model is: ; Among them, L i Let i be the illuminance at the i-th sampling point. R is the standard deviation of reflectance. avg U represents the average reflectance and is used to describe the uniformity of illumination. The U value is compared with a set uniformity threshold, and the judgment rule is as follows: If U < 0.85, it is determined that the current illumination uniformity is insufficient, and brightness uniformity adjustment is initiated to increase the uniformity of brightness distribution. If 0.85≤U<0.90, then enter the color temperature equalization adjustment stage to optimize the color temperature distribution; If U ≥ 0.90, then the illumination uniformity is considered to meet the standard.

[0008] Preferably, step 2 includes: The collected 3D lighting data is modeled and reconstructed using deep neural networks to generate high-quality lighting prediction maps. By combining self-supervised learning algorithms, the illumination feature extraction process is optimized, enhancing adaptability and accuracy in complex environments; Multilayer perceptron was used to perform regression analysis on the illumination residuals to obtain the parameter values ​​that need to be adjusted.

[0009] Preferably, step 3 includes: A deep learning-based illumination prediction model was established, which can adaptively optimize light brightness and color temperature and perform real-time environmental prediction. The dimming strategy of the lamps is trained by a reinforcement learning dimming engine, and the optimal dimming strategy is generated by using Q-learning and deep Q network algorithms. By dynamically adjusting the learning rate and exploration strategy based on environmental changes and sensor data, the optimization process can be ensured to be real-time and accurate.

[0010] Preferably, the objective function J of the dimming strategy is: ; Where U represents illumination uniformity, and F v For the brightness fluctuation rate, F c E represents the color temperature shift rate. sum λ1 is the total energy of the light field residual map, λ2 is the weight of illumination uniformity, λ3 is the weight of brightness fluctuation rate, and λ4 is the weight of light field residual. After calculating the objective function J, it is compared with a preset optimization threshold, and the judgment rule is as follows: If J < 0.75, then proceed to the brightness optimization stage to improve brightness uniformity and comfort; If 0.75≤J<0.85, then enter the color temperature adjustment stage to optimize color temperature consistency; If J≥0.85, then the dimming effect is considered to meet the standard.

[0011] Preferably, the maximum tolerance of the color temperature shift rate is adjusted as the number of exhibits increases to avoid local overexposure or underexposure. Based on historical data and real-time feedback, the weights of illumination uniformity λ1, brightness fluctuation rate λ2, color temperature shift rate λ3, and light field residual λ4 are dynamically adjusted. Among them, the weights of illumination uniformity λ1 ∈ [0.2, 0.5], brightness fluctuation rate λ2 ∈ [0.1, 0.3], color temperature shift rate λ3 ∈ [0.1, 0.25], and light field residual λ4 ∈ [0.05, 0.2].

[0012] Preferably, the adjustment process in step 4 includes: When the light field residual E(x,y) > 0.05, the luminaires in this area are adjusted, with the adjustment angle ranging from 1° to 5°. When the brightness fluctuation rate F v When the value is greater than 0.1, execute the brightness update command to increase the light intensity of the exhibits; When the color temperature shift rate F c When the value is greater than 0.15, the color temperature calibration mode is activated to adjust the color temperature to the predetermined range; After dimming is completed, a scene lighting fingerprint is generated and stored to support rapid scene reproduction in the future.

[0013] Preferably, the dimming method includes a dimming stability control and security verification mechanism: The brightness change amplitude ΔL of the lamp is detected in three consecutive dimming iterations. When the condition is met in any iteration... When the brightness exceeds 15%, the dimming process is deemed unstable, and the system enters a brightness buffer mode, where the brightness adjustment step size is fixed at 2%. Apply a safety constraint to the angle adjustment range Δθ of the model output, when When the angle is greater than 8°, angle protection is activated, limiting the angle adjustment to 5°. Before each dimming, a consistency check of illumination parameters is performed, specifically for the target lighting parameter P. target Compared with the previous round's fixed parameter P stable When comparing:

[0014] During this process, a transition dimming step is inserted to keep the parameter change rate below 10% in order to avoid visual discomfort or transient changes in illumination caused by abrupt changes.

[0015] A smart dimming system, applied to any of the smart dimming methods described above, comprising: A multimodal sensor module is used to collect data on brightness, color temperature, reflectivity, and background illumination of the exhibit area. The AI ​​lighting analysis and modeling module is used to generate a three-dimensional lighting model through deep learning and to predict and optimize the light field. A reinforcement learning dimming engine module is used to generate optimal lighting adjustment strategies and control the brightness, color temperature and angle of the lamps. A lighting control and execution module is used to automatically adjust the lighting according to an optimization strategy to ensure the lighting effect of the exhibits; The feedback learning and data accumulation module is used to record the lighting effect and environmental feedback during the dimming process, thereby improving the system's adaptability and optimization capabilities. The scene reproduction and management module is used to store the fingerprint of the lighting scene and supports fast reproduction and real-time adjustment.

