Multi-scene-oriented intelligent light adaptive control method and system

Abstract

The application provides a multi-scene-oriented intelligent light adaptive control method and system. The method comprises the following steps: identifying the current behavior state of a user and predicting the behavior state of the user based on visual sensor data, and outputting a light control behavior label; combining the behavior label with a scene type and an ambient light intensity to generate a target lighting intention vector; calculating the lighting weight of each lamp on the user according to the target lighting intention vector, combining a lighting control conflict situation, real-time lamp parameter and user space state data, and distributing brightness and color temperature parameters, combining the target dimming process duration to generate a lamp control parameter set; converting the lamp control parameter set into a dimming instruction sequence, performing time synchronization and delay compensation, and delivering the dimming instruction sequence to each lamp for execution. The application realizes accurate mapping from behavior intention to physical light, and has high universality, deployability and engineering controllability.

Description

Technical Field

[0001] This invention belongs to the field of intelligent adaptive lighting control, and particularly relates to intelligent adaptive lighting control methods and systems for multiple scenarios. Background Technology

[0002] With the rapid development of smart space applications, lighting systems are widely used in various scenarios such as homes, offices, education, and commerce. Their control requirements are no longer limited to simple on / off or timer mechanisms, but are evolving towards intelligence, adaptability, and behavior-driven approaches. Some existing systems use infrared or photosensitive sensors to detect the presence of people, but these only achieve basic on / off control and lack behavioral intent recognition and continuous response mechanisms. In recent years, some research has attempted to introduce interaction methods based on image recognition or voice control, but most can only react one-time based on the user's current state and cannot predict or infer the evolution of user behavior, resulting in lighting adjustments often lagging behind actual changes in demand.

[0003] Furthermore, the currently prevalent lighting zoning methods struggle to accurately respond to usage scenarios involving multiple users, multiple viewing angles, and multiple areas, especially in shared spaces such as meetings, offices, and homes. The lack of spatial matching mechanisms based on factors like user head orientation and focus leads to issues such as coarse light distribution, redundant energy consumption, and unbalanced user experience. On the other hand, existing systems generally employ group control or broadcast strategies at the lighting control layer, lacking the ability to model control trajectories at the individual lamp level. This makes it difficult to achieve flexible dimming, synchronized lamp rhythms, or response delay compensation during execution. Especially in practical deployments involving heterogeneous control protocols and significant differences in lamp response speeds, traditional control architectures struggle to guarantee the consistency and stability of lighting changes.

[0004] Therefore, a hierarchical adaptive control system is needed that is user behavior prediction-driven, combined with real-time spatial state modeling, and has the ability to perform dimming at the device level, so as to comprehensively improve the system's response quality and deployment adaptability in multi-scenario, multi-state, and multi-lighting environment. Summary of the Invention

[0005] The purpose of this invention is to propose an intelligent adaptive lighting control method and system for multiple scenarios, thereby solving the above-mentioned problems.

[0006] To achieve the above objectives, a first aspect of the present invention provides an intelligent adaptive lighting control method for multiple scenarios, the method comprising the following steps:

[0007] S1. Identify the user's current behavior state based on visual sensor data and predict the user's behavior state, and output a light control behavior label;

[0008] S2. Combine the behavior tag with scene type and ambient light intensity to generate a target lighting intent vector; wherein, the target lighting intent vector includes target ambient brightness, target color temperature and target dimming process duration;

[0009] S3. Based on the target lighting intent vector, combined with the lighting control conflict situation, real-time luminaire parameters and user space state data, calculate the lighting weight of each luminaire to the user, and assign brightness and color temperature parameters. Combined with the duration of the target dimming process, generate a set of luminaire control parameters.

[0010] S4. Convert the set of lighting control parameters into a dimming command sequence, perform time synchronization and delay compensation, and send it to each lighting fixture for execution.

[0011] Furthermore, the visual sensor data consists of a sequence of user action states within the current time period collected by an RGB-D visual perception device, as well as previously stored historical sample data of user behavior.

[0012] Further, S1 includes:

[0013] S101. Acquire visual sensor data and encode it into a behavior state vector to construct a behavior state sequence;

[0014] S102. Input the behavior state sequence into the pre-built LSTM model and output the predicted user behavior state at the next time step.

