An intelligent lighting control method based on people flow monitoring

Abstract

The present application belongs to the technical field of intelligent lighting, and particularly relates to a kind of intelligent lighting control method based on people flow monitoring, wherein real-time people flow distribution and ambient light original data are obtained, natural light background, shadow disturbance and high-order optical turbulence characteristics are separated through physical guidance decoupling network, holographic people flow thermographic image is generated with the aid of space-time convolution recurrent network, non-uniform people flow illumination demand prediction model is constructed, people flow aggregation and movement trend are predicted, dynamic light field regulation target is generated by fusing visual physiology and rhythm stimulation model, multi-lamp light spot superposition illumination calculation model is constructed, and the solution is obtained through virtual particle swarm optimization algorithm, multi-lamp collaborative regulation instruction is generated after illumination accuracy and glare check, lamp parameters are adjusted, calibration deviation is adjusted through digital twin light decay prediction compensation model, instantaneous physiological equivalent illumination is controlled, and dynamic lighting regulation record atlas is generated. Thus, the problems of low regulation accuracy and poor lighting comfort in the prior art are solved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent lighting technology, specifically relating to an intelligent lighting control method based on pedestrian flow monitoring. Background Technology

[0002] With the rapid growth in smart city construction and building energy conservation demands, smart lighting, as a core component of energy consumption management and living environment improvement in public buildings, is experiencing a continuous increase in market demand. Adaptive pedestrian flow control is a core function of smart lighting. The lighting system dynamically adjusts illuminance levels and light environment parameters based on the distribution of pedestrian traffic in the area. The accuracy of this adjustment directly determines lighting comfort, energy efficiency, and visual health. Existing smart lighting control technologies mostly employ fixed threshold-based tiered control modes, which struggle to respond in real-time to non-uniform pedestrian distribution, dynamic changes in natural light, and multi-source optical disturbances. This results in delayed control response, limited illuminance matching accuracy, and problems such as uneven illuminance, excessive glare, and poor energy-saving effects.

[0003] Specifically, existing technologies suffer from the following key deficiencies: First, insufficient decoupling between pedestrian flow perception and ambient light. They often rely on single light intensity sensors to collect pedestrian flow and ambient light data, failing to effectively separate the background of natural light, shadow disturbances, and high-order optical turbulence interference. This results in low accuracy in pedestrian trajectory and posture recognition, making it difficult to support refined illuminance control. Second, the illuminance demand model is coarse, often setting fixed illuminance values ​​based solely on pedestrian density, without considering pedestrian movement trends, individual visual physiological characteristics, and circadian rhythm health needs. This makes it unsuitable for adapting to differentiated illuminance requirements in non-uniform pedestrian flow scenarios, easily leading to insufficient or excessive local illuminance. Third, insufficient precision in multi-lamp collaborative control. It lacks accurate calculation of light spot superposition illuminance and a multi-parameter global optimization mechanism, and does not consider output deviations caused by long-term light decay of lamps. Furthermore, it lacks closed-loop feedback and dynamic optimization mechanisms, failing to link control parameters with final lighting quality indicators. It cannot adjust control strategies in real-time based on comfort fluctuations during operation, resulting in poor lighting quality stability. As smart lighting develops towards refinement, health, and low energy consumption, the market demand for dynamic and precise control of pedestrian-adaptive lighting is becoming increasingly urgent, but existing technologies are unable to meet these needs due to the aforementioned shortcomings. Summary of the Invention

[0004] Therefore, the purpose of this invention is to provide an intelligent lighting control method based on pedestrian flow monitoring, so as to at least solve the shortcomings of the above-mentioned technology.

[0005] This invention proposes an intelligent lighting control method based on pedestrian flow monitoring, comprising the following steps: acquiring real-time pedestrian flow distribution data and ambient light raw data within the lighting area; adaptively separating natural light background, shadow disturbance, and higher-order optical turbulence features from the ambient light raw data using a physically guided decoupling network, and generating a holographic thermal image of pedestrian flow carrying trajectory and posture using a spatiotemporal convolutional recurrent network; constructing a non-uniform pedestrian flow illuminance demand prediction model based on the real-time pedestrian flow distribution data, natural light background parameters, and holographic pedestrian flow thermal image, predicting pedestrian flow aggregation patterns and movement trends, and integrating individual visual physiological models and rhythmic stimulation curves to construct a personalized isooptic model. A brightness and illuminance demand model is established to form a dynamic light field control target. Based on the dynamic light field control target, and combining the inverse square law of illuminance and the cosine law of incident angle, a multi-lamp spot superposition illuminance calculation model is constructed. Multi-parameter collaborative solution is performed through virtual particle swarm optimization algorithm. After illuminance accuracy and glare value gradient verification, a multi-lamp collaborative dynamic control command is generated. According to the multi-lamp collaborative dynamic control command, the output power, light distribution angle and projection orientation of the lamps in the lighting circuit are adjusted. The output deviation is calibrated online through a digital twin light decay prediction compensation model to control the instantaneous physiological equivalent illuminance during the dynamic change of pedestrian flow, and a dynamic lighting control record map is generated.

[0006] Preferably, a physical-guided decoupling network adaptively separates the natural light background, shadow perturbation, and higher-order optical turbulence features from the raw ambient light data. A spatiotemporal convolutional recurrent network is then used to generate a holographic thermal image of the pedestrian flow carrying trajectory and posture. This process includes: constructing a multi-source ambient light perturbation feature database; based on the database, and combining the spatial light intensity distribution, spectral component proportions, and temporal fluctuation amplitudes in the raw ambient light data, separating the natural light background component and shadow perturbation component through the orthogonal constraint decomposition layer of the physical-guided decoupling network; based on the natural light background component and shadow perturbation component, introducing a higher-order turbulence regularization term to extract higher-order optical turbulence features from the ambient light; and combining this with a spatiotemporal convolutional recurrent network to encode and fit the temporal trajectory of the pedestrian flow and human posture features to generate a holographic thermal image of the pedestrian flow carrying trajectory and posture.

[0007] Preferably, the formula for calculating the illuminance of the multi-light spot superposition is as follows: The formula for calculating the horizontal illuminance of a single point light source is: ; In the formula, Let i be the base horizontal illuminance produced by the i-th lamp in the j-th lighting zone; The initial luminous intensity of the light source; The angle between the incident direction of the light ray and the normal to the horizontal plane; The azimuth angle of the light distribution of the lamp; This is the straight-line distance from the light source of the luminaire to the calculation point of the lighting zone; The pedestrian flow distribution weighting correction term is: ; In the formula, The pedestrian flow distribution weighting coefficient for the j-th lighting zone; This serves as a baseline coefficient for personnel density. Let be the real-time pedestrian density of the j-th partition; This is a correction factor for movement speed; Let be the average movement speed of people in the j-th partition; Combining the above two equations, we derive the formula for calculating the illuminance of multiple light spots superimposed on each other: ; In the formula, Let be the instantaneous equivalent illuminance of the j-th partition; Let be the actual output light intensity of the i-th lamp at time t; Let be the angle of incidence of the light from the i-th lamp at time t; Let be the light distribution azimuth angle of the i-th lamp at time t; Let be the distance from the i-th lamp to the calculation point at time t; The total number of luminaires involved in lighting control.

[0008] Preferably, based on the real-time pedestrian flow distribution data, natural light background parameters, and holographic pedestrian flow thermal imaging, a non-uniform pedestrian flow illuminance demand prediction model is constructed to predict pedestrian flow aggregation patterns and movement trends. This model is then integrated with individual visual physiological models and rhythmic stimulation curves to construct a personalized isopyroluminance illuminance demand model, forming a dynamic light field control target. This includes: constructing the non-uniform pedestrian flow illuminance demand prediction model; based on the non-uniform pedestrian flow illuminance demand prediction model, combined with real-time pedestrian flow distribution data, natural light background parameters, and holographic pedestrian flow thermal imaging, predicting pedestrian flow aggregation patterns and movement trends to obtain basic illuminance demand thresholds for each area; and based on these basic illuminance demand thresholds, solving for personalized isopyroluminance illuminance demands for each area through fitting individual visual physiological models and rhythmic stimulation curves, extracting spatial illuminance distribution gradient features, and fitting to obtain a dynamic light field control target that meets visual comfort and rhythmic health requirements.

[0009] Preferably, the non-uniform pedestrian flow illumination demand prediction model includes: constructing a historical database of lighting areas; based on the historical database of lighting areas, and combining real-time pedestrian flow distribution density, aggregation patterns, and body trajectory features in holographic pedestrian flow thermal imaging, extracting the spatiotemporal evolution features of the dynamic distribution of pedestrian flow through a spatiotemporal graph convolution algorithm; based on the spatiotemporal evolution features of the dynamic distribution of pedestrian flow, introducing a natural light background illuminance compensation term, fitting the regional illuminance demand curves under different pedestrian flow scenarios, and establishing a non-uniform pedestrian flow illumination demand prediction model.

[0010] Preferably, generating a dynamic lighting control record map includes: acquiring actual illuminance, luminaire operating parameters, pedestrian flow distribution data, and visual comfort data in the area during the control process; constructing a lighting control effect evaluation model based on the actual illuminance, luminaire operating parameters, pedestrian flow distribution data, and visual comfort data, and calculating illuminance uniformity, uniform glare value, energy saving rate, and visual satisfaction quality indicators; and associating the lighting control parameters, luminaire operating status, and the quality indicators along a time axis to generate a dynamic lighting control record map that marks control nodes, abnormal events, and comfort fluctuations.

