Smart city lighting advertisement management method and system

Abstract

The invention provides a smart city lighting advertisement management method and system, and the method comprises the steps: collecting the state data of a lighting device, an advertisement resource and an environment scene through a multi-source sensing network, and carrying out the preprocessing and fusion through an edge calculation node, and obtaining a standardized data flow; based on the standardized data stream, the lighting equipment and the advertisement resources are subjected to bucket separation by adopting locality sensitive hashing, optimization is carried out through iterative local search, and a resource grouping result is obtained; constructing a random regular graph model, optimizing a graph structure through a Markov chain Monte Carlo algorithm, extracting a mapping relationship between the lighting equipment and the advertisement resources, and generating a resource allocation strategy; generating a cooperative control strategy through a multi-objective optimization algorithm on the basis of a resource configuration strategy in combination with the user behavior data and the state data of the environment scene; and based on the cooperative control strategy, controlling the lighting equipment and the advertisement resources and collecting feedback data. According to the invention, intelligent collaborative management of urban lighting and advertisement resources is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of smart city management, and in particular to a smart city lighting advertising management method and system. Background Art

[0002] Smart city development is a key trend in current urban development. Urban lighting systems and outdoor advertising, as integral components of a city's image, are crucial for enhancing its quality and economic value. Smart city lighting and advertising management involve core technologies such as large-scale distributed device control, optimized resource allocation, and diversified content management.

[0003] Traditional urban lighting systems are typically managed through centralized control or pre-set schedules, such as time-zoned lighting control systems or simple light-sensing technology for brightness adjustment. Traditional outdoor advertising management relies primarily on manual inspections and regular maintenance, with updates to ad content reliant on manual intervention, lacking intelligent and automated management.

[0004] Currently, more advanced smart city lighting systems are attempting to integrate IoT technology, big data analytics, and lighting systems. By establishing a centralized management platform, they enable remote monitoring and control of urban lighting equipment. These systems typically employ a hierarchical data processing architecture, performing simple cluster analysis on collected environmental data and device status, and then making control decisions based on pre-set rules. Regarding advertising resource management, these systems employ fixed resource allocation strategies, making it impossible to dynamically optimize advertising content based on regional characteristics and traffic flow.

[0005] However, existing technologies suffer from inefficient data processing and insufficiently optimized resource allocation. Traditional clustering algorithms, in particular, face high computational complexity when processing large-scale urban lighting equipment data, making it difficult to meet real-time requirements. Furthermore, the lack of an effective resource allocation model leads to a conflict between advertising display effectiveness and urban lighting needs, preventing the coordinated optimization of the two. Furthermore, existing systems typically treat lighting and advertising as independent systems, lacking a unified resource scheduling mechanism, resulting in energy waste and visual pollution.

[0006] Therefore, how to build an efficient smart city lighting advertising management method and system to achieve coordinated optimization of urban lighting equipment and advertising resources is a technical problem that urgently needs to be solved. Summary of the Invention

[0007] The purpose of the present invention is to provide a smart city lighting advertising management method and system, aiming to solve the technical problems of low data processing efficiency, insufficient resource allocation optimization and lack of a unified resource scheduling mechanism in the existing technology, and realize the coordinated optimization management of urban lighting equipment and advertising resources.

[0008] To achieve the above objectives, the present invention provides the following technical solutions:

[0009] A smart city lighting advertising management method, comprising:

[0010] The state data of urban lighting equipment, advertising resources and environmental scenes are collected through a multi-source perception network, and the state data are pre-processed and integrated using edge computing nodes to obtain a standardized data stream;

[0011] Based on the standardized data stream, locality-sensitive hashing is used to bucket the lighting equipment and advertising resources, and optimization is performed through iterative local search to obtain resource grouping results;

[0012] Based on the resource grouping results, a random regular graph model is constructed, the graph structure is optimized using a Markov chain Monte Carlo algorithm, the mapping relationship between lighting equipment and advertising resources is extracted, and a resource allocation strategy is generated;

[0013] Based on the resource allocation strategy, combined with user behavior data and status data of the environmental scenario, a collaborative control strategy is generated through a multi-objective optimization algorithm;

[0014] Based on the collaborative control strategy, lighting equipment and advertising resources are controlled; at the same time, the operating status data and display effect data of both the lighting equipment and the advertising resources are collected, and the operating status data and the display effect data are fed back to the multi-source perception network as optimization input to realize smart city lighting advertising management.

[0015] Optionally, the state data of urban lighting equipment, advertising resources, and environmental scenes are collected through a multi-source perception network, and the state data are preprocessed and integrated using edge computing nodes to obtain a standardized data stream, including:

[0016] Collecting status data of the lighting device, the advertising resources, and the environmental scene, and transmitting the status data to the edge computing node through a preset data acquisition protocol; wherein the status data includes power parameters, brightness parameters, size parameters of the advertising screen, resolution parameters, and working data of the environmental sensor;

[0017] Based on the state data, data preprocessing and spatiotemporal synchronization are performed at the edge computing node to generate a data stream;

[0018] The data stream is cleaned and features are extracted to obtain the standardized data stream.

[0019] Optionally, the method of bucketing lighting equipment and advertising resources using locality-sensitive hashing and optimizing through iterative local search to obtain resource grouping results includes:

[0020] Based on the standardized data stream, construct a feature vector using dimension reduction and feature extraction techniques;

[0021] Based on the feature vector, initially bucketing the lighting devices and advertising resources using a locality-sensitive hash function;

[0022] Iterative local search optimization is performed on the initial bucketing result to obtain the resource grouping result.