[0016] The technical effects and advantages of this invention are as follows: This invention performs three-dimensional modeling and intelligent analysis of the lighting in the exhibit area based on multimodal sensing data, and automatically generates the optimal dimming strategy through deep learning and reinforcement learning. This enables adaptive adjustment of the brightness, color temperature and angle of the lamps, thereby significantly improving dimming efficiency, avoiding the instability of manual adjustment, enhancing the uniformity of lighting and the consistency of color temperature, and maintaining real-time optimization of lighting effects when the environment changes. This makes the display lighting effects more standardized, reproducible and highly stable with visual comfort. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram of the illumination uniformity judgment process of the present invention; Figure 3 This is a schematic diagram of the objective function optimization process for the dimming strategy of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] This invention provides, for example Figures 1-3 The intelligent dimming method shown includes the following steps: Step 1: Collect data on brightness, color temperature, and reflectance distribution of the exhibit area using a magnetic or suction cup illuminance meter. Step 1 includes: Multiple sensor nodes were deployed around the exhibits to collect data on brightness L(t), color temperature C(t), reflectivity R(t), and background illumination B(t). Preprocess the time-series data collected by the sensor to eliminate noise and interference and extract key illumination features; We use deep convolutional neural networks to extract features from the data, generate a light field model in three-dimensional space, and calculate the uniformity of illumination and the rate of color temperature change.

[0020] The uniformity evaluation formula for the three-dimensional light field model is as follows: ; Among them, L i Let i be the illuminance at the i-th sampling point. R is the standard deviation of reflectance. avg U represents the average reflectance, and is used to describe the uniformity of illumination. The U value is compared with a set uniformity threshold. In this implementation, in order to achieve accurate perception of the lighting status of the exhibit area, the system first deploys multiple magnetic or suction cup illuminance meters around the exhibit. Each meter acts as an independent sensing node and is positioned according to the outline of the exhibit and the distribution of the lighting environment, so that it can cover key reflective surfaces, shadow edges and highlight areas. Through these sensor nodes, brightness L(t), color temperature C(t), reflectivity R(t) and background illumination B(t) can be continuously collected to form a multi-dimensional, full-area original lighting dataset. The acquired time-series data undergoes noise filtering and outlier removal through a preprocessing unit. Preprocessing includes median filtering, moving average, and outlier detection based on statistical distribution to eliminate interference caused by environmental disturbances, sensor jitter, or transient light changes, ensuring stable and reliable input data. Since the exhibition lighting environment may experience transient interference such as people passing by or sudden increases in reflectivity, causing extreme transient points in sensor data, a planting filter is used to remove impulse noise. This effectively suppresses transient flicker caused by equipment jitter or short-term occlusion, preserving the true trend of lighting changes. Unlike average filtering, it does not cause edge blurring, eliminating "peak" interference and providing a basis for subsequent algorithms. To further reduce high-frequency jitter in the acquired signals, a moving average filter is introduced, effectively reducing high-frequency noise caused by sensor sampling errors or slight environmental fluctuations. This results in smoother output brightness and color temperature curves, improving the stability of the light field modeling input for deep convolutional neural network models and making model training and inference more reliable. To eliminate outliers that do not conform to normal illumination distribution patterns, outlier detection based on statistical distribution is used to remove abnormal data points. This effectively eliminates outliers caused by sudden strong illumination, personnel occlusion, and equipment jitter, while retaining diverse data within the normal range. This makes the light field modeling more consistent with real-world scenarios and significantly improves the accuracy of subsequent illumination uniformity index U calculations.

[0021] Subsequently, the system performs multi-feature fusion on the preprocessed data based on a deep convolutional neural network. The deep convolutional neural network model uses the spatial coordinates of the sensor nodes as a reference to map brightness, color temperature, and reflectivity features into three-dimensional space, thereby constructing a three-dimensional illumination distribution model. This model can generate the distribution pattern of illumination in three-dimensional space and further calculate the illumination uniformity index U and the color temperature change rate, providing crucial basis for subsequent dimming strategies. The illumination uniformity index U comprehensively reflects the uniformity of light received on the exhibit surface, and its judgment rule is as follows: If U < 0.85, it indicates that there is a significant difference in brightness between different areas. Therefore, it is determined that the current illumination uniformity is insufficient, and brightness uniformity adjustment is initiated to increase the uniformity of brightness distribution. If 0.85≤U<0.90, the uniformity of illumination basically meets the requirements, but there may be a slight deviation in color temperature. In this case, the color temperature equalization adjustment stage will be entered to optimize the color temperature distribution and further improve the display effect. If U≥0.90, the illumination uniformity is considered to meet the standard. Since the illumination uniformity has met the standard, no correction is needed at the illumination level, and the existing settings can be maintained. Through the above steps, this method can ensure the accurate reproduction and quantification of the brightness, color temperature and reflection state of the light-receiving area of ​​the exhibit, thereby providing a reliable data basis for subsequent dimming. This step employs a multi-sensor acquisition and deep neural network modeling approach, where multiple sensor nodes simultaneously perceive brightness, color temperature, and reflectivity. This effectively avoids deviations caused by single-point measurements, thereby obtaining three-dimensional data on the exhibit's lighting status. Through multi-level preprocessing and time-series analysis, the system can automatically eliminate environmental interference fluctuations, making lighting calculations more stable and reliable. By using a deep convolutional neural network model for spatial mapping and feature fusion, a high-precision three-dimensional light field model can be generated, reflecting the true structure of the lighting distribution. This provides a clear quantitative basis for subsequent dimming operations. Through a unified formula for calculating lighting uniformity and multi-level threshold judgment, the system can automatically identify areas of uneven lighting and adopt targeted optimization strategies, achieving intelligent zoned dimming. Based on achieving uniform lighting, the system can further adjust the color temperature to ensure that exhibits present the best visual effect in different environments, improving the accuracy and repeatability of lighting design. Through the above implementation methods and effects, this invention achieves a high degree of automation and intelligence in lighting data acquisition, modeling, and analysis, providing a solid foundation and excellent results for subsequent lighting control and optimization.