[0011] Preferably, adjusting the output power, light distribution angle, and projection orientation of the luminaires in the lighting circuit, and calibrating the output deviation online through a digital twin light decay prediction and compensation model to control the instantaneous physiological equivalent illuminance during the dynamic change phase of pedestrian flow, includes: constructing a digital twin light decay prediction and compensation model; based on the digital twin light decay prediction and compensation model, adjusting the output power of the luminaires through a dimming drive module and adjusting the light distribution angle and projection orientation through a stepper motor according to the multi-lamp collaborative dynamic control command, controlling the instantaneous physiological equivalent illuminance of each zone; when the illuminance deviation of a zone exceeds a preset threshold, triggering a self-calibration compensation mechanism, correcting the luminaire output parameters based on the light decay prediction results, and simultaneously generating an abnormal warning signal.

[0012] This invention provides an intelligent lighting control system based on pedestrian flow monitoring, comprising: an acquisition module for acquiring real-time pedestrian flow distribution data and ambient light raw data within the lighting area; a separation module for adaptively separating natural light background, shadow disturbances, and higher-order optical turbulence features from the ambient light raw data using a physically guided decoupling network, and generating a holographic thermal image of pedestrian flow carrying trajectories and body postures using a spatiotemporal convolutional recurrent network; and a prediction module for constructing a non-uniform pedestrian flow illuminance demand prediction model based on the real-time pedestrian flow distribution data, natural light background parameters, and the holographic thermal image of pedestrian flow, predicting pedestrian flow aggregation patterns and movement trends, and integrating individual visual physiological models and rhythmic stimulation curves to construct personalized... A dynamic light field control target is formed based on the illuminance demand model. A generation module, based on the dynamic light field control target and combining the inverse square law of illuminance and the cosine law of incident angle, constructs a multi-lamp spot superposition illuminance calculation model. This model is solved collaboratively using a virtual particle swarm optimization algorithm for multiple parameters. After illuminance accuracy and glare value gradient verification, a multi-lamp collaborative dynamic control command is generated. A control module, based on the multi-lamp collaborative dynamic control command, adjusts the output power, light distribution angle, and projection orientation of the lamps in the lighting circuit. It uses a digital twin light decay prediction and compensation model to calibrate output deviations online, controlling the instantaneous physiological equivalent illuminance during the dynamic changes in pedestrian flow, and generating a dynamic lighting control record map.

[0013] The present invention also proposes a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described intelligent lighting control method based on pedestrian flow monitoring.

[0014] The present invention also proposes a readable storage medium storing a computer program that, when executed by a processor, implements the above-described intelligent lighting control method based on pedestrian flow monitoring.

[0015] This invention discloses an intelligent lighting control method based on pedestrian flow monitoring. By collecting real-time pedestrian flow distribution and ambient light data in the lighting area, a physically guided decoupling network is constructed to accurately separate multi-source optical disturbances, generating a high-precision holographic thermal image of pedestrian flow. This solves the problems of weak anti-interference capability and insufficient accuracy of pedestrian flow feature recognition in traditional sensing methods. Based on a non-uniform pedestrian flow illuminance demand prediction model, the method predicts the evolution trend of pedestrian flow and integrates visual physiology and circadian rhythm health models to construct dynamic light field control targets, adapting to the differentiated lighting needs in non-uniform pedestrian flow scenarios. Through a multi-lamp spot superposition illuminance calculation model and a virtual particle swarm optimization algorithm, multi-lamp collaborative precise control is achieved. Combined with a digital twin light decay compensation model, output deviation is calibrated online to ensure the accuracy of instantaneous physiological equivalent illuminance control. Simultaneously, a correlation is established between control parameters and lighting quality indicators, generating a dynamic lighting control record map, enabling traceability and continuous optimization of the operation process. Therefore, this method solves the problems of low control accuracy and poor lighting comfort in existing technologies. Attached Figure Description

[0016] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart illustrating an intelligent lighting control method based on pedestrian flow monitoring, according to an embodiment of this application. Figure 2 This is a schematic diagram of a smart lighting control system based on pedestrian flow monitoring, according to an embodiment of this application. Figure 3 This is a schematic diagram of a computer structure provided according to an embodiment of this application. Detailed Implementation

[0017] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0019] The following description, with reference to the accompanying drawings, illustrates an intelligent lighting control method and system based on pedestrian flow monitoring, according to an embodiment of this application. Addressing the issue of poor lighting comfort mentioned in the background section, this application provides an intelligent lighting control method based on pedestrian flow monitoring. This method involves real-time acquisition of pedestrian flow distribution and ambient light data in the lighting area, constructing a physically guided decoupling network to accurately separate multi-source optical disturbances, and generating a high-precision holographic thermal imaging map of pedestrian flow. This solves the problems of weak anti-interference capability and insufficient accuracy in pedestrian flow feature recognition inherent in traditional sensing methods. Based on a non-uniform pedestrian flow illuminance demand prediction model, the method predicts the evolution trend of pedestrian flow and integrates visual physiology and circadian rhythm health models to construct a dynamic light field control target, adapting to the differentiated lighting needs under non-uniform pedestrian flow scenarios. Through a multi-lamp spot superposition illuminance calculation model and a virtual particle swarm optimization algorithm, the method achieves precise multi-lamp collaborative control. Combined with a digital twin light decay compensation model, the method calibrates output deviations online, ensuring the accuracy of instantaneous physiological equivalent illuminance control. Simultaneously, it establishes a correlation between control parameters and lighting quality indicators, generating a dynamic lighting control record map, enabling traceability and continuous optimization of the operation process. Thus, it solves the problems of low control accuracy and poor lighting comfort in existing technologies.

[0020] Specifically, Figure 1 This is a flowchart illustrating an intelligent lighting control method based on pedestrian flow monitoring, provided in an embodiment of this application.

[0021] Please see Figure 1 The intelligent lighting control method based on pedestrian flow monitoring includes the following steps: In step S101, real-time pedestrian flow distribution data and ambient light raw data within the lighting area are acquired.

[0022] Among them, the raw ambient light data refers to the natural light intensity, spectral distribution, shadow distribution and optical fluctuation data within the lighting area, which directly affects the basic illuminance level of the area and the accuracy of pedestrian flow recognition.

[0023] It is understood that by acquiring the raw ambient light data within the illumination area in real time, the embodiments of this application can promptly capture the dynamic changes in natural light intensity, spectral distribution, and multi-source optical disturbances, providing real-time input for the subsequent physical guidance decoupling network and holographic crowd flow thermal imaging generation module. This avoids deviations in crowd flow recognition and illuminance calculation caused by natural light fluctuations, shadow occlusion, and high-order optical turbulence interference. At the same time, the background parameters of natural light can correct the regional basic illuminance calculation results, ensuring the accuracy of the dynamic light field control target.

[0024] In step S102, the natural light background, shadow disturbance and higher-order optical turbulence features are adaptively separated from the raw ambient light data through a physical guided decoupling network, and a holographic thermal image of the crowd carrying the trajectory and body posture is generated using a spatiotemporal convolutional recurrent network.

[0025] Among them, the physical-guided decoupling network is a deep learning network that embeds optical physical orthogonal constraints and turbulence regularization mechanisms. It can adaptively separate natural light background, shadow disturbances and high-order optical turbulence features from complex ambient light signals, and achieve accurate decoupling and descrambling of multi-source optical components.

[0026] It is understood that the embodiments of this application use a physically guided decoupling network to process the raw ambient light data, adaptively separating the natural light background, shadow disturbances and high-order optical turbulence features, effectively filtering out multi-source optical interference, providing high signal-to-noise ratio feature inputs for the subsequent spatiotemporal convolutional recurrent network to generate holographic thermal images of people flow, avoiding deviations in people flow recognition and trajectory tracking distortion caused by natural light fluctuations, shadow occlusion and turbulence disturbances; at the same time, the decoupled pure light field components can improve the body detail restoration of the holographic thermal images of people flow, ensuring the accuracy of people flow distribution and movement trend prediction.

[0027] In this embodiment, a physically guided decoupling network adaptively separates the natural light background, shadow perturbation, and higher-order optical turbulence features from the raw ambient light data. A spatiotemporal convolutional recurrent network is then used to generate a holographic thermal image of pedestrian flow carrying trajectories and body postures. The process includes: constructing a multi-source perturbation feature database of ambient light; based on the multi-source perturbation feature database of ambient light, and combining the spatial light intensity distribution, spectral component proportion, and temporal fluctuation amplitude in the raw ambient light data, separating the natural light background component and shadow perturbation component through the orthogonal constraint decomposition layer of the physically guided decoupling network; based on the natural light background component and shadow perturbation component, introducing a higher-order turbulence regularization term, extracting higher-order optical turbulence features in the ambient light, and combining the spatiotemporal convolutional recurrent network to encode and fit the temporal trajectory of pedestrian flow and human body posture features to generate a holographic thermal image of pedestrian flow carrying trajectories and body postures.

[0028] Among them, the spatiotemporal convolutional recurrent network is a deep learning network that integrates the spatial feature extraction capability of convolutional neural networks with the temporal modeling capability of recurrent neural networks. It can encode and fit the temporal movement trajectory of pedestrian flow and human body shape characteristics to generate a holographic thermal image of pedestrian flow carrying complete dynamic information.