[0015] S103. The predicted user behavior state at the next moment is mapped into a light control behavior label through a behavior mapping function.

[0016] Furthermore, the scenario types include family bedrooms, meeting rooms, corridors, and shared office areas;

[0017] The lighting control behavior labels include reading mode, leave mode, enter mode, and communication mode.

[0018] Further, S2 includes:

[0019] S201. Construct a spatial scene recognition factor to make weighted fine-tuning of lighting output for the scene, corresponding to the correction coefficients of the three components of brightness, color temperature and change rhythm in turn. That is, the spatial scene recognition factor includes brightness correction coefficient, color temperature correction coefficient and change rhythm coefficient.

[0020] S202. Real-time acquisition of ambient light intensity from ambient light sensor, combined with standard ambient illuminance reference value, standard brightness template value corresponding to the light control behavior label, and brightness correction coefficient to calculate target ambient brightness;

[0021] S203. Calculate the target color temperature based on the template color temperature value corresponding to the light control behavior label, combined with the current time and color temperature correction coefficient.

[0022] S204. Calculate the duration of the target dimming process based on the default change duration corresponding to the light control behavior label and the softmax confidence level of the light control behavior label.

[0023] S205. The target lighting intent vector is obtained by combining the target ambient brightness, target color temperature and target dimming process duration.

[0024] Furthermore, the lighting control conflict scenarios include overlapping light focal points when multiple users are together, how the system can continuously identify the focal point of a user's field of vision when the user is stationary, and different luminaire distribution densities caused by spatial asymmetry.

[0025] The real-time lighting parameters include spatial location, emission angle, power limit, and color temperature range; the user spatial status data includes the user's head orientation angle and the user's position coordinates in three-dimensional space.

[0026] Further, S3 includes:

[0027] S301. Obtain the result of whether the user is inside the lamp cone, as a geometric occlusion judgment item, where 1 indicates that the user is inside the lamp cone; obtain the angle between the user's viewpoint and the lamp direction, as a viewpoint matching item; obtain whether other lamps cause occlusion between the lamp and the user, as a spatial coverage suppression item.

[0028] S302. Calculate the effective lighting weight of the luminaire for the user by combining the geometric occlusion judgment item, the viewing angle matching item, and the spatial coverage suppression item.

[0029] S303, aggregate the effective lighting weights of all luminaires to users into the total spatial contribution weight of the luminaires;

[0030] S304. Based on the total spatial contribution weight, the target ambient brightness and target color temperature are allocated to each lamp to form the final control parameters, including the brightness value finally allocated to the lamp, the control color temperature, and the duration of the target dimming process.

[0031] Furthermore, the spatial coverage suppression term is calculated using a Boolean indicator function.

[0032] Further, S4 includes:

[0033] S401. Obtain the set of lighting control parameters and, in conjunction with the sampled execution time, obtain the dimming control sequence of the lighting fixture;

[0034] S402. Align the dimming control sequence of all lamps with the time axis, and set a unified start time point according to the current system rhythm to generate the actual time when the lamp command should be sent.

[0035] S403. Generate a control sequence set for each lamp by combining the dimming control sequence and the actual time when the lamp command should be sent.

[0036] A second aspect of the invention provides an intelligent adaptive lighting control system for multiple scenarios, the system comprising:

[0037] The visual sensor data acquisition module is used to identify the user's current behavior state and predict the user's behavior state based on visual sensor data, and output the light control behavior label.

[0038] The data analysis module is used to combine the behavior tags with scene type and ambient light intensity to generate a target lighting intent vector; wherein, the target lighting intent vector includes target ambient brightness, target color temperature and target dimming process duration;

[0039] The adaptive lighting adjustment module is used to calculate the lighting weight of each lamp to the user based on the target lighting intention vector, combined with the lighting control conflict situation, real-time lamp parameters and user space state data, and to allocate brightness and color temperature parameters. Combined with the duration of the target dimming process, it generates a set of lamp control parameters.

[0040] The parameter execution module is used to convert the set of lighting control parameters into a dimming command sequence, perform time synchronization and delay compensation, and send it to each lighting fixture for execution.