[0022] Step 2: Based on the multimodal artificial intelligence model, feature extraction and light field reconstruction are performed on the collected data to generate a three-dimensional illumination field; Step 2 includes: The collected 3D lighting data is modeled and reconstructed using deep neural networks to generate high-quality lighting prediction maps. By combining self-supervised learning algorithms, the illumination feature extraction process is optimized, enhancing adaptability and accuracy in complex environments; Regression analysis of illumination residuals was performed using a multilayer perceptron to obtain the parameter values ​​that need to be adjusted; In this embodiment, to achieve high-precision prediction and reconstruction of the illumination distribution in the exhibit area, the present invention employs a multimodal artificial intelligence model to extract and fuse features from the collected brightness, color temperature, reflectivity, and spatial coordinate data, thereby constructing a three-dimensional illumination field and generating an illumination prediction map, specifically as follows: First, the sensor node data obtained in step 1 is structured and encoded according to sampling time, spatial coordinates, and illumination characteristics, serving as the input feature vector of a deep neural network. The deep neural network includes an input layer, multiple nonlinear hidden layers, and an output layer. Its main functions include: multimodal fusion of brightness L(t), color temperature C(t), reflectivity R(t), background illumination B(t), and spatial location information (x, y, z); extraction of the nonlinear correlation structure between illumination characteristics; generation of a preliminary illumination distribution prediction map through feature mapping; the model uses the ReLU activation function to improve feature expression ability; and the upper network uses batch normalization to improve training stability. Through this process, a preliminary predicted distribution of illumination in three-dimensional space can be generated, providing a foundation for further reconstruction of the three-dimensional illumination field. It can automatically extract illumination relationships from multi-dimensional inputs without relying on manual rules, and has stronger adaptability to different materials and different illumination interferences of exhibits. The generated illumination prediction map has high spatial coherence. To further improve the performance of the lighting model in complex exhibition environments, this implementation introduces a self-supervised learning mechanism. By constructing a prediction task, the model performs feature optimization under unlabeled conditions, specifically including: The model attempts to reconstruct the input illumination distribution data, and through mean square error constraints, enables the model to learn the true statistical distribution characteristics of illumination. By partially occluding the illumination data, the model can predict the brightness and color temperature of the occluded area based on the remaining data, thus simulating real-world situations such as occlusion and human interference. By comparing and learning the interrelationships among the three modes of brightness, color temperature, and reflectivity, the robustness of the model in scenarios with rapid changes in illumination is improved. This allows the model to achieve adaptive optimization without relying on a large amount of labeled data, significantly enhancing its adaptability to scenarios with sudden changes in illumination (such as people passing by or local reflection enhancement), effectively improving the stability and accuracy of illumination prediction maps, and reducing the impact of bias noise. To achieve precise dimming, this implementation method further introduces illumination residual map calculation and regression analysis. The specific processing flow is as follows: The illumination prediction map generated by the deep neural network is compared point by point with the actual observed illumination data in step 1 to obtain the illumination residual map: , Among them, L pred C pred For the predicted brightness and color temperature, L real C real For actual measurement of brightness and color temperature, β is the residual weighting coefficient, used to adjust the importance of brightness and color temperature residuals, and β is between 0.1 and 0.3; The error information from the residual map is input into the multilayer perceptron regression network, and the network outputs the dimming parameters that need to be adjusted, including: the adjusted target brightness value L. target Adjusted target color temperature value C target The multilayer sensor achieves precise fitting of lighting adjustment parameters through multilayer linear mapping and nonlinear activation functions, including lamp angle adjustment amount △θ and lamp power correction amount △P. It can dynamically provide the optimal adjustment direction for areas of lighting deviation, making lighting compensation more intelligent, avoiding manual lamp-by-lamp testing, and making the dimming strategy more refined, thus improving the consistency of the overall display effect. Through the above implementation steps, this invention achieves high-precision three-dimensional lighting reconstruction through deep neural networks in terms of lighting data processing and model construction. It can accurately reflect the spatial distribution relationship of brightness, color temperature and reflectivity around the exhibit. The self-supervised mechanism improves the adaptive performance of the model in complex environments and enhances the robustness of the system to dynamic light changes, occlusion and reflection interference. The lighting residual analysis makes the generation of dimming parameters more accurate, without relying on manual experience, making dimming more scientific and intelligent, providing reliable input for subsequent dimming strategies, and enabling the third step of lighting prediction and reinforcement learning dimming engine to have high-quality input and more stable results.

[0023] Step 3: Generate lighting parameter optimization schemes through deep learning models and reinforcement learning dimming engines. The optimization targets include brightness uniformity, color temperature stability, and visual comfort. Step 3 includes: A deep learning-based illumination prediction model was established, which can adaptively optimize light brightness and color temperature and perform real-time environmental prediction. The dimming strategy of the lamps is trained by a reinforcement learning dimming engine, and the optimal dimming strategy is generated by using Q-learning and deep Q network algorithms. By dynamically adjusting the learning rate and exploration strategy based on environmental changes and sensor data, the optimization process can be ensured to be real-time and accurate.