[0029] It is understood that the embodiments of this application use a spatiotemporal convolutional recurrent network to process the decoupled light field features and pedestrian signals, simultaneously extracting the spatial details of human body posture and the temporal evolution of pedestrian movement, accurately fitting the movement trajectory and body shape of people, and generating a holographic thermal image of pedestrian flow carrying trajectory and body shape information. This avoids the deviation in pedestrian positioning and the lag in movement trend prediction caused by single spatial recognition or temporal tracking. At the same time, the thermal image that integrates spatiotemporal dual-dimensional information can provide high-quality feature input for the subsequent non-uniform pedestrian flow illumination demand prediction model, ensuring the accuracy of pedestrian aggregation pattern and movement trend prediction.

[0030] Furthermore, based on the full-scene sample prior and ground truth annotation support provided by the ambient light multi-source perturbation feature database, and combined with three types of multi-dimensional feature inputs from the original ambient light data—spatial light intensity distribution representing the spatial brightness pattern, spectral component proportion representing the color composition characteristics, and temporal fluctuation amplitude representing the dynamic change law of light intensity—the orthogonal constraint decomposition layer of the physically guided decoupling network, with the physical laws of optical propagation as its intrinsic constraint, first establishes an orthogonal decomposition model for the ambient light composite signal: ,in Spatial coordinates ,wavelength ,time The original ambient light composite signal at the location; Spatial coordinates ,wavelength ,time The background component of natural light at that location; Spatial coordinates ,wavelength ,time The shadow perturbation component at the location; Spatial coordinates ,wavelength ,time The residual higher-order turbulent perturbation components at the location. Utilizing the inherent properties of the natural light background component—spatially gradual continuity, uniform distribution across the entire spectrum, and low-frequency stability in the temporal domain—and its differences from the shadow perturbation component—which exhibits spatial abrupt changes, selective attenuation in specific bands, and high-frequency fluctuations in the temporal domain, two sets of mutually orthogonal feature subspaces are constructed through algorithmic constraints, and orthogonal subspace projection operations are performed: ,in The natural light background component vector obtained by solving the projection; The shadow perturbation component vector obtained by solving the projection; This is the orthogonal projection matrix corresponding to the feature subspace of natural light background; This is the orthogonal projection matrix corresponding to the shadow perturbation feature subspace; This is the vectorized original ambient light composite signal. Simultaneously, an orthogonal regularization loss function is introduced to constrain the inner product of the features of the two types of components to approach zero. ,in This represents the orthogonal regularization loss value of the orthogonal constraint decomposition layer. The natural light background component vector obtained by solving the projection; The shadow perturbation component vector obtained by solving the projection; This represents the residual higher-order turbulence component vector; This is the vectorized original ambient light composite signal; These are the weighting coefficients for the signal reconstruction error term; This is the transpose operation for a matrix or vector; The L2 norm operation of the vector forces the separation of components to avoid feature aliasing and cross-coupling. This process accurately decouples the pure natural light background component and shadow perturbation component from the complex original ambient light signal, providing high signal-to-noise ratio basic light field data for subsequent high-order turbulence feature extraction and crowd flow thermal imaging reconstruction.

[0031] Based on the orthogonal decoupling of the background component of natural light and the shadow perturbation component, a power spectrum model of atmospheric turbulence refractive index fluctuations is first established based on Kolmogorov's atmospheric turbulence theory and the power-law spectral characteristics of light propagation refractive index fluctuations: ,in The spatial power spectral density of atmospheric turbulent refractive index fluctuations; The atmospheric refractive index structure constant characterizes the intensity level of turbulence. For space wavenumber; This represents the low-frequency cutoff wavenumber corresponding to the external scale of turbulence. Let be the high-frequency cutoff wavenumber corresponding to the internal scale of turbulence. Based on this, a higher-order turbulence regularization term is constructed that integrates fractional-order differential constraints and sparse L1 regularization: ,in This is the penalty value for the higher-order turbulence regularization term; These are residual higher-order turbulent disturbance components; for Fractional-order differential operators; The order of the fractional derivative is used to match the fractal characteristics of turbulence; These are the weighting coefficients for the fractional-order differential constraint terms; These are the weighting coefficients for the L1 sparse constraint terms; This involves L1 norm operations on the vector. The residual light field signal, after subtracting the natural light background and shadow perturbations, is used as input. The high-frequency sparse distribution characteristics of turbulent perturbations are used as a priori constraints. An iterative threshold shrinkage algorithm is employed to solve for the optimal sparse solution. The iterative update formula is as follows: ,in For the first The optimal estimate of the turbulent components output in the next iteration; For the first The estimated turbulence components obtained from the next iteration; For the step size coefficient of the iterative update; Gradient operator for data fidelity terms; This is a soft threshold shrinkage function; The weighting coefficients of the L1 sparse constraint term are used to accurately separate and extract high-order optical turbulence features from the residual light field, such as light spot drift and light intensity flicker caused by air convection, temperature gradient, equipment micro-vibration, etc., and simultaneously perform inverse correction of turbulence disturbances, outputting a pure substrate light field that eliminates all types of optical interference. Subsequently, a spatiotemporal convolutional recurrent network is constructed to complete the holographic thermal imaging reconstruction of the crowd flow. The reconstruction formula is: ,in Spatial coordinates ,time Holographic thermal imaging reconstruction results of the crowd flow at the location; To achieve a pure substrate light field after de-interference for all product categories; for Temporal feature sequence of human body position at a given moment; This is a multi-scale dilated convolutional encoding operation used to extract spatial volumetric features; It is used for bidirectional gated loop unit operations, which are used for time-series trajectory modeling; This is a concatenation and fusion operation based on the feature channel dimension; For deconvolutional decoding upsampling operations used in pixel-level thermal imaging reconstruction, refined spatial body features such as human contours and limb shapes are extracted from different receptive fields through convolutional coding layers. The human position features of continuous frames are temporally modeled through bidirectional GRU recurrent layers to capture the speed, direction and trajectory continuity of people's movement. The spatial body features and temporal trajectory features are fused and stitched together in a high-dimensional feature space at the channel dimension, and then input into the deconvolutional decoding layer for pixel-level upsampling reconstruction. Finally, a holographic thermal image of people flow is generated that simultaneously carries the spatial body details, continuous movement trajectory and thermal radiation distribution information of people. This provides a high-confidence spatiotemporal fusion feature basis for subsequent prediction of non-uniform pedestrian flow illumination demand and forward-looking lighting control.

[0032] In step S103, based on real-time pedestrian flow distribution data, natural light background parameters, and holographic pedestrian flow thermal imaging, a non-uniform pedestrian flow illumination demand prediction model is constructed to predict the pedestrian flow aggregation pattern and movement trend. The individual visual physiological model and rhythmic stimulation curve are integrated to construct a personalized isopaque brightness illumination demand model, forming a dynamic light field control target.

[0033] Among them, the rhythmic stimulation curve is a quantitative relationship curve characterizing the intensity of light stimulation on the human circadian rhythm system (typically the melatonin secretion inhibition effect) under different spectral compositions, illuminance levels and irradiation periods, and is used to support the regulation of rhythm-friendly light parameters in healthy lighting scenarios.

[0034] It is understood that the embodiments of this application, by using rhythmic stimulation curves to quantify the stimulating effects of light on the human diurnal rhythm under different spectra, illuminances, and time periods, can be deeply integrated with individual visual physiological models. While ensuring the clarity of visual tasks, it can accurately match rhythm-friendly light parameters for different crowd scenes and time periods, providing quantitative constraints on the rhythmic health dimension for constructing personalized isopaque brightness illuminance demand models. This avoids the problem of traditional lighting control focusing only on visual brightness while ignoring rhythmic interference, leading to the disruption of people's biological clock and visual fatigue. At the same time, the illuminance and spectrum combination optimized based on rhythmic stimulation curves can take into account both visual comfort and diurnal rhythm health, ensuring the human-centered adaptability and scientific rationality of dynamic light field control targets.

[0035] In this embodiment, a non-uniform pedestrian flow illumination demand prediction model is constructed based on real-time pedestrian flow distribution data, natural light background parameters, and holographic pedestrian flow thermal imaging. This model predicts pedestrian flow aggregation patterns and movement trends. A personalized isopyroluminance illumination demand model is then constructed by integrating individual visual physiological models and rhythmic stimulation curves to form a dynamic light field control target. This includes: constructing a non-uniform pedestrian flow illumination demand prediction model; based on this model, combined with real-time pedestrian flow distribution data, natural light background parameters, and holographic pedestrian flow thermal imaging, predicting pedestrian flow aggregation patterns and movement trends to obtain basic illumination demand thresholds for each area; and based on these thresholds, solving for personalized isopyroluminance illumination demands for each area through fitting individual visual physiological models and rhythmic stimulation curves, extracting spatial illumination distribution gradient features, and fitting a dynamic light field control target that meets visual comfort and rhythmic health requirements.

[0036] Among them, the individual visual physiological model is an ergonomic model based on the spectral response mechanism of human eye photopic vision, scotopic vision and mesovision. It combines the difficulty of visual tasks and viewing conditions to quantify the correspondence between individual visual perception brightness, visual efficacy and illuminance, and spectral parameters. This model is used to support the accurate calculation of personalized visual brightness and illuminance requirements.