[0041] The beneficial technical effects of the present invention are at least as follows:

[0042] This invention proposes an intelligent adaptive lighting control method and system for multiple scenarios. By constructing a four-level control architecture driven by user behavior tags and spanning time perception and spatial reasoning, it realizes the prediction of behavior trends, dynamic generation of lighting intentions, fine allocation of spatial lighting resources, and unified execution of multi-lamp control commands. It can systematically solve the technical bottlenecks of traditional lighting control systems such as response lag, generalized space, rigid control, and incompatible deployment. This method first predicts the user's upcoming behavioral state through structured time series modeling, outputting predictive lighting behavior labels. Second, the system maps these labels to structured lighting intentions, dynamically generating target brightness, color temperature, and change duration by combining information such as scene type, ambient light conditions, and user behavior confidence. Based on this, the spatial control module constructs a view-oriented lighting response function by combining luminaire deployment information and the user's real-time spatial distribution, modeling the illuminance contribution of each luminaire and generating final dimming parameters. Finally, the controller constructs executable segmented dimming control commands based on the luminaire target parameters, uniformly schedules and sends them to the heterogeneous luminaire system, ensuring that all luminaires complete the brightness and color temperature transitions consistently over time. This invention decouples the four tasks of "perception-reasoning-scheduling-execution" in its system structure, using behavior labels as the main axis of the entire process control, achieving precise mapping from behavioral intentions to physical illumination. It possesses extremely high versatility, deployability, and engineering controllability, making it a smart lighting control solution with broad adaptability and distributed implementability. Attached Figure Description

[0043] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0044] Figure 1 This is a flowchart of the intelligent adaptive lighting control method for multiple scenarios according to the present invention. Detailed Implementation

[0045] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0046] like Figure 1 As shown in the figure, the intelligent adaptive lighting control method for multiple scenarios provided by the embodiments of the present invention includes:

[0047] S1. Identify the user's current behavior state based on visual sensor data and predict the user's behavior state, and output a light control behavior label.

[0048] Specifically, the goal of this step is to identify the user's current activity state based on their temporal behavior data and predict the behavioral scenarios they will enter in the near future, thereby providing a preliminary decision-making basis for subsequent lighting strategy generation. This prediction strategy not only enhances the system's proactive response to user behavior but also provides a unified behavioral abstraction entry point for multi-scenario adaptation. The output is a lighting behavior label defined internally by the system, used to guide the next step of lighting intent generation.

[0049] This step receives a data stream from RGB-D visual sensing devices deployed within the lighting control area, including two parts: (1) the user action state sequence X within the current time period. t (2) Prior stored user behavior historical sample data H t .

[0050] RGB images are used for pose and motion classification, while depth images are used for spatial location and human motion intensity assessment. The acquisition frequency is set to 5 frames per second to reduce system load while ensuring real-time performance. The sensor is installed at a top-down angle on the edge of the ceiling, ensuring complete acquisition of the user's contours and spatial coordinates across the entire area. The system does not store images, only performs real-time recognition and extraction. The motion recognition component uses YOLOv5 for initial human detection, combined with skeleton extraction tools such as OpenPose or MediaPipe to capture the 3D coordinates of key points such as the user's head, shoulders, and hips in each frame, thereby constructing motion codes.

[0051] Furthermore, after each frame is processed, the system encodes it into a behavior state vector x. i Typically, the following fields are included: whether the person is sitting (judged by a shoulder-to-hip height difference of <20cm), whether they have just stood up (by a hip height change rate >60% over two consecutive frames), whether they are moving slowly (horizontal displacement <1 meter within 3 seconds), and whether they are standing still (stable position, no significant hand movements, duration >10 seconds). These states are mapped to a multi-dimensional vector, where each dimension is a binary or normalized amplitude feature, ultimately forming X. t =[x1,x2,...,x n ], used for sequence modeling.

[0052] H is also reserved in system memory t As a cache of historical behavior data, it contains the encoded historical behavior sequence of the target user for the past 30 minutes. If the identity of the current user is uncertain, other typical user behavior templates in the space are used. tIts main purpose is to help the model stabilize its boundaries and avoid misjudging extreme actions.