[0024] The objective function J for optimizing the dimming strategy is: ; Where U represents illumination uniformity, and F v For the brightness fluctuation rate, F c E represents the color temperature shift rate. sum λ1 is the total energy of the light field residual map, λ2 is the weight of illumination uniformity, λ3 is the weight of brightness fluctuation rate, and λ4 is the weight of light field residual. After calculating the objective function J, it is compared with a preset optimization threshold, and the judgment rule is as follows: If J < 0.75, the light quality is poor, so the brightness optimization stage is entered to improve brightness uniformity and comfort. If 0.75≤J<0.85, the brightness is acceptable but the color temperature is uneven, then the color temperature adjustment stage is entered to optimize the color temperature consistency. If J≥0.85, then the dimming effect is considered to meet the standard.

[0025] Among them, the maximum tolerance of color temperature shift rate is adjusted when the number of exhibits increases in order to avoid local overexposure or underexposure. Based on historical data and real-time feedback, the weights of illumination uniformity λ1, brightness fluctuation rate λ2, color temperature shift rate λ3, and light field residual λ4 are dynamically adjusted. Among them, the weights of illumination uniformity λ1 ∈ [0.2, 0.5], brightness fluctuation rate λ2 ∈ [0.1, 0.3], color temperature shift rate λ3 ∈ [0.1, 0.25], and light field residual λ4 ∈ [0.05, 0.2]. According to the number of exhibits, illumination complexity, and historical dimming effects, the weights can be dynamically adjusted. When there are many exhibits and complex reflections, the illumination uniformity weight λ1 is appropriately increased. When the ambient light changes significantly, the brightness fluctuation rate weight λ2 is increased to maintain brightness stability. In multi-exhibit scenarios, the color temperature shift rate weight λ3 is automatically increased to avoid local color temperature inconsistencies. When the light field difference is significant, the light field residual weight λ4 is increased to achieve more accurate compensation. In this embodiment, to achieve intelligent adjustment of lamp brightness, angle, and color temperature, the present invention establishes a dimming decision system through a deep learning model and a reinforcement learning engine, enabling the system to automatically generate optimal dimming parameters under different lighting conditions. The specific technical process is as follows: First, a deep learning model is used to predict the three-dimensional lighting field. The model takes the lighting feature map, light field residual information and spatial coordinates obtained in step 2 as input, and extracts the spatiotemporal features of the lighting through a multi-layer convolutional structure and attention mechanism to achieve adaptive prediction of brightness and color temperature distribution. In order to cope with the dynamic nature of the scene such as exhibit movement and background light changes, this model adopts an online update strategy, which automatically optimizes the internal feature expression based on real-time sensor data during operation, making the lighting prediction more in line with the current environment. It can predict in advance the areas of insufficient local lighting, sudden increase in brightness or color temperature drift, so that the dimming strategy no longer depends on human experience, the dimming response speed is faster, and the training stability of the subsequent reinforcement learning model is improved. To achieve automatic dimming, this invention constructs a dimming decision model based on Q-learning and a deep Q-network, specifically implemented as follows: The system uses the following variables as input for dimming status: current illumination uniformity U, brightness fluctuation rate F. v Color temperature shift rate F c Light field residual energy E sum The current brightness, color temperature, angle of the luminaire, and the ambient light intensity and rate of change; The model can select the following dimming behaviors: increase or decrease brightness △L, increase or decrease color temperature △C, adjust lamp angle △θ, and adjust supplementary lighting method; A reward function is constructed based on the lighting effect and dimming stability, including: reward for improved lighting uniformity, reward for color temperature approaching the target range, reward for reduced brightness fluctuation, penalty for excessive dimming, and penalty for angle jump. The dimming strategy is learned automatically by the model, without the need for manual annotation or predefined rules. The strategy becomes more and more accurate with each round of system operation, achieving self-optimization and having a strong generalization ability, which can adapt to different exhibition layouts and lighting types. To improve the response speed of the dimming engine in different environments, this invention introduces a dynamic learning rate α(t) and exploration rate ε(t). When the illumination changes drastically (such as frequent pedestrian traffic causing changes in reflection), the learning rate is increased to enable the model to adapt quickly. When the illumination is stable, the learning rate is reduced to avoid excessive oscillation of the strategy. The exploration rate gradually decays over time, allowing the model to gradually transition from "exploration action" to "utilizing the optimal action". When abnormal illumination residuals are detected (E(x,y) exceeds 0.05), the exploration rate is briefly increased to allow the system to re-explore the strategy in the abnormal area. By using a combination of deep learning and reinforcement learning, this step automatically generates the "optimal dimming strategy," eliminating reliance on manual dimming and providing a complete closed loop of "learning-prediction-adjustment-verification." Dimming decisions do not depend on a single indicator but rather on a comprehensive assessment of four types of lighting quality indicators, with weights that can be automatically adjusted. This gives the system a high degree of intelligence and adaptability, significantly improving the uniformity of lighting, color temperature consistency, and viewing comfort of exhibits.

[0026] Step 4: The lighting control console adaptively adjusts the angle, brightness, and color temperature of the lights, forming a closed-loop optimization to ensure the display effect of the exhibits.