[0037] It is understood that the embodiments of this application quantify the perceived brightness and visual efficacy response patterns of the human eye under different illuminance and spectral conditions by using individual visual physiological models. Combining the visual task difficulty and pedestrian activity characteristics of each region, the basic illuminance requirement threshold is modified according to human factors and deeply integrated with the rhythmic stimulus curve to construct a personalized isopaque illuminance requirement model. This avoids the problem that traditional fixed illuminance standards ignore the differences in the human eye's light / dark visual spectral response, leading to insufficient actual visual brightness or excessive lighting. At the same time, the illuminance parameters calibrated based on visual physiological characteristics can accurately match the actual visual perception needs of personnel, ensuring visual task efficiency and visual comfort, effectively reducing visual fatigue in long-term lighting environments, and ensuring that the dynamic light field control target takes into account both visual clarity and physiological health.

[0038] Furthermore, after obtaining the basic illuminance requirement thresholds for each region based on pedestrian density and movement trends, an individual visual physiological model is first introduced. Based on the photopic vision spectral luminous efficiency function V(λ) and the scotopic vision spectral luminous efficiency function V'(λ), an equivalent visual luminance calculation model is constructed: ,in, To match the equivalent visual luminance benchmark value of actual human visual perception; For visual weighting coefficients, These are the scotopic vision weighting coefficients, and both are dynamically adjusted according to the ambient brightness level to match the visual adaptation state. For maximum spectral light efficiency of photopic vision; It has the highest spectral light visibility efficiency for dark vision; λ represents the spectral irradiance at wavelength λ. The spectral luminous efficiency function for photopic vision; For dark-vision spectral luminous efficiency function; The wavelength is in the visible light band; Adjustment factor for visual task difficulty; The reflectivity of the working surface; Pi is a constant used to perform dimensional conversion from incident illuminance to reflected apparent luminance; The algorithm uses a differential operator; it combines the current ambient brightness level to determine the visual adaptation state and matches the corresponding visual effect function weights. Simultaneously, based on the body posture and activity characteristics extracted from the holographic thermal imaging of pedestrian flow, it distinguishes different levels of visual task difficulty, such as walking, lingering, and gathering. It converts the basic illuminance threshold of physical dimensions into an equivalent visual brightness benchmark value that matches the actual visual perception of the personnel, thus completing the illuminance requirement correction for the visual clarity dimension. Furthermore, it integrates rhythmic stimulus curves, using the spectral response law of the human melatonin secretion inhibition rate as the core, to construct a rhythmic stimulus intensity calculation model. ,in, This represents the intensity of light stimulation on the human circadian rhythm. λ represents the spectral irradiance at wavelength λ. Spectral function of the human circadian rhythm; The wavelength is in the visible light band; This refers to the time-period rhythm weighting coefficient; The algorithm uses a differential operator. It combines the time of day and the proportion of natural light to determine the rhythm regulation target. During the day, it moderately increases the proportion of short-wavelength spectrum to enhance rhythmic alertness and improve alertness, while at night, it reduces the weight of short-wavelength spectrum to decrease melatonin suppression and protect the diurnal rhythm. It then optimizes the spectral composition and illuminance amplitude of the visual brightness reference value based on rhythmic health constraints, solving for personalized isovisual brightness illuminance parameters for each area that simultaneously meet visual perception and rhythmic needs. Subsequently, it iterates through all lighting zones to construct a spatial illuminance gradient calculation model. ,in, The spatial illuminance gradient between adjacent partitions i and j; The target illuminance value for partition i; Let j be the target illuminance value for partition j; Let i be the spatial distance between the center points of partition i and partition j. Calculate the ratio of the illuminance difference between adjacent areas to the spatial distance, extract the gradient characteristics of the spatial illuminance distribution, and set a gradient safety threshold to avoid frequent eye adaptation load caused by abrupt changes in brightness in adjacent areas. Use a cubic spline interpolation algorithm to smoothly fit the illuminance and spectral parameters of the partition boundaries. Finally, integrate the target illuminance, target spectral distribution, and spatial gradient constraints of each partition to generate a continuous spatial light field parameter matrix covering the entire area. This forms a dynamic light field control target that simultaneously satisfies visual clarity and comfort, rhythmic health and friendliness, and natural spatial transition, providing accurate target constraint basis for subsequent multi-lamp collaborative light spot superposition calculations.

[0039] In this embodiment of the application, a non-uniform pedestrian flow illumination demand prediction model is constructed, including: constructing a historical database of lighting areas; based on the historical database of lighting areas, combined with real-time pedestrian flow distribution density, aggregation patterns, and body trajectory features in holographic pedestrian flow thermal imaging, extracting the spatiotemporal evolution features of the dynamic distribution of pedestrian flow through a spatiotemporal graph convolution algorithm; based on the spatiotemporal evolution features of the dynamic distribution of pedestrian flow, introducing a natural light background illumination compensation term, fitting the regional illumination demand curves under different pedestrian flow scenarios, and establishing a non-uniform pedestrian flow illumination demand prediction model.

[0040] Among them, the natural light background illuminance compensation item is an artificial light supplementary illuminance value calculated based on the difference between the natural light background illuminance obtained by decoupling and separation and the regional target illuminance in dynamic lighting control. It is used to offset natural light fluctuations, ensure stable illuminance on the working surface, and achieve energy-saving control.

[0041] It is understood that the embodiments of this application introduce a natural light background illuminance compensation term, embedding the real-time natural light background illuminance obtained by decoupling and separation as a constraint variable into the illuminance demand fitting process. Based on fitting the regional illuminance demand curve based on the dynamic spatiotemporal evolution characteristics of pedestrian flow, the effective contribution of natural light to illuminance in each region is accurately calculated. The difference between the target illuminance and the natural light background is dynamically calculated as the artificial light supplementation benchmark, effectively offsetting the illuminance prediction deviation caused by temporal fluctuations and uneven spatial distribution of natural light, and avoiding over-illumination and energy waste caused by the traditional prediction model ignoring the contribution of natural light. At the same time, it can adapt to regional scenes with different pedestrian densities and aggregation patterns, matching differentiated compensation weights so that the illuminance demand prediction can simultaneously meet the visual comfort needs of people and the energy-saving goal of maximizing the utilization of natural light, significantly improving the accuracy and energy-saving adaptability of illuminance demand prediction in non-uniform pedestrian flow scenarios.

[0042] In step S104, based on the dynamic light field control target, and combined with the inverse square law of illuminance and the cosine law of incident angle, a multi-lamp spot superposition illuminance calculation model is constructed. The multi-parameter collaborative solution is performed through the virtual particle swarm optimization algorithm. After the illuminance accuracy and glare value gradient verification are performed, a multi-lamp collaborative dynamic control command is generated.

[0043] Among them, the virtual particle swarm optimization algorithm is an improved swarm intelligence optimization algorithm. It optimizes the search strategy by introducing virtual particle individuals or virtual extreme value guidance mechanisms. It iteratively updates the velocity and position of particles in the feasible solution space to search for the global optimal solution. It can effectively avoid premature convergence problems and improve the solution accuracy and convergence efficiency of complex optimization problems.

[0044] It is understood that the embodiments of this application utilize a virtual particle swarm optimization algorithm to map the multi-dimensional control parameters of each lighting fixture, such as luminous flux output, projection angle, and spectral ratio, into the position vector of the particle swarm. The dynamic light field control target is used as the core fitness function, combined with a light spot superposition illuminance calculation model constructed based on the inverse square law of illuminance and the cosine law of incident angle as physical constraints. The optimization of the population search strategy is achieved through the guidance mechanism of virtual particle individuals and virtual extrema, effectively avoiding the premature convergence problem of traditional particle swarm algorithms that easily get trapped in local optima. This efficiently completes global collaborative optimization for multiple lighting fixtures and multiple parameters. Simultaneously, gradient verification constraints for illuminance accuracy deviation and unified glare value can be embedded in the iterative optimization process to ensure that the optimal parameter combination output simultaneously meets the requirements of light field distribution accuracy and glare control, significantly improving the solution efficiency, parameter adaptation accuracy, and human comfort of multi-lamp collaborative control.

[0045] In this embodiment of the application, the formula for calculating the illuminance of multiple light spot superposition is as follows: The formula for calculating the horizontal illuminance of a single point light source is: ; In the formula, Let i be the base horizontal illuminance produced by the i-th lamp in the j-th lighting zone; The initial luminous intensity of the light source; The angle between the incident direction of the light ray and the normal to the horizontal plane; The azimuth angle of the light distribution of the lamp; This is the straight-line distance from the light source of the luminaire to the calculation point of the lighting zone; The pedestrian flow distribution weighting correction term is: ; In the formula, The pedestrian flow distribution weighting coefficient for the j-th lighting zone; This serves as a baseline coefficient for personnel density. Let be the real-time pedestrian density of the j-th partition; This is a correction factor for movement speed; Let be the average movement speed of people in the j-th partition; Combining the above two equations, we derive the formula for calculating the illuminance of multiple light spots superimposed on each other: ; In the formula, Let be the instantaneous equivalent illuminance of the j-th partition; Let be the actual output light intensity of the i-th lamp at time t; Let be the angle of incidence of the light from the i-th lamp at time t; Let be the light distribution azimuth angle of the i-th lamp at time t; Let be the distance from the i-th lamp to the calculation point at time t; The total number of luminaires involved in lighting control.