[0053] Behavioral trend prediction relies on a three-layer LSTM model, which possesses temporal memory and multi-behavior classification capabilities. The first layer is the embedding layer, with the input vector x... i The first layer is a fully connected transformation to a fixed-length internal representation; the second layer is a bidirectional LSTM structure, which captures the behavioral evolution trend in the forward direction and is used for pattern correction in the backward direction; the third layer is a classification layer that outputs the probability distribution of the behavioral state. During the training phase, the model uses pre-labeled real action samples for supervised training, and the model weights are frozen after deployment to avoid runtime learning errors.

[0054] For example: Suppose a user maintains a seated posture in an office area for more than 20 minutes. However, at the end of the first time window, the system detects an increase in hip height, an increase in the forward shoulder angle, and significant foot displacement. The system identifies this as a "getting up + leaning forward + moving forward" pattern for five consecutive frames. This is combined with historical H... t The model predicts that the user's next action is "leaving" more than 85% of the time, based on the fact that the user's past similar behavior is usually accompanied by a "leaving" label. Therefore, the output predicted label is: Then, the mapping function g maps it to the lighting behavior label L. t =Leave mode.

[0055]

[0056] Among them, f LSTM This represents a three-layer LSTM network structure, including an embedding layer, a bidirectional LSTM layer (with a hidden unit size of 64), and an output classification layer (softmax predicts 4 types of behavior labels). The input is the state sequence X within the current time window. t and historical behavior samples H t The output is the prediction result of the next row's state. X t H is a time-series state vector array encoded by the camera; t Cache historical behavior sequences; Predict the user behavior state for the next moment.

[0057]

[0058] Where g is the system-defined behavior mapping function, which maps the prediction results... Mapped to lighting control behavior labels; L t The behavior label specifically for lighting control will be used as input for the next step.

[0059] This step outputs a behavior label L. tThe label is one of the finite enumeration sets predefined by the system (such as reading mode, leaving mode, entering mode, interaction mode, etc.), and it will directly drive the generation of target lighting state in the next step.

[0060] S2. Combine the behavior tag with the scene type and ambient light intensity to generate a target lighting intent vector; wherein, the target lighting intent vector includes the target ambient brightness, the target color temperature and the duration of the target dimming process.

[0061] Specifically, the core task of this step is to process the behavior label L output in the previous step. t This is transformed into the target lighting state that should be achieved in the current space, that is, generating a lighting intent vector that is timely, behaviorally dependent, and environmentally adaptable. These correspond to the target brightness, target color temperature, and light transition time, respectively. This step not only plays a crucial role in connecting the preceding and following steps, but is also the central logic for realizing "feedforward intelligent lighting control based on user behavior trends" in this invention. All subsequent spatial lighting resource allocation and lamp dimming control are based on this vector.

[0062] The input is the behavior label L output from the previous step. t The label is a discrete enumeration variable defined within a set of behavioral categories in the system, such as {reading mode, leave mode, enter mode, communication mode, focus mode}. The behavioral label is generated by the LSTM sequence model in step one, mapping to the user's continuous posture trajectory and historical behavior probabilities; therefore, it is a control signal with a certain predictive capability.

[0063] Furthermore, considering the specific needs of this invention to be applicable to multiple scenarios (such as family bedrooms, conference rooms, corridors, and shared office areas), this step introduces a multi-dimensional environment adaptation mechanism based on behavior tags during the lighting strategy generation process, specifically reflected in three aspects:

[0064] First, considering that the lighting requirements corresponding to the same behavior label vary in different scenarios (for example, "leaving" in a corridor means turning off the lights, while in a meeting room it means slowly dimming the lights while maintaining background brightness), this invention introduces a spatial scene recognition factor C. s This is configured during the system deployment phase. s It is a set of weighted coefficients that participate in the target brightness calculation during runtime, used to fine-tune the lighting output for the specific scene. This factor is set according to the room function during the deployment phase, such as {0.9, 1.0, 0.8} for a bedroom and {1.1, 1.0, 1.2} for a conference room, corresponding to the correction coefficients for the three components of brightness, color temperature, and change rate, respectively.

[0065] Secondly, in order to adapt to real-time ambient brightness interference (such as enhanced natural light during the day and dark background at night), the system introduces ambient light sensor data collection value E. s The system dynamically corrects the illumination brightness by comparing it with a set reference value E0, ensuring that the total illuminance remains constant within the target comfort range. Unlike traditional additive correction methods, this invention introduces a nonlinear suppression term to prevent the system from mistakenly pushing the supplementary light into an overly bright area in high-brightness scenes, thereby causing user discomfort.