[0027] The adjustment process in step 4 includes: When the light field residual E(x,y) > 0.05, the luminaires in this area are adjusted, with the adjustment angle ranging from 1° to 5°. When the brightness fluctuation rate F v When the value is greater than 0.1, execute the brightness update command to increase the light intensity of the exhibits; When the color temperature shift rate F c When the value is greater than 0.15, the color temperature calibration mode is activated to adjust the color temperature to the predetermined range; After dimming is completed, a scene lighting fingerprint is generated and stored to support rapid scene reproduction in the future; In this embodiment, to achieve stability, uniformity, and visual consistency of the exhibit lighting, the present invention relies on a lighting control console to adaptively adjust the brightness, angle, and color temperature of the lamps. Based on the dimming strategy instructions generated in step 3 and combined with real-time light field analysis results, the lighting control console performs lighting optimization through a zoned adjustment and closed-loop verification mechanism. The specific technical process is as follows: The system first analyzes the light field residual map E(x,y) obtained in step 3 to identify areas with significant illumination deviations. When a certain area satisfies E(x,y) > 0.05, it is considered that the illumination distribution in that area deviates from the ideal value. The lighting control console automatically positions the corresponding lighting fixtures and performs fine adjustment according to the angle adjustment parameters generated by the prediction model. To avoid illumination instability caused by angle jumps, this invention constrains the angle adjustment range to 1°-5°. The adjustment methods include: rotating the lighting fixtures to reposition the center of the light spot, adjusting the projection direction to increase or decrease the local illuminance, and achieving illumination compensation through multi-light fixture collaborative adjustment. This can effectively eliminate local shadows, light spot shifts, and sudden changes in brightness on the surface of exhibits. Multi-light fixture collaborative supplementary lighting makes the light field smoother. The system calculates the brightness fluctuation rate F in real time. v When F is satisfied v A value greater than 0.1 indicates that the exhibition area is experiencing ambient light interference or unstable luminous output, causing brightness fluctuations exceeding the acceptable range. The system sends a brightness update command to the lighting control console, automatically increasing the luminous flux output of the luminous fixtures to gradually bring the actual brightness closer to the target brightness L. target The brightness adjustment process adopts a graded brightness gain mode (2%-5% per step), amplitude limiting protection (to prevent sudden brightness increase), and multi-lamp joint adjustment to avoid local overexposure. It can effectively suppress brightness flicker caused by environmental interference, improve the brightness consistency of exhibits, make the visual experience more comfortable, ensure stable illuminance output, and improve the overall controllability of lighting. The system detects the color temperature offset rate F in real time. c When F is satisfied c When the value is greater than 0.15, it indicates a significant color temperature drift, such as being too cool, too warm, or having uneven light color. The system then activates the color temperature calibration mode, automatically adjusting the color temperature in the following ways: adjusting the RGB channel ratio of the LED light source; if the luminaire has an adjustable color temperature function, then adjusting the target color temperature C output by the prediction model. target Calibration is performed to unify the color temperature of multiple lamps, avoiding color deviation in individual lamps. The calibration process adopts a small step calibration strategy, with a typical adjustment step of 50-100K, to avoid excessive color temperature changes that could lead to abrupt changes in light color. This results in more realistic color reproduction on the surface of exhibits, avoids inconsistent warm and cool colors in exhibit areas, improves overall light color consistency, and provides viewers with a more balanced visual experience. After the brightness, angle, and color temperature are adjusted on the lighting control panel, the system restarts the sensor sampling process to perform closed-loop verification of the current lighting conditions. The verification result will be valid if the following conditions are met: lighting uniformity U ≥ 0.85, color temperature offset F... c ≤0.15 and optical field residual energy E sum When the brightness drops significantly, the dimming is considered to have reached a stable state. Subsequently, a scene lighting fingerprint is generated, which consists of the following information: the final brightness, color temperature and angle parameters of each lamp, key indicators such as lighting uniformity and residual distribution, a snapshot of the 3D light field model of the exhibition area, and a unique scene identifier ID and timestamp. This lighting fingerprint is stored in the system database and is used for rapid scene reproduction (such as when repeating the setup, relighting, or changing scenes), comparative analysis of lighting effects, and optimization training of future lighting strategies. This ensures that the dimming closed loop can verify the effectiveness of each adjustment. The lighting fingerprint transforms exhibition lighting from "experience-based" to "data-based," significantly shortening the dimming time for re-setting up the exhibition and improving efficiency. Through the above technical methods, this step can achieve precise regional supplemental lighting and angle correction, solve problems such as local dark areas, light spot deviation, and disordered reflection, automatically eliminate lighting deviations caused by brightness flicker and unstable lamps, maintain stable illuminance, improve visual comfort, unify color temperature, ensure consistent color reproduction in the exhibition area, avoid color confusion caused by different lamps, provide closed-loop dimming and reproducible light fingerprints, and provide efficient and standardized lighting solutions for large-scale exhibitions.

[0028] The dimming method includes dimming stability control and security verification mechanisms: The brightness change amplitude ΔL of the lamp is detected in three consecutive dimming iterations. When the condition is met in any iteration... When the brightness exceeds 15%, the dimming process is deemed unstable, and the system enters a brightness buffer mode, where the brightness adjustment step size is fixed at 2%. Apply a safety constraint to the angle adjustment range Δθ of the model output, when When the angle is greater than 8°, angle protection is activated, limiting the angle adjustment to 5°. Before each dimming, a consistency check of illumination parameters is performed, specifically for the target lighting parameter P. target Compared with the previous round's fixed parameter P stable When comparing:

[0029] During this process, a transition dimming step is inserted to keep the parameter change rate below 10% in order to avoid visual discomfort or transient changes in illumination caused by abrupt changes. To ensure that the lighting fixtures do not over-adjust, abruptly adjust, or remain unstable during automatic dimming, this invention establishes a complete dimming stability control and safety verification mechanism. This mechanism constrains the variation range of brightness, angle, and dimming parameters, making the dimming process smoother, safer, and more predictable. The specific implementation steps are as follows: To prevent excessive changes in brightness adjustment within a short period during continuous dimming, this implementation method monitors the brightness output for three consecutive dimming iterations. The brightness change amplitude is defined as follows: When any iteration satisfies When the brightness exceeds 15%, it is determined that there is an unstable risk in this round of dimming, such as sluggish output response of the lamps, large error in the lighting model, or sudden change in ambient light. At this time, the system automatically enters the brightness buffer mode. This mode has the following characteristics: fixing the brightness adjustment step size to 2%, prohibiting large brightness jumps, and gradually approaching the target value without excessive correction. The buffer mode can only be exited after at least two consecutive dimming cycles, thereby effectively preventing sudden brightness increases and decreases from causing light flicker, improving dimming stability and lamp response consistency, avoiding abrupt changes in brightness and darkness for the audience in the exhibition hall, and improving visual comfort. When adjusting the projection direction of the lighting fixture, this invention establishes a safety verification mechanism for angle changes. The system reads the angle adjustment value Δθ generated by the deep learning and reinforcement learning models and determines whether there is excessive angle fluctuation. When the angle is greater than 8°, the system determines that the angle adjustment may be too large, which may cause the light spot to deviate from the main body of the exhibit or cause local overexposure. At this time, the angle protection mechanism is triggered, which automatically limits the angle adjustment to △θ=5° to ensure that the adjustment action is within a safe range. This avoids the instantaneous displacement of the light spot due to the angle jump, prevents the sudden appearance of strong shadows or overexposed areas on the surface of the exhibit, improves the stability and controllability of the lighting adjustment, and makes the system dimming softer and more natural. To prevent dimming parameter P target When significant changes occur between two consecutive dimming cycles, this invention performs a consistency check of illumination parameters before each dimming cycle begins, calculating the ratio of change between the current target parameters and the previous stable parameters. When the above changes satisfy If the system determines that the strategy change generated by the dimming system is too large and may cause a sudden change in illumination, it will automatically insert a transition dimming step: limiting the single parameter change to below 10%, and gradually approaching P in 2-4 rounds. targetBy progressively adjusting brightness, color temperature, and angle, the visual impact caused by abrupt changes in dimming parameters is reduced. This avoids color reproduction errors or visual discomfort caused by sudden changes in color temperature and prevents instantaneous redistribution of light caused by sudden changes in angle. The dimming process is smoother and more controllable, improving the user experience. Through the above technical measures, this invention improves the stability, safety, and visual comfort of the dimming process, prevents "hard jumps" in brightness, angle, and color temperature, and makes dimming soft and natural. It fundamentally reduces the interference of sudden changes in light on the visual presentation of exhibits. The robustness of the dimming system is enhanced through continuous multi-round parameter verification, forming a closed-loop control system of "safety-verification-transition-stability".