[0046] It is understood that the embodiments of this application utilize a multi-lamp spot superposition illuminance calculation model to accurately quantify the illuminance contribution of each lamp at various points in space. Through linear superposition operations, it quickly solves the continuous illuminance distribution and spatial gradient characteristics of the entire area under the synergistic effect of multiple lamps, providing accurate physical calculation basis for the multi-parameter collaborative optimization algorithm of virtual particle swarm optimization. This replaces the traditional empirical illuminance estimation method, significantly improving the calculation accuracy and parameter extrapolation efficiency of light field distribution. At the same time, the model can adapt to the dynamic adjustment of parameters such as luminous flux, projection angle, and installation height of lamps in real time, quickly verifying the illuminance compliance rate, spatial uniformity, and transition effect of partition boundaries under different control schemes. Combined with the unified glare value gradient verification process, it simultaneously ensures the visual comfort and quality stability of the light field. This not only ensures that the final control result accurately matches the dynamic light field control target and avoids local over-illumination or under-illumination problems, but also provides reliable physical calculation support for the generation of multi-lamp collaborative dynamic control commands in non-uniform pedestrian flow scenarios, taking into account the multiple needs of lighting quality, human health, and energy-saving control.

[0047] In step S105, according to the multi-lamp collaborative dynamic control command, the output power, light distribution angle and projection orientation of the lamps in the lighting circuit are adjusted, the output deviation is calibrated online through the digital twin light decay prediction compensation model, the instantaneous physiological equivalent illuminance during the dynamic change of the pedestrian flow is controlled, and a dynamic lighting control record map is generated.

[0048] Among them, the digital twin light decay prediction and compensation model is an intelligent operation and maintenance model that uses a digital twin of the lighting system to construct a virtual-real synchronous mapping, predicts the light flux decay trend of the lamps through a self-supervised learning algorithm and dynamically adjusts the output parameters for compensation, thus ensuring the long-term stability of the illuminance on the work surface.

[0049] It is understood that the embodiments of this application utilize a digital twin light decay prediction and compensation model. Based on a virtual-real synchronous mapping mechanism, it compares the simulated light field output of the digital twin with the actual operating parameters of the on-site lighting fixtures in real time. It calibrates the execution deviation of control commands such as the output power, light distribution angle, and projection orientation of the lighting fixtures online. At the same time, it uses a self-supervised learning algorithm to mine the light decay evolution law in the historical operating data of the lighting fixtures, accurately predicts the light flux decay amplitude of different service cycles, and dynamically generates power compensation. This effectively offsets the long-term illuminance decay problem caused by lighting fixture aging and environmental damage. It can participate in the instantaneous physiological equivalent illuminance control during the dynamic change of pedestrian flow, ensuring that the illuminance and spectral parameters always match the dual goals of visual comfort and circadian rhythm health. It can also simultaneously accumulate full-cycle control data to generate a dynamic lighting control record map, which not only greatly improves the long-term stability and execution accuracy of dynamic light field control, but also reduces the maintenance cost of manual inspection and calibration.

[0050] In this embodiment of the application, generating a dynamic lighting control record map includes: acquiring the actual illuminance of the area, luminaire operating parameters, pedestrian flow distribution data, and visual comfort data during the control process; constructing a lighting control effect evaluation model based on the actual illuminance of the area, luminaire operating parameters, pedestrian flow distribution data, and visual comfort data, and calculating illuminance uniformity, uniform glare value, energy saving rate, and visual satisfaction quality indicators; and associating the lighting control parameters, luminaire operating status, and quality indicators along a time axis to generate a dynamic lighting control record map that marks control nodes, abnormal events, and comfort fluctuations.

[0051] Among them, the lighting control effect evaluation model is an analytical model that comprehensively evaluates the operation effect of dynamic lighting control schemes and supports strategy iteration and optimization by quantifying and calculating multiple dimensions such as illuminance compliance rate, spatial uniformity, human comfort, rhythm adaptability and energy efficiency.

[0052] It is understood that the embodiments of this application utilize a lighting control effect evaluation model. Based on multi-source data such as actual illuminance, luminaire operating parameters, pedestrian flow distribution, and visual comfort collected during the control process, the model calculates core quality indicators such as illuminance uniformity, uniform glare value, energy saving rate, and visual satisfaction. This enables a multi-dimensional quantitative evaluation of the operational effect of dynamic lighting control schemes. It can accurately identify problems such as uneven illuminance in local areas, excessive glare, redundant energy consumption, or insufficient human comfort, providing clear data support for online calibration of control parameters and strategy iteration. It can also deeply correlate various quality indicators with lighting control parameters and luminaire operating status along the time axis, helping the management to intuitively grasp the changes in control effectiveness and user experience throughout the entire cycle, and continuously optimize the light field control strategy in non-uniform pedestrian flow scenarios.

[0053] Furthermore, when constructing the lighting control effect evaluation model, the model first uses multi-source inputs: the actual illuminance time-series data collected by distributed photoelectric sensors, the operating parameters such as lamp output power, dimming depth, light distribution angle, and running time transmitted back by the intelligent drive circuit, the zonal pedestrian density, dwell time, and movement trajectory distribution data obtained from holographic pedestrian thermal imaging analysis, and the visual comfort quantification data calculated by integrating physiological equivalent illuminance and rhythmic stimulus adaptation. This allows for the construction of a three-dimensional hierarchical evaluation index system encompassing luminous quality, human comfort, and energy efficiency, thus completing the framework construction and index weight configuration of the evaluation model. In the index calculation stage, illuminance uniformity follows the CIE lighting standard calculation paradigm. First, the arithmetic mean and minimum illuminance values ​​of all measuring points on the working surface are calculated. The point illuminance uniformity is calculated by the ratio of minimum illuminance to average illuminance. Simultaneously, the mean illuminance gradient between adjacent measuring points is calculated to characterize the smoothness of spatial transition. Measuring points in densely populated areas are given higher weights to match actual usage needs. The unified glare value is calculated strictly according to the CIE117 standard model, incorporating the luminance of each luminaire's luminous surface, solid angle of observation direction, position index, and ambient background luminance parameters. The value is iteratively solved point-by-point, and the maximum value within the region is taken as the final result. Corrections are made based on typical viewing height and angle in the personnel's residence area to improve the scene adaptability of the glare assessment. The energy saving rate is based on... The quasi-comparative method is used to calculate energy consumption. First, a theoretical benchmark energy consumption model for the traditional fixed illuminance mode under the same pedestrian flow scenario and illuminance standard is constructed. The benchmark energy consumption value corresponding to hourly natural light compensation and pedestrian flow demand is calculated. Then, the actual total energy consumption of the lamps during the control cycle is statistically analyzed. The energy saving rate is calculated by the ratio of the difference between the benchmark energy consumption and the actual energy consumption. The calculation process includes drive loss and standby loss to ensure the accuracy of energy consumption calculation. Visual satisfaction is calculated using a weighted comprehensive evaluation method. After normalization, five sub-indicators, namely illuminance compliance rate, glare compliance rate, rhythmic stimulation adaptation, illuminance gradient comfort, and pedestrian flow-illuminance matching degree, are weighted and fused according to preset weights to obtain a quantitative satisfaction score in the range of 0-100. It comprehensively covers the multi-dimensional human needs of visual clarity, glare control, and circadian rhythm health. After calculating all quality indicators, a time-series correlation matrix is ​​constructed with a unified timestamp as the alignment benchmark. The lighting control parameters, real-time operating status of lamps, and various quality indicators are mapped and bound to each time point. Through a threshold detection algorithm, the dimming action nodes corresponding to sudden changes in control parameters, the abnormal operation events corresponding to exceeding the limits of indicators, and the comfort fluctuation range where satisfaction is continuously lower than the threshold are automatically identified and labeled. Finally, all time-series data and event labels are integrated along the time axis to generate a traceable and quantifiable dynamic lighting control record map, providing complete data support for the review and iterative optimization of control strategies.

[0054] In this embodiment, the output power, light distribution angle, and projection orientation of the luminaires in the lighting circuit are adjusted, and the output deviation is calibrated online through a digital twin light decay prediction and compensation model to control the instantaneous physiological equivalent illuminance during the dynamic change phase of pedestrian flow. This includes: constructing a digital twin light decay prediction and compensation model; based on the digital twin light decay prediction and compensation model, adjusting the output power of the luminaires through a dimming drive module and adjusting the light distribution angle and projection orientation through a stepper motor according to the multi-lamp collaborative dynamic control command, controlling the instantaneous physiological equivalent illuminance of each zone; when the illuminance deviation of a zone exceeds a preset threshold, triggering a self-calibration compensation mechanism, correcting the luminaire output parameters based on the light decay prediction results, and generating an abnormal warning signal.