[0066]

[0067] Among them, B * B represents the target ambient brightness. L For behavior label L t The corresponding standard brightness template value (e.g., 400 for reading mode); E is the brightness correction factor for the current space type. s The ambient light intensity is derived from a light sensor placed on the wall or near the light fixture, with a sampling period of 20 seconds; E0 is the standard ambient light reference value (e.g., 300); α is the supplementary light sensitivity coefficient (default 0.6); ∈ is a small constant to avoid the denominator being zero.

[0068] The tanh term introduced in this formula has good nonlinear convergence characteristics, making E s The higher the value, the faster the brightness correction decreases, effectively suppressing excessive fill light. This is a structural improvement of traditional linear dimming logic in scenarios with natural light intensity interference.

[0069] Next is the color temperature parameter. The generation of color temperature is crucial. Color temperature is typically closely related to behavioral labels; for example, "Focus Mode" tends to be cooler (e.g., 5500K), while "Communication Mode" tends to be warmer (e.g., 3000K). However, under specific time periods or user preferences, a certain degree of adjustability is still necessary. This invention introduces a time-adjustment function φ(t), which takes the current system time t as input to simulate the daily rhythm, for example, cooler in the morning, neutral at noon, and warmer in the evening. This function is constructed using a piecewise cosine function and superimposed on the base template color temperature.

[0070]

[0071] in, For the final target color temperature; T L For the label L t Mapped template color temperature; β is the color temperature correction factor corresponding to the space type; t is the current time (minutes), with a cycle of 1440 minutes per day; β is the color temperature adjustment range (default 300K), used to adapt to the gradual change in the circadian rhythm.

[0072] This color temperature adjustment strategy embodies a design concept that integrates time perception and behavior perception, improves the synchronization between lighting and human physiological rhythms, and is a technical route for adaptive control of circadian rhythm lighting.

[0073] Finally, considering the potential impact of uncertainties in behavior tag recognition on the rhythm of light changes, this invention introduces a confidence level P consistent with step one. L Control the dimming time D * The generation of . Consistent with the aforementioned formula:

[0074] D * =D L ·(1+γ·(1-P L (5)

[0075] Among them, D * D represents the duration of the target dimming process. L For behavior label L t The corresponding default change duration; P L γ represents the softmax confidence level for the model to predict the label of this behavior; γ is the adjustment sensitivity coefficient.

[0076] The output of this step is the target lighting intent vector. It is a structured control signal that includes brightness, color temperature, and adjustment rhythm. It will directly enter the next step of the spatial lighting distribution module to determine the power and color temperature adjustment range of each lamp.

[0077] S3. Based on the target lighting intent vector, combined with the lighting control conflict situation, real-time luminaire parameters and user space state data, calculate the lighting weight of each luminaire to the user, and assign brightness and color temperature parameters. Combined with the duration of the target dimming process, generate a set of luminaire control parameters.

[0078] Specifically, this step follows the target lighting intent vector generated in the previous step. The aim is to rationally allocate lighting fixtures in a specific space according to this intention, so that the final output brightness B j ′、Color Temperature T j ′ and duration D * In terms of spatial distribution, it closely follows the user's location and area of ​​interest, achieving a high degree of coupling and scheduling among space, behavior, and lighting.

[0079] Furthermore, this step uses T * As input, combined with the lighting parameters configured during the deployment phase (such as spatial location (x...) j ,y j ,z j ), light emission angle φ j Power limit and color temperature range ), and real-time acquired user space status data U t ={(p k ,θ k )},θ k Let p be the head orientation angle of the k-th user. k To determine the position coordinates of the k-th user in 3D space, weight modeling and lighting resource allocation are performed. User space state data comes from a top RGB-D camera, and the system uses a skeleton point + continuous frame difference method to identify the user coordinates p. k and head orientation θ k And maintain independent modeling in multi-user scenarios.

[0080] To address the "multi-scenario" requirements of this invention (such as conference rooms, corridors, and family reading areas), the modeling in this step must be compatible with the following three common lighting control conflict scenarios:

[0081] Overlapping lighting focus when multiple users are present (e.g., in a conference room where speakers and recorders are present simultaneously);

[0082] How does the system continuously identify the user's field of view focus when the user is stationary?