[0030] An intelligent dimming system, applied to any of the intelligent dimming methods described above, includes a multimodal sensor module, an AI lighting analysis and modeling module, a reinforcement learning dimming engine module, a lighting control and execution module, a feedback learning and data accumulation module, and a scene reproduction and management module. The multimodal sensor module is used to collect brightness, color temperature, reflectivity, and background lighting data of the exhibit area. The AI ​​lighting analysis and modeling module is used to generate a three-dimensional lighting model through deep learning and perform light field prediction and optimization. The reinforcement learning dimming engine module is used to generate the optimal lighting adjustment strategy and control the brightness, color temperature, and angle of the lamps. The lighting control and execution module is used to automatically adjust the lighting according to the optimization strategy to ensure the lighting effect of the exhibits. The feedback learning and data accumulation module is used to record the lighting effect and environmental feedback during the dimming process to improve the system's adaptability and optimization capabilities. The scene reproduction and management module is used to store the lighting scene fingerprint and support rapid reproduction and real-time adjustment. This embodiment provides an intelligent dimming system for optimizing the lighting effect in the exhibition area. The system uses multiple modules to work together, collects data using multimodal sensors, and combines deep learning and reinforcement learning algorithms to generate the optimal dimming strategy, thereby achieving automatic dimming and real-time optimization of the lighting effect. The multimodal sensor module collects real-time data on brightness, color temperature, reflectivity, and background illumination of the exhibit area through multiple sensor nodes installed in the exhibit area. Each sensor node includes: Illuminance sensor: Used to collect brightness data of the exhibit area and provide real-time feedback; Color temperature sensor: Collects color temperature data for the exhibit area; Reflectivity sensor: Used to measure the reflectivity of exhibit surfaces; Background light sensor: used to capture the background light intensity of the surrounding environment; The sensors transmit the collected data to the system center via wireless communication, forming a complete multidimensional dataset. This ensures that the collected illumination data covers multiple parameters such as brightness, color temperature, and reflectivity, providing comprehensive data support for subsequent dimming strategies. The collaborative work of multimodal sensors can quickly reflect changes in illumination in the exhibit environment, ensuring the real-time response of the dimming system. High-precision sensors can capture subtle changes in illumination, providing a reliable basis for precise dimming. The AI ​​lighting analysis and modeling module analyzes and processes the collected lighting data using deep learning algorithms. This module employs convolutional neural networks and self-supervised learning techniques. Data input and feature extraction: Spatial and temporal features in sensor data are extracted using convolutional neural networks, and a multi-dimensional illumination dataset is generated; 3D lighting model generation: Based on the collected data, a 3D lighting model of the exhibit area is established to simulate the distribution and changing trend of lighting; Light field optimization and prediction: Predict future lighting conditions using a deep learning model and optimize the current lighting configuration to make the lighting distribution more uniform; Deep learning algorithms can automatically generate high-precision 3D lighting models based on real-time data, reflecting the real lighting distribution in the exhibition area. Through the light field prediction function, the system can predict the trend of lighting changes in advance, providing data support for the formulation of dimming strategies. This module can automatically optimize the lighting field, reduce manual intervention, and improve the system's dimming efficiency. The reinforcement learning dimming engine module uses Q-learning and deep Q-network algorithms to optimize and train the dimming strategy. The specific process is as follows: State space definition: The dimming engine defines multiple lighting parameters (such as brightness, color temperature, and lighting uniformity) as the state space; Motion space definition: The system adjusts the lighting effect by adjusting motion parameters such as brightness, color temperature, and lamp angle; Reward function design: Design a reward function based on the target lighting effect (such as brightness uniformity, color temperature stability, etc.) to update the strategy; By continuously interacting with the environment, the system continuously optimizes the dimming strategy based on feedback, and gradually obtains the optimal dimming solution. Through continuous training and adjustment, the system can automatically optimize the dimming strategy without human intervention. The reinforcement learning model can quickly make the optimal dimming decision in complex environmental changes. The system can adjust the dimming strategy in real time based on sensor data to ensure the real-time performance and stability of the lighting effect. The lighting control and execution module is responsible for automatically adjusting the brightness, color temperature, and angle of the luminaires according to the optimized dimming strategy. This module precisely adjusts the luminaire parameters through the actuators of the control system. The specific steps are as follows: Brightness adjustment: Adjust the brightness of the lights according to the optimization strategy to ensure the uniformity of brightness and visual comfort of the exhibits; Color temperature adjustment: Adjust the color temperature of the lamps according to the color temperature optimization scheme to keep it stable within the predetermined range; Angle adjustment: Based on the analysis of illumination uniformity, adjust the projection angle of the lamps to ensure that the illumination covers the exhibit area; The lighting control and execution module also monitors and adjusts various parameters of the lamps in real time through dimming algorithms to ensure that the dimming effect meets the predetermined target. The system can accurately control the brightness, color temperature and angle of the lamps to ensure the stability and balance of the lighting effect of the exhibits. The lighting control module can automatically execute optimization strategies without manual intervention, improving work efficiency. Through meticulous adjustment of brightness, color temperature and angle, it ensures that the display effect of the exhibits is more vivid and realistic. The feedback learning and data accumulation module records the lighting effects and environmental feedback during the dimming process, and combines historical data for self-learning and optimization. The specific process is as follows: Data accumulation: The system records the status and results of each dimming in real time, including lighting parameters and feedback data during the dimming process; Adaptive optimization: The system automatically adjusts the dimming strategy and weight parameters based on historical data and feedback information to improve the system's adaptability; Knowledge graph construction: Through accumulated data, the system constructs a knowledge graph of illumination regulation to support subsequent intelligent reasoning and decision-making; As data accumulates, the system can self-adjust and optimize, gradually improving the dimming effect. Based on data accumulation, the system can analyze the advantages and disadvantages of dimming modes and perform real-time optimization. Through accumulated dimming data and environmental feedback, the system can achieve personalized optimization for different exhibition scenarios. The scene reproduction and management module stores and manages lighting scene fingerprints to ensure rapid reproduction of each exhibit's lighting adjustment. This module includes: Light fingerprint generation: After each dimming is completed, the system generates a light fingerprint for the current dimming scene, which includes key information such as lamp parameters, light uniformity, and color temperature stability; Scene storage and management: The system stores the light fingerprint data in the database for subsequent rapid dimming reproduction; Scene Reproduction: During the next dimming process, the system can quickly retrieve historical lighting fingerprints and reproduce the corresponding dimming parameters; By storing light fingerprints, the system can quickly reproduce historical lighting scenes, saving time on setup and dimming. Each time the scene is reproduced, it can be accurately matched to ensure a high degree of consistency in the display effect of exhibits. Through scene management, the system can easily adjust and optimize the dimming effect of different exhibits.

[0031] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A smart dimming method, characterized in that, Includes the following steps: Step 1: Collect data on brightness, color temperature, and reflectance distribution of the exhibit area using a magnetic or suction cup illuminance meter. Step 2: Based on the multimodal artificial intelligence model, feature extraction and light field reconstruction are performed on the collected data to generate a three-dimensional illumination field; Step 3: Generate lighting parameter optimization schemes through deep learning models and reinforcement learning dimming engines. The optimization targets include brightness uniformity, color temperature stability, and visual comfort. Step 4: The lighting control console adaptively adjusts the angle, brightness, and color temperature of the lights, forming a closed-loop optimization to ensure the display effect of the exhibits.

2. The intelligent dimming method according to claim 1, characterized in that, Step 1 includes: Multiple sensor nodes were deployed around the exhibits to collect data on brightness L(t), color temperature C(t), reflectivity R(t), and background illumination B(t). Preprocess the time-series data collected by the sensor to eliminate noise and interference and extract key illumination features; We use deep convolutional neural networks to extract features from the data, generate a light field model in three-dimensional space, and calculate the uniformity of illumination and the rate of color temperature change.