[0055] It is understood that the embodiments of this application, by combining a digital twin light decay prediction and compensation model with a multi-dimensional control mechanism for luminaires, precisely adjust the output power, light distribution angle, and projection orientation of the luminaires by dimming drive and stepper motor according to the multi-lamp collaborative dynamic control command. This allows for precise control of the instantaneous physiological equivalent illuminance of each zone throughout the entire time period of dynamic changes in pedestrian flow, ensuring that the light environment matches the dual needs of pedestrian flow distribution, visual clarity, and circadian rhythm health in real time. Relying on the virtual-real synchronous mapping capability of the digital twin, the deviation between the actual output of the luminaires and the target parameters can be verified online. When the illuminance deviation of a zone exceeds a preset threshold, a self-calibration compensation mechanism is automatically triggered. Based on the light decay prediction results, the output parameters of the luminaires are dynamically corrected, effectively offsetting the luminous flux attenuation caused by long-term service of the luminaires and the illuminance drift caused by the aging of optical components. This ensures the long-term stability of the light field control accuracy and simultaneously generates anomaly warning signals to assist maintenance personnel in quickly locating fault points. This significantly improves the real-time response accuracy and long-term operational reliability of dynamic lighting control, while also reducing the maintenance cost of manual inspection and calibration.

[0056] Furthermore, the virtual-real synchronous mapping mechanism of the digital twin light decay prediction and compensation model receives and parses multi-lamp collaborative dynamic control commands, decomposes the target light field parameters into single-lamp level output power, beam distribution angle, and projection orientation control quantities, and sends them to the execution end. The dimming drive module linearly dims the LED driver power supply via DALI or 0-10V dimming protocols, precisely adjusting the luminous flux output of the lamps to match the target illuminance amplitude. The stepper motor drive module controls the axial displacement of the lamp's optical lens group and the horizontal rotation and vertical pitch mechanisms of the lamp body, synchronously adjusting the beam distribution angle and beam projection orientation to achieve dynamic adaptation of light intensity distribution and illumination area. During this process, real-time data from the spectral illuminance sensors of each zone is accessed, and instantaneous physiological equivalent illuminance values ​​are calculated based on the photopic vision function and the circadian rhythm spectral conversion, forming a transient control link to precisely manage the visual and circadian dual light environment indicators of each zone during dynamic changes in pedestrian flow. When the real-time monitored zone... When the relative deviation between the physiological equivalent illuminance and the target value exceeds a preset threshold, a self-calibration compensation mechanism is automatically triggered. First, the digital twin calls the light decay prediction model trained by self-supervised learning, and calculates the current luminous flux attenuation rate and optical loss by combining parameters such as the cumulative running time of the luminaire, the junction temperature of the light source, and the ambient temperature and humidity. The output pre-compensation correction amount is superimposed on the original control command, and then the instantaneous deviation is corrected in a closed loop through power fine-tuning and angle fine-shifting, quickly pulling the illuminance deviation back to the allowable range. If the deviation continues to exceed the limit after multiple rounds of compensation calibration, or if the single lamp control parameter reaches the output upper limit but still cannot meet the standard, it is determined that the luminaire light source is aging, the optical device is faulty, or the circuit is abnormal. At the same time, an abnormal warning signal containing the fault location, deviation value, and duration is generated and pushed to the operation and maintenance management platform to support rapid troubleshooting. In this way, while ensuring the real-time control accuracy of the dynamic light field, the long-term light decay effect is offset by predictive compensation, thereby improving the operational stability and operation and maintenance efficiency of the system throughout its entire life cycle.

[0057] This application proposes an intelligent lighting control method based on pedestrian flow monitoring. By real-time acquisition of pedestrian flow distribution and ambient light data in the lighting area, a physically guided decoupling network is constructed to accurately separate multi-source optical disturbances, generating a high-precision holographic thermal image of pedestrian flow. This solves the problems of weak anti-interference capability and insufficient accuracy of pedestrian flow feature recognition in traditional sensing methods. Based on a non-uniform pedestrian flow illuminance demand prediction model, the method predicts the evolution trend of pedestrian flow and integrates visual physiology and circadian rhythm health models to construct dynamic light field control targets, adapting to the differentiated lighting needs in non-uniform pedestrian flow scenarios. Through a multi-lamp spot superposition illuminance calculation model and a virtual particle swarm optimization algorithm, multi-lamp collaborative precise control is achieved. Combined with a digital twin light decay compensation model, output deviation is calibrated online to ensure the accuracy of instantaneous physiological equivalent illuminance control. Simultaneously, a correlation is established between control parameters and lighting quality indicators, generating a dynamic lighting control record map, enabling traceability and continuous optimization of the operation process. Therefore, this method solves the problems of low control accuracy and poor lighting comfort in existing technologies.

[0058] The following will illustrate a smart lighting control method based on pedestrian flow monitoring through a specific embodiment, including: The application scenario is based on the core maintenance depot and storage area of ​​a 32,000-square-meter EMU depot. This area encompasses five functional zones: inspection lines, storage lines, maintenance trenches, side crossings, and personnel gathering and dispersal areas. During peak periods, multiple activities overlap, including personnel operations, train entry and exit, and shunting track changes, with hundreds of personnel on duty at any given time. Dynamic events such as train arrivals, wheel stops for maintenance, and joint commissioning tests also occur. The quantitative and qualitative requirements for lighting differ significantly between zones and must be deeply coordinated with train arrival and exit times and maintenance workstation scenarios. A total of 128 intelligent dimmable LED high-bay lights and 96 adjustable-angle floodlights are deployed within the area, with a single lamp rated power ranging from 30 to 120 watts. All lights are equipped with DALI dimming drive modules and two-dimensional stepper motorized pan-tilt units, supporting linear adjustment of output power and multi-angle projection adjustment. The ceiling is equipped with spectral illuminance sensors arranged in an eight-meter grid. Thermal imaging cameras for passenger flow and train occupancy detection are installed at the depot doors and key passageways. High-definition binocular vision cameras are deployed in inspection trenches and key work areas to capture personnel's working postures and gathering patterns. The entire lighting system is connected to the EMU depot's intelligent management and control platform, and integrates entry and exit plans, track occupancy, and maintenance time information provided by the train dispatching system. It supports timetable-based scene presets and real-time dynamic light field fine-tuning, meeting multiple requirements for visual accuracy, personnel circadian rhythm health, and energy conservation under different work scenarios.

[0059] To achieve precise intelligent lighting control in complex spatiotemporal scenarios involving personnel and trains, a comprehensive data acquisition system is constructed, encompassing four categories: real-time personnel and train distribution, raw ambient light data, lighting fixture operating status, and visual comfort feedback. Real-time personnel and train distribution data is jointly collected via thermal imaging cameras, binocular vision cameras, and a train dispatching interface. The thermal imaging and vision cameras output data on personnel density, dwell time, and movement trajectories for each zone, identifying four behavioral states: walking, standing, gathering, and standardized operations. Simultaneously, the system integrates train entry / exit plans, track occupancy signals, and real-time train positions to obtain the train's track location, speed, and estimated stopping time. Based on this, the system divides the entire area into dynamic operational scenarios such as train reception and preparation, pit maintenance, roof inspection, waiting for departure, and unmanned inspection. The statistical error for personnel in a single area is controlled within 3%, and the accuracy of train positioning and track matching is no less than 99%. Raw ambient light data is collected via a gridded spectral illuminance sensor, covering the entire visible light spectrum, capturing natural light fluctuations introduced by depot door openings, optical disturbances caused by building shadows, and train body obstruction. The operating status data of the lighting fixtures is transmitted in real time via the DALI bus, including parameters such as output power, dimming depth, and projection angle, with a transmission latency of no more than 100 milliseconds. Visual comfort data is acquired through human factors data acquisition terminals worn by operators, and comfort quantification values ​​are calculated based on physiological indicators such as blink frequency and pupil diameter changes. Data processing employs sliding window filtering to remove outliers, and transmission and storage utilize an edge gateway plus local server architecture. Edge processing latency is controlled within 50 milliseconds, while the local server stores 24 months of historical data with daily automatic backups to ensure data security and traceability.

[0060] Based on Python 3.8 and the PyTorch deep learning framework, a joint processing architecture combining a physics-guided decoupling network and a spatiotemporal convolutional recurrent network was constructed to separate ambient light components and generate holographic thermal images of people and trains. An ambient light component sample database was established initially, collecting raw and baseline data under different time periods, weather conditions, depot door opening / closing states, and personnel and train distributions, totaling 120,000 labeled samples. The physics-guided decoupling network, using spectral data as input, embeds physical constraints on natural light radiative transmission, adaptively separating the natural light background illuminance, shadow disturbances caused by train obstruction, and higher-order optical turbulence features, with a natural light background calculation error of less than 4%. The spatiotemporal convolutional recurrent network, using thermal imaging, personnel trajectory, and train occupancy status data as input, extracts personnel aggregation patterns, train movement trends, and interaction evolution patterns through spatiotemporal convolution kernels, generating holographic thermal images of people and trains carrying motion trajectories and body features, accurately distinguishing multiple composite forms such as single-person inspections, group operations, low-speed train passage, and wheel-stopping maintenance. Based on this, a non-uniform personnel-train illumination demand prediction model is constructed. Inputs include personnel density, aggregation characteristics, train position and speed, natural light background, and scheduling timetable. A spatiotemporal evolution feature is extracted using a spatiotemporal graph convolution algorithm to predict the distribution pattern and movement trend of personnel and trains within 15 minutes. The average absolute error of personnel density prediction is less than 0.2 people per square meter, and the deviation of train position prediction does not exceed 0.5 meters. Integrating individual visual physiological models and rhythmic stimulus curves, a natural light background illumination compensation term is introduced to fit regional illumination demand curves under different scenarios. The personalized isopaque brightness parameters for each zone are solved, and combined with spatial illumination gradient smoothing constraints, a dynamic light field control target that meets visual comfort and rhythmic health requirements is ultimately formed and pushed to the lighting management platform as a constraint basis.