[0083] Asymmetrical spatial structures (such as L-shaped corridors) result in varying density of lighting fixtures.

[0084] To this end, this invention incorporates the following key mechanisms into the weighting modeling formula and designs a structured spatial lighting response function A. jk Used to indicate lamps j For user u k The actual lighting contribution. The function consists of three terms: cone projection term, view matching term, and spatial coverage suppression term, which respectively capture the three dimensions of "whether it can be illuminated", "whether the illumination is reasonable", and "whether it is blocked by other lights".

[0085]

[0086] Among them, A jk Indicates lighting fixtures j For user u k Effective lighting weight; δ jk For geometric occlusion determination, 1 indicates that the user is inside the luminaire cone, calculated by the projection function; Indicates the angle between the user's viewing angle and the direction of the light fixture. λ1 is the main illumination vector of the luminaire; λ1 is the view offset penalty coefficient, default 0.7, the larger the angle, the faster the attenuation; ρ is the occlusion suppression coefficient (default 0.5); 1 overlap(·) A Boolean indicator function to determine whether other lights are obstructing the connection between the light and the user.

[0087] The most innovative aspect of this expression is the third suppression mechanism: in environments with dense lighting (such as conference room lighting grids), adjacent lights may overlap in illumination of the same area. Without suppression, this can easily lead to overexposure or light "competition." This invention introduces an occlusion judgment function, which dynamically suppresses occlusion by judging the intersection of the light cones in space and their positional relationship relative to the user, thereby enhancing the spatial balance of the light field distribution.

[0088] Complete A jk After calculation, the system aggregates the weights between all users and lighting fixtures into a contribution value w for the lighting fixture layer. j ,Right now:

[0089]

[0090] Among them, w j For lighting fixtures j Total spatial contribution weight; ω k For user u k The behavior weight (default mean value is 1) can be determined by the tag priority, such as "reading" being higher than "leaving"; A jk Let be the spatial lighting response function defined in the above equation.

[0091] Next, the system will use the global target brightness B from the previous step. * and color temperature The parameters are assigned to each luminaire to form the final control parameter B. j ′ and T j The specific method involves proportional allocation and applying hard-constraint clipping, calculated as follows:

[0092]

[0093] Among them, B j ′ for final allocation to lighting fixtures l j The brightness value; T j ' represents its color temperature, and the clip function is used for boundary protection;

[0094] The duration of change shared by all luminaires D * It can be directly inherited from step two without any changes.

[0095] The actual operation flow of this strategy is as follows: In each frame update, the system recalculates A based on the latest user pose state. jk and w j However, only in behavior label L t Switching or lighting parameter T * A new round of allocation and dimming commands is only triggered when changes exceed a certain threshold. This ensures both real-time control and avoids excessive system fluctuations.

[0096] The final output is a set of lighting control parameters. This output will be used directly as input for the next step of control execution.

[0097] S4. Convert the set of lighting control parameters into a dimming command sequence, perform time synchronization and delay compensation, and send it to each lighting fixture for execution.

[0098] Specifically, this step is responsible for processing the control parameters of each lamp generated in step three. This process transforms the signals into control commands that can be received and executed by the actual lighting equipment, and ensures that all lighting fixtures complete the lighting transition in a coordinated manner through a unified scheduling mechanism. This step is the final output node of the entire system's transition from "perception-reasoning-distribution" to "physical control," and it determines whether the system's control effect has spatial consistency, smooth response, and controllability and feasibility.

[0099] The control module needs to construct a dimming command frame for each luminaire based on the above three parameters. This frame should include the brightness value, color temperature value, and execution time, while also considering control accuracy and luminaire response lag. This system adopts a linear interpolation-type command encoding mechanism, which constructs a dimming sequence, quantizes it into sampling points, and then generates a command packet to be transmitted to each luminaire.