3. The intelligent dimming method according to claim 2, characterized in that, The uniformity evaluation formula for the three-dimensional light field model is as follows: ； Among them, L i Let i be the illuminance at the i-th sampling point. R is the standard deviation of reflectance. avg U represents the average reflectance and is used to describe the uniformity of illumination. The U value is compared with a set uniformity threshold, and the judgment rule is as follows: If U < 0.85, it is determined that the current illumination uniformity is insufficient, and brightness uniformity adjustment is initiated to increase the uniformity of brightness distribution. If 0.85≤U<0.90, then enter the color temperature equalization adjustment stage to optimize the color temperature distribution; If U ≥ 0.90, then the illumination uniformity is considered to meet the standard.

4. The intelligent dimming method according to claim 1, characterized in that, Step 2 includes: The collected 3D lighting data is modeled and reconstructed using deep neural networks to generate high-quality lighting prediction maps. By combining self-supervised learning algorithms, the illumination feature extraction process is optimized, enhancing adaptability and accuracy in complex environments; Multilayer perceptron was used to perform regression analysis on the illumination residuals to obtain the parameter values ​​that need to be adjusted.

5. The intelligent dimming method according to claim 1, characterized in that, Step 3 includes: A deep learning-based illumination prediction model was established, which can adaptively optimize light brightness and color temperature and perform real-time environmental prediction. The dimming strategy of the lamps is trained by a reinforcement learning dimming engine, and the optimal dimming strategy is generated by using Q-learning and deep Q network algorithms. By dynamically adjusting the learning rate and exploration strategy based on environmental changes and sensor data, the optimization process can be ensured to be real-time and accurate.

6. The intelligent dimming method according to claim 5, characterized in that, The objective function J of the dimming strategy is: ； Where U represents illumination uniformity, and F v For the brightness fluctuation rate, F c E represents the color temperature shift rate. sum λ1 is the total energy of the light field residual map, λ2 is the weight of illumination uniformity, λ3 is the weight of brightness fluctuation rate, and λ4 is the weight of light field residual. After calculating the objective function J, it is compared with a preset optimization threshold, and the judgment rule is as follows: If J < 0.75, then proceed to the brightness optimization stage to improve brightness uniformity and comfort; If 0.75≤J<0.85, then enter the color temperature adjustment stage to optimize color temperature consistency; If J≥0.85, then the dimming effect is considered to meet the standard.

7. The intelligent dimming method according to claim 6, characterized in that, The color temperature shift rate is adjusted to its maximum tolerance as the number of exhibits increases, in order to avoid local overexposure or underexposure. Based on historical data and real-time feedback, the weights of illumination uniformity λ1, brightness fluctuation rate λ2, color temperature shift rate λ3, and light field residual λ4 are dynamically adjusted. Among them, the weights of illumination uniformity λ1 ∈ [0.2, 0.5], brightness fluctuation rate λ2 ∈ [0.1, 0.3], color temperature shift rate λ3 ∈ [0.1, 0.25], and light field residual λ4 ∈ [0.05, 0.2].

8. The intelligent dimming method according to claim 1, characterized in that, The adjustment process in step 4 includes: When the light field residual E(x,y) > 0.05, the luminaires in this area are adjusted, with the adjustment angle ranging from 1° to 5°. When the brightness fluctuation rate F v When the value is greater than 0.1, execute the brightness update command to increase the light intensity of the exhibits; When the color temperature shift rate F c When the value is greater than 0.15, the color temperature calibration mode is activated to adjust the color temperature to the predetermined range; After dimming is completed, a scene lighting fingerprint is generated and stored to support rapid scene reproduction in the future.

9. The intelligent dimming method according to claim 1, characterized in that, The dimming method includes dimming stability control and security verification mechanisms: The brightness change amplitude ΔL of the lamp is detected in three consecutive dimming iterations. When the condition is met in any iteration... When the brightness exceeds 15%, the dimming process is deemed unstable, and the system enters a brightness buffer mode, where the brightness adjustment step size is fixed at 2%. Apply a safety constraint to the angle adjustment range Δθ of the model output, when When the angle is greater than 8°, angle protection is activated, limiting the angle adjustment to 5°. Before each dimming, a consistency check of illumination parameters is performed, specifically for the target lighting parameter P. target Compared with the previous round's fixed parameter P stable When comparing: During this process, a transition dimming step is inserted to keep the parameter change rate below 10% in order to avoid visual discomfort or transient changes in illumination caused by abrupt changes.

10. An intelligent dimming system, applied to the intelligent dimming method according to any one of claims 1-9, characterized in that, include: A multimodal sensor module is used to collect data on brightness, color temperature, reflectivity, and background illumination of the exhibit area. The AI ​​lighting analysis and modeling module is used to generate a three-dimensional lighting model through deep learning and to predict and optimize the light field. A reinforcement learning dimming engine module is used to generate optimal lighting adjustment strategies and control the brightness, color temperature and angle of the lamps. A lighting control and execution module is used to automatically adjust the lighting according to an optimization strategy to ensure the lighting effect of the exhibits; The feedback learning and data accumulation module is used to record the lighting effect and environmental feedback during the dimming process, thereby improving the system's adaptability and optimization capabilities. The scene reproduction and management module is used to store the fingerprint of the lighting scene and supports fast reproduction and real-time adjustment.

Images