[0061] Based on the dynamic light field control target, and combining the inverse square law of illuminance and the cosine law of incident angle, a multi-lamp spot superposition illuminance calculation model is constructed. Input parameters include the flux of a single lamp, installation height, projection angle, and reflectivity of the working surface. The illuminance contribution of a single lamp at each measuring point is calculated point-by-point. The full-area illuminance distribution under multi-lamp collaboration is solved through linear superposition, with the calculation error controlled within ±3%. A virtual particle swarm optimization algorithm is used for multi-parameter collaborative solution. The output power, horizontal projection angle, and vertical pitch angle of each lamp are encoded as particle position vectors. The fitness function is a weighted comprehensive index of regional illuminance compliance rate, spatial uniformity, glare control value, and scene mode matching degree. The scene mode matching degree is constrained by the baseline illuminance and brightness range of predefined scenes such as vehicle preparation, pit maintenance, and vehicle waiting to depart. A virtual global extremum guidance mechanism is introduced to optimize the search strategy, avoiding the algorithm from getting trapped in local optima. After iterative optimization, the globally optimal lamp parameter combination is output. After parameter calculation, a two-dimensional verification is performed. The illuminance accuracy verification ensures that the deviation of more than 95% of the measuring points is within ±10%. The unified glare value verification calculates the values ​​of each typical viewpoint according to the CIE117 standard to ensure that the glare values ​​in the maintenance work area and the passage area meet the specification requirements. After the verification is passed, a standardized multi-lamp collaborative dynamic control command is generated and sent to the execution terminal of each lamp via the DALI bus and the pan-tilt control protocol. The total process takes no more than 0.8 seconds, which can meet the real-time control needs under dynamic changes of personnel and trains.

[0062] Based on the multi-lamp collaborative dynamic control commands, the system, combined with train arrival and departure times and maintenance plans, pre-sets the depot lighting into typical scenarios such as train reception mode, pit maintenance mode, rooftop operation mode, waiting-to-departure mode, and clearing and inspection mode. Each scenario has preset reference illuminance and color temperature ranges. On this basis, the dimming drive module linearly adjusts the lamp output power, and the stepper motor precisely adjusts the light distribution angle and projection direction, synchronously controlling the instantaneous physiological equivalent illuminance of each zone. This indicator is calculated based on the photopic vision spectral luminous efficiency function and the spectral weighting of diurnal rhythm effects, simultaneously reflecting visual clarity and the intensity of rhythmic stimulation. To ensure long-term control accuracy, a digital twin light decay prediction and compensation model is constructed. Based on historical lamp operating data and environmental temperature and humidity parameters, a self-supervised learning algorithm is used to fit the luminous flux decay curve, predicting the light decay amplitude at different service cycles. The model prediction error is less than 2.5%. During the control process, the digital twin synchronizes the operating status of the lighting fixtures in real time, comparing the deviation between the measured illuminance and the target illuminance. When the illuminance deviation of a zone exceeds a preset threshold of ±8%, a self-calibration compensation mechanism is triggered. Based on the light decay prediction results, the output parameters are corrected to quickly offset the illuminance drift caused by light decay and loss. If the deviation still exceeds the limit after two rounds of compensation, it is determined to be an equipment malfunction and an early warning signal is generated and pushed to the operation and maintenance end. Multi-source data is continuously collected throughout the operation to build a lighting control effect evaluation model and calculate four core quality indicators: illuminance uniformity, unified glare value, energy saving rate, and visual satisfaction. The control parameters, operating status, and quality indicators are correlated and mapped along the time axis, automatically marking dimming action nodes, train entry and exit times, equipment malfunction events, and comfort fluctuation periods, generating a dynamic lighting control record map to support strategy backtracking and problem localization.

[0063] In summary, this invention provides reliable data support for intelligent lighting control by collecting comprehensive data on personnel and train distribution, ambient light, lighting operation, and visual comfort, combined with an edge gateway and local storage architecture. It separates natural light background and optical disturbances using a physically guided decoupling network, generates holographic personnel-train thermal images using a spatiotemporal convolutional recurrent network, infers personnel and train distribution trends using an illuminance demand prediction model, and generates dynamic light field targets by integrating visual physiology and circadian rhythm stimulation models, achieving on-demand dimming and efficient utilization of natural light. Based on a multi-light spot overlay model and a virtual particle swarm optimization algorithm, it performs global optimization of multiple parameters, issuing commands after double verification of illuminance accuracy and unified glare value, ensuring light field uniformity and human comfort. During the control phase, it uses a digital twin light decay prediction model to calibrate deviations online, and a performance evaluation model to quantify core indicators, generating a dynamic control record map. This improves lighting quality, significantly reduces energy consumption and maintenance costs, extends lighting fixture lifespan, and balances visual comfort, circadian rhythm health, and energy-saving benefits.

[0064] This invention also proposes an intelligent lighting control system based on pedestrian flow monitoring; please refer to [link / reference needed]. Figure 2 , Figure 2This is a schematic diagram of the structure of an intelligent lighting control system based on pedestrian flow monitoring, according to an embodiment of this application.

[0065] The intelligent lighting control system 10 based on pedestrian flow monitoring includes: an acquisition module 100, a separation module 200, a prediction module 300, a generation module 400, and a control module 500.

[0066] The system comprises the following modules: Acquisition module 100 acquires real-time pedestrian flow distribution data and ambient light raw data within the illuminated area; Separation module 200 adaptively separates natural light background, shadow disturbances, and higher-order optical turbulence features from the ambient light raw data using a physically guided decoupling network, and generates a holographic thermal image of pedestrian flow carrying trajectories and body postures using a spatiotemporal convolutional recurrent network; Prediction module 300 constructs a non-uniform pedestrian flow illuminance demand prediction model based on real-time pedestrian flow distribution data, natural light background parameters, and the holographic pedestrian flow thermal image, predicts pedestrian flow aggregation patterns and movement trends, and integrates individual visual physiological models and rhythmic stimulus curves to construct a personalized isopyroluminance illuminance demand model. The system generates a dynamic light field control target. The generation module 400, based on this target and combining the inverse square law of illuminance and the cosine law of incident angle, constructs a multi-lamp spot superposition illuminance calculation model. It then uses a virtual particle swarm optimization algorithm to solve multiple parameters collaboratively. After illuminance accuracy and glare value gradient verification, it generates a multi-lamp collaborative dynamic control command. The control module 500, according to the multi-lamp collaborative dynamic control command, adjusts the output power, light distribution angle, and projection orientation of the lamps in the lighting circuit. It uses a digital twin light decay prediction compensation model to calibrate output deviations online, controlling the instantaneous physiological equivalent illuminance during the dynamic changes in pedestrian flow, and generating a dynamic lighting control record map.

[0067] Furthermore, the functions or operation steps implemented by the above modules and units when executed are largely the same as those in the above method embodiments, and will not be repeated here.

[0068] The intelligent lighting control system based on pedestrian flow monitoring provided in this embodiment of the invention has the same implementation principle and technical effects as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the system embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0069] According to the embodiments of this application, an intelligent lighting control system based on pedestrian flow monitoring is proposed. This system collects real-time pedestrian flow distribution and ambient light data in the lighting area, constructs a physically guided decoupling network to accurately separate multi-source optical disturbances, and generates a high-precision holographic thermal image of pedestrian flow. This solves the problems of weak anti-interference capability and insufficient accuracy of pedestrian flow feature recognition in traditional sensing methods. Based on a non-uniform pedestrian flow illuminance demand prediction model, it predicts the evolution trend of pedestrian flow and integrates visual physiology and circadian rhythm health models to construct dynamic light field control targets, adapting to the differentiated lighting needs under non-uniform pedestrian flow scenarios. Through a multi-lamp spot superposition illuminance calculation model and a virtual particle swarm optimization algorithm, it achieves precise multi-lamp collaborative control. Combined with a digital twin light decay compensation model, it calibrates output deviations online to ensure the accuracy of instantaneous physiological equivalent illuminance control. Simultaneously, it establishes a correlation between control parameters and lighting quality indicators, generating a dynamic lighting control record map, enabling traceability and continuous optimization of the operation process. Therefore, it solves the problems of low control accuracy and poor lighting comfort in existing technologies.

[0070] This invention also proposes a computer, please refer to [link / reference]. Figure 3 The diagram shows a computer structure in an embodiment of the present invention, including a memory 10, a processor 20, and a computer program 30 stored in the memory 10 and executable on the processor 20. When the processor 20 executes the computer program 30, it implements the above-mentioned intelligent lighting control method based on pedestrian flow monitoring.

[0071] The memory 10 includes at least one type of readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 10 can be an internal storage unit of a computer, such as the computer's hard disk. In other embodiments, the memory 10 can be an external storage device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Furthermore, the memory 10 can include both internal and external storage units of the computer. The memory 10 can be used not only to store application software and various types of data installed on the computer, but also to temporarily store data that has been output or will be output.

[0072] In some embodiments, the processor 20 may be an electronic control unit (ECU), a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip, used to run program code stored in the memory 10 or process data, such as executing access restriction programs.

[0073] It should be pointed out that, Figure 3 The structure shown does not constitute a limitation on the computer. In other embodiments, the computer may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0074] This invention also proposes a readable storage medium storing a computer program that, when executed by a processor, implements the intelligent lighting control method based on pedestrian flow monitoring as described above.