[0100] The control sequence is generated using the following expression:

[0101]

[0102] Among them, C j Indicates lighting fixtures j The dimming control sequence; This indicates the execution time of the i-th sampling step; n is the control resolution, which divides the dimming process into n uniform time periods (e.g., n=10); each step transitions uniformly and linearly to the target brightness B. j ′ and color temperature T j The structural innovation of this expression lies in using the standard structure of the triplet set C. j Expressing the dimming process of each luminaire does not require defining complex function trajectories and avoids strong assumptions about the luminaire response model, making the instruction structure compatible with various control interfaces (DALI, Zigbee, etc.). In addition, by selecting an appropriate n value (default 10-20), the gradual dimming process can be fitted without increasing communication pressure, effectively avoiding "jumpy transitions".

[0103] Furthermore, in actual implementation, the system is equipped with a scheduling controller, whose core task is to manage the C-level control of all lighting fixtures. j Timeline alignment is performed, and a unified start time point t0 is set according to the current system rhythm. To improve synchronization accuracy, a fine-tuning phase delay compensation term Δt is introduced into the system. jBased on the historical response characteristics of each lamp, if a lamp's response generally delays by 30ms, its command will be issued in advance:

[0104]

[0105] in, Δt represents the actual time when the i-th step instruction for lamp j should be sent; j The offset compensation value is measured by the system during deployment.

[0106] This compensation mechanism can effectively eliminate the "spatial inconsistency in light perception" caused by inconsistent control paths and different protocol delays, which is one of the technical highlights of this invention in the "unified dimming synchronization control" stage.

[0107] For example, in a meeting room scenario, when the user behavior is labeled "speaking mode," the target lighting intent is T. * = (300, 4000, 3), the system allocates this brightness and color temperature to the three lamps in the main presentation area, namely B j =280,260,240, T j = 4000 (consistent). The controller divides each lamp into 12 dimming sequences, and finally sends 12 command sets containing triplets to each lamp, with dimming duration D. * =3s, resolution is 250ms. If lamp 2 has an average lag of 50ms in the historical response, the system sends its each step instruction 50ms earlier, i.e. Δt2 = 0.05.

[0108] The final output is a control sequence set C = {C1,...,C} for each lamp. N The system then distributes these signals to various device ports after unified scheduling. All lights will be located in [0, D]. * Within a short time, the system smoothly transitions between brightness and color temperature, completing the closed-loop dimming behavior.

[0109] This invention also provides an intelligent adaptive lighting control system for multiple scenarios, the system comprising:

[0110] The visual sensor data acquisition module is used to identify the user's current behavior state and predict the user's behavior state based on visual sensor data, and output the light control behavior label.

[0111] The data analysis module is used to combine the behavior tags with scene type and ambient light intensity to generate a target lighting intent vector; wherein, the target lighting intent vector includes target ambient brightness, target color temperature and target dimming process duration;

[0112] The adaptive lighting adjustment module is used to calculate the lighting weight of each lamp to the user based on the target lighting intention vector, combined with the lighting control conflict situation, real-time lamp parameters and user space state data, and to allocate brightness and color temperature parameters. Combined with the duration of the target dimming process, it generates a set of lamp control parameters.

[0113] The parameter execution module is used to convert the set of lighting control parameters into a dimming command sequence, perform time synchronization and delay compensation, and send it to each lighting fixture for execution.

[0114] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0115] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of apparatuses or units may be electrical, mechanical, or other forms.