[0075] Those skilled in the art will understand that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a ordered 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-included 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 mean 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.

[0076] 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.

[0077] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0078] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0079] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this invention should be determined by the appended claims.

Claims

1. A smart lighting control method based on pedestrian flow monitoring, characterized in that, include: Acquire real-time pedestrian traffic distribution data and raw ambient light data within the illuminated area; A physical-guided decoupling network adaptively separates the natural light background, shadow disturbances, and higher-order optical turbulence features from the raw ambient light data, and a spatiotemporal convolutional recurrent network is used to generate a holographic thermal image of the crowd carrying trajectory and body posture. Based on the real-time pedestrian flow distribution data, natural light background parameters and holographic pedestrian flow thermal imaging, a non-uniform pedestrian flow illumination demand prediction model is constructed to predict the pedestrian flow aggregation pattern and movement trend. A personalized isopaque brightness illumination demand model is constructed by integrating individual visual physiological models and rhythmic stimulation curves to form a dynamic light field control target. Based on the aforementioned dynamic light field control target, and combining the inverse square law of illuminance and the cosine law of incident angle, a multi-lamp spot superposition illuminance calculation model is constructed. Multi-parameter collaborative solution is performed through virtual particle swarm optimization algorithm. After illuminance accuracy and glare value gradient verification, multi-lamp collaborative dynamic control command is generated. According to the multi-lamp collaborative dynamic control command, the output power, light distribution angle and projection orientation of the lamps in the lighting circuit are adjusted. The output deviation is calibrated online through the digital twin light decay prediction compensation model to control the instantaneous physiological equivalent illuminance during the dynamic change of the pedestrian flow and generate a dynamic lighting control record map.

2. The intelligent lighting control method based on pedestrian flow monitoring according to claim 1, characterized in that, A physically guided decoupling network adaptively separates natural light background, shadow disturbances, and higher-order optical turbulence features from the raw ambient light data. A spatiotemporal convolutional recurrent network is then used to generate a holographic thermal image of the crowd carrying trajectory and posture, including: Construct a database of multi-source ambient light disturbance characteristics; Based on the aforementioned ambient light multi-source perturbation feature database, and combined with the spatial light intensity distribution, spectral component proportion, and temporal fluctuation amplitude in the original ambient light data, the natural light background component and shadow perturbation component are separated through the orthogonal constraint decomposition layer of the physically guided decoupling network. Based on the background component of natural light and the shadow perturbation component, a higher-order turbulence regularization term is introduced to extract higher-order optical turbulence features in ambient light. Combined with a spatiotemporal convolutional recurrent network, the temporal trajectory of pedestrian flow and human body posture features are encoded and fitted to generate a holographic thermal image of pedestrian flow carrying trajectory and posture.

3. The intelligent lighting control method based on pedestrian flow monitoring according to claim 1, characterized in that, The formula for calculating the illuminance of the multi-light spot superposition model is as follows: The formula for calculating the horizontal illuminance of a single point light source is: ; In the formula, Let i be the base horizontal illuminance produced by the i-th lamp in the j-th lighting zone; The initial luminous intensity of the light source; The angle between the incident direction of the light ray and the normal to the horizontal plane; The azimuth angle of the light distribution of the lamp; This is the straight-line distance from the light source of the luminaire to the calculation point of the lighting zone; The pedestrian flow distribution weight correction term is: ; In the formula, The pedestrian flow distribution weighting coefficient for the j-th lighting zone; This serves as a baseline coefficient for personnel density. Let be the real-time pedestrian density of the j-th partition; This is a correction factor for movement speed; Let be the average movement speed of people in the j-th partition; Combining the above two equations, we derive the formula for calculating the illuminance of multiple light spots superimposed on each other: ; In the formula, Let be the instantaneous equivalent illuminance of the j-th partition; Let be the actual output light intensity of the i-th lamp at time t; Let be the angle of incidence of the light from the i-th lamp at time t; Let be the light distribution azimuth angle of the i-th lamp at time t; Let be the distance from the i-th lamp to the calculation point at time t; The total number of luminaires involved in lighting control.

4. The intelligent lighting control method based on pedestrian flow monitoring according to claim 1, characterized in that, Based on the real-time pedestrian flow distribution data, natural light background parameters, and holographic thermal imaging of pedestrian flow, a non-uniform pedestrian flow illumination demand prediction model is constructed to predict pedestrian flow aggregation patterns and movement trends. A personalized isopaque brightness illumination demand model is then constructed by integrating individual visual physiological models and rhythmic stimulus curves, forming a dynamic light field control target, including: Construct a prediction model for illumination demand of non-uniform pedestrian flow; Based on the non-uniform pedestrian flow illumination demand prediction model, combined with real-time pedestrian flow distribution data, natural light background parameters and holographic pedestrian flow thermal imaging, the pattern of pedestrian flow aggregation and movement trend are predicted, and the basic illumination demand threshold of each area is obtained. Based on the basic illuminance requirement thresholds for each region, the personalized isoluminance illuminance requirement for each region is solved by fitting an individual visual physiological model and a rhythmic stimulus curve. The spatial illuminance distribution gradient features are extracted, and a dynamic light field regulation target that meets the requirements of visual comfort and rhythmic health is obtained through fitting.

5. The intelligent lighting control method based on pedestrian flow monitoring according to claim 4, characterized in that, Construct a prediction model for illuminance demand during non-uniform pedestrian traffic, including: Build a historical database of lighting areas; Based on the historical database of the lighting area, combined with the real-time pedestrian flow distribution density, aggregation pattern and body trajectory features in the holographic pedestrian flow thermal imaging, the spatiotemporal evolution features of the dynamic distribution of pedestrian flow are extracted by the spatiotemporal graph convolution algorithm. Based on the spatiotemporal evolution characteristics of the dynamic distribution of pedestrian flow, a natural light background illuminance compensation term is introduced to fit the regional illuminance demand curve under different pedestrian flow scenarios and establish a non-uniform pedestrian flow illuminance demand prediction model.

6. The intelligent lighting control method based on pedestrian flow monitoring according to claim 1, characterized in that, Generate a dynamic lighting control record map, including: Acquire data on actual illuminance, lighting operating parameters, pedestrian flow distribution, and visual comfort in the area during the control process; Based on the actual illuminance of the area, the operating parameters of the luminaires, the data on pedestrian flow distribution and visual comfort, a lighting control effect evaluation model is constructed to calculate the illuminance uniformity, uniform glare value, energy saving rate and visual satisfaction quality index. By associating lighting control parameters, luminaire operating status, and the aforementioned quality indicators along a time axis, a dynamic lighting control record map is generated, annotating control nodes, abnormal events, and comfort fluctuations.

7. The intelligent lighting control method based on pedestrian flow monitoring according to claim 1, characterized in that, Adjusting the output power, beam distribution angle, and projection orientation of luminaires in the lighting circuit, and calibrating output deviations online using a digital twin light decay prediction and compensation model, controls the instantaneous physiological equivalent illuminance during dynamic changes in pedestrian flow, including: Constructing a digital twin optical attenuation prediction and compensation model; Based on the aforementioned digital twin light decay prediction and compensation model, according to the multi-lamp collaborative dynamic control command, the output power of the lamps is adjusted by the dimming drive module, and the light distribution angle and projection direction are adjusted by the stepper motor to control the instantaneous physiological equivalent illuminance of each zone. When the illuminance deviation of a zone exceeds a preset threshold, a self-calibration compensation mechanism is triggered to correct the output parameters of the lamps based on the light decay prediction results, and at the same time, an abnormal warning signal is generated.

8. An intelligent lighting control system based on pedestrian flow monitoring, characterized in that, include: The acquisition module is used to acquire real-time pedestrian flow distribution data and raw ambient light data within the lighting area; The separation module is used to adaptively separate the natural light background, shadow disturbance and higher-order optical turbulence features from the raw ambient light data through a physically guided decoupling network, and to generate a holographic thermal image of the crowd carrying the trajectory and body posture using a spatiotemporal convolutional recurrent network. The prediction module is used to construct a non-uniform pedestrian flow illumination demand prediction model based on the real-time pedestrian flow distribution data, natural light background parameters and holographic pedestrian flow thermal imaging, predict the pedestrian flow aggregation pattern and movement trend, integrate individual visual physiological models and rhythmic stimulation curves to construct a personalized isopaque brightness illumination demand model, and form a dynamic light field control target. The generation module is used to construct a multi-lamp spot superposition illuminance calculation model based on the dynamic light field control target, combined with the inverse square law of illuminance and the cosine law of incident angle. The model is then solved by multi-parameter collaborative solution through virtual particle swarm optimization algorithm. After illuminance accuracy and glare value gradient verification, a multi-lamp collaborative dynamic control command is generated. The control module is used to adjust the output power, light distribution angle and projection orientation of the lamps in the lighting circuit according to the multi-lamp collaborative dynamic control command, calibrate the output deviation online through the digital twin light decay prediction compensation model, control the instantaneous physiological equivalent illuminance during the dynamic change of the pedestrian flow, and generate a dynamic lighting control record map.

9. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements an intelligent lighting control method based on pedestrian flow monitoring as described in any one of claims 1 to 7.

10. A readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements an intelligent lighting control method based on pedestrian flow monitoring as described in any one of claims 1 to 7.

Images