[0116] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0117] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A smart lighting adaptive control method for multiple scenarios, characterized in that, The method includes the following steps: S1. Identify the user's current behavior state based on visual sensor data and predict the user's behavior state, and output a light control behavior label; S2. Combine the behavior tag with scene type and ambient light intensity to generate a target lighting intent vector; wherein, the target lighting intent vector includes target ambient brightness, target color temperature and target dimming process duration; S3. Based on the target lighting intent vector, combined with the lighting control conflict situation, real-time luminaire parameters and user space state data, calculate the lighting weight of each luminaire to the user, and assign brightness and color temperature parameters. Combined with the duration of the target dimming process, generate a set of luminaire control parameters. S4. Convert the set of lighting control parameters into a dimming command sequence, perform time synchronization and delay compensation, and send it to each lighting fixture for execution; Wherein, S2 includes: S201. Construct a spatial scene recognition factor to make weighted fine-tuning of lighting output for the scene, corresponding to the correction coefficients of the three components of brightness, color temperature and change rhythm in turn. That is, the spatial scene recognition factor includes brightness correction coefficient, color temperature correction coefficient and change rhythm coefficient. S202. Real-time acquisition of ambient light intensity from ambient light sensor, combined with standard ambient illuminance reference value, standard brightness template value corresponding to the light control behavior label, and brightness correction coefficient to calculate target ambient brightness; S203. Calculate the target color temperature based on the template color temperature value corresponding to the light control behavior label, combined with the current time and color temperature correction coefficient. S204. Calculate the duration of the target dimming process based on the default change duration corresponding to the light control behavior label and the softmax confidence level of the light control behavior label. S205. Combine the target ambient brightness, target color temperature, and target dimming process duration to obtain the target lighting intent vector; The S3 includes: S301. Obtain the result of whether the user is inside the lamp cone, as a geometric occlusion judgment item, where 1 indicates that the user is inside the lamp cone; obtain the angle between the user's viewpoint and the lamp direction, as a viewpoint matching item; obtain whether other lamps cause occlusion between the lamp and the user, as a spatial coverage suppression item. S302. Calculate the effective lighting weight of the luminaire for the user by combining the geometric occlusion judgment item, the viewing angle matching item, and the spatial coverage suppression item. S303, aggregate the effective lighting weights of all luminaires to users into the total spatial contribution weight of the luminaires; S304. Based on the total spatial contribution weight, the target ambient brightness and target color temperature are allocated to each lamp to form the final control parameters, including the brightness value finally allocated to the lamp, the control color temperature, and the duration of the target dimming process. The S4 includes: S401. Obtain the set of lighting control parameters and, in conjunction with the sampled execution time, obtain the dimming control sequence of the lighting fixture; S402. Align the dimming control sequence of all lamps with the time axis, and set a unified start time point according to the current system rhythm to generate the actual time when the lamp command should be sent. S403. Generate a control sequence set for each lamp by combining the dimming control sequence and the actual time when the lamp command should be sent.

2. The intelligent adaptive lighting control method for multiple scenarios according to claim 1, characterized in that, The visual sensor data consists of user action state sequences within the current time period collected by an RGB-D visual perception device, as well as previously stored historical sample data of user behavior.

3. The intelligent adaptive lighting control method for multiple scenarios according to claim 1, characterized in that, S1 includes: S101. Acquire visual sensor data and encode it into a behavior state vector to construct a behavior state sequence; S102. Input the behavior state sequence into the pre-built LSTM model and output the predicted user behavior state at the next time step. S103. The predicted user behavior state at the next moment is mapped into a light control behavior label through a behavior mapping function.

4. The intelligent adaptive lighting control method for multiple scenarios according to claim 1, characterized in that, The scenario types include family bedrooms, meeting rooms, corridors, and shared office areas; The lighting control behavior labels include reading mode, leave mode, enter mode, and communication mode.

5. The intelligent adaptive lighting control method for multiple scenarios according to claim 1, characterized in that, The lighting control conflict scenarios include overlapping light focus when multiple users are together, how the system can continuously identify the field of view focus when a user is stationary, and different luminaire distribution densities caused by spatial asymmetry. The real-time lighting parameters include spatial location, emission angle, power limit, and color temperature range; the user spatial status data includes the user's head orientation angle and the user's position coordinates in three-dimensional space.

6. The intelligent adaptive lighting control method for multiple scenarios according to claim 1, characterized in that, The spatial coverage suppression term is calculated using a Boolean indicator function.

7. A system for implementing the intelligent adaptive lighting control method for multiple scenarios as described in claim 1, characterized in that, The system includes: The visual sensor data acquisition module is used to identify the user's current behavior state and predict the user's behavior state based on visual sensor data, and output the light control behavior label. The data analysis module is used to combine the behavior tags with scene type and ambient light intensity to generate a target lighting intent vector; wherein, the target lighting intent vector includes target ambient brightness, target color temperature and target dimming process duration; The adaptive lighting adjustment module is used to calculate the lighting weight of each lamp to the user based on the target lighting intention vector, combined with the lighting control conflict situation, real-time lamp parameters and user space state data, and to allocate brightness and color temperature parameters. Combined with the duration of the target dimming process, it generates a set of lamp control parameters. The parameter execution module is used to convert the set of lighting control parameters into a dimming command sequence, perform time synchronization and delay compensation, and send it to each lighting fixture for execution.

Images