[0023] Optionally, performing iterative local search optimization on the initial bucketing result to obtain the resource grouping result includes:

[0024] Using each bucket of the initial bucketing result as an initial clustering cluster, a clustering objective function is constructed; wherein the clustering objective function includes intra-cluster similarity, spatial continuity, energy load balance, and advertising coverage efficiency;

[0025] Based on the clustering objective function, performing local search on the data points within each of the initial clusters, and updating the clustering objective function by using an incremental calculation technique;

[0026] When it is determined that the clustering objective function converges, the resource grouping result is obtained.

[0027] Optionally, the constructing of a random regular graph model includes:

[0028] Based on the resource grouping result and in combination with the state data of the environmental scene, a heterogeneous graph structure of lighting device nodes, advertising resource nodes and environmental scene nodes is constructed;

[0029] Degree distribution regularization and edge weight randomization are performed on the heterogeneous graph structure to obtain the random regular graph model.

[0030] Optionally, the optimizing the graph structure by using a Markov Chain Monte Carlo algorithm includes:

[0031] Based on the random regular graph model, an energy function including resource allocation balance and control system responsiveness is constructed;

[0032] Iteratively optimizing the random regular graph model using a block parallel Markov chain Monte Carlo algorithm to obtain an optimized graph structure;

[0033] Based on the optimized graph structure and in combination with the energy function, the mapping relationship between device nodes and resource nodes is extracted to generate the resource configuration strategy.

[0034] Optionally, generating a collaborative control strategy by a multi-objective optimization algorithm includes:

[0035] Based on the resource allocation strategy, a multi-objective function including energy efficiency and visual comfort is constructed;

[0036] Optimizing and solving the multi-objective function to obtain a Pareto optimal solution for resource allocation;

[0037] Based on the Pareto optimal solution, the collaborative control strategy is generated.

[0038] Optionally, the state data combining the user behavior data and the environmental scenario includes:

[0039] Collecting data on ambient light intensity, weather conditions, and crowd density to obtain status data of the environmental scene;

[0040] Obtain user behavior data by obtaining user dwell time and attention heat map based on video analysis technology;

[0041] Performing spatiotemporal correlation analysis on the state data of the environmental scene and the user behavior data to obtain a scene feature vector.

[0042] Optionally, the controlling lighting equipment and advertising resources based on the collaborative control strategy includes:

[0043] Converting the collaborative control strategy into brightness parameters of the lighting device and display parameters of the advertising resources;

[0044] Adjusting the brightness of the lighting device and the display content of the advertising resources through an Internet of Things control system;

[0045] Collect energy consumption data of the lighting equipment and display effect data of the advertising resources.

[0046] The present invention also provides a smart city lighting advertising management device, comprising:

[0047] A data acquisition module is used to collect status data of urban lighting equipment, advertising resources, and environmental scenes through a multi-source perception network, and use edge computing nodes to preprocess and fuse the status data to obtain a standardized data stream;

[0048] A resource grouping module, configured to bucket lighting equipment and advertising resources using locality-sensitive hashing based on the standardized data stream, and optimize through iterative local search to obtain resource grouping results;

[0049] A resource mapping module is used to construct a random regular graph model based on the resource grouping results, optimize the graph structure using a Markov chain Monte Carlo algorithm, extract the mapping relationship between lighting equipment and advertising resources, and generate a resource allocation strategy;

[0050] A control decision module, configured to generate a collaborative control strategy based on the resource allocation strategy, combined with user behavior data and status data of the environmental scenario, through a multi-objective optimization algorithm;

[0051] An execution feedback module is used to control lighting equipment and advertising resources based on the collaborative control strategy; at the same time, the operating status data and display effect data of both the lighting equipment and the advertising resources are collected, and the operating status data and the display effect data are fed back to the multi-source perception network as optimization input to realize smart city lighting advertising management.

[0052] The beneficial effects of the present invention are:

[0053] 1. By combining an improved locality-sensitive hashing (LSH) with iterative local search, we achieve near-linear time clustering with O(nlog n) time complexity, solving the real-time problem of large-scale urban lighting equipment and advertising resource analysis.

[0054] 2. A random regular graph model that transcends uniqueness was designed, and a Markov chain Monte Carlo (MCMC) algorithm with fast mixing characteristics was introduced to achieve the optimal mapping relationship between lighting equipment and advertising resources.

[0055] 3. A four-layer architecture system based on the Internet of Things perception layer, edge computing layer, cloud platform layer and application service layer was constructed to realize the collection, processing and integration of multi-source heterogeneous data.

[0056] 4. A multi-objective optimization decision-making system integrating energy efficiency, visual comfort, advertising display effect and environmental impact was developed to achieve coordinated control of lighting and advertising.

[0057] An adaptive execution and feedback mechanism was designed, and a closed-loop optimization was formed through a real-time performance evaluation model, which improved the system's adaptability and resource utilization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0059] Figure 1 A flowchart of a smart city lighting advertising management method provided by an embodiment of the present invention;

[0060] Figure 2 A flowchart of a near-linear time clustering algorithm provided by an embodiment of the present invention;

[0061] Figure 3 A schematic diagram of the structure of a collaborative optimization decision-making system provided by an embodiment of the present invention;

[0062] Figure 4 This is a module composition diagram of the smart city lighting advertising management system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0063] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0064] like Figure 1 As shown, the smart city lighting advertising management method provided by the embodiment of the present invention includes:

[0065] Step S1: collecting status data of urban lighting equipment, advertising resources, and environmental scenes through a multi-source perception network, and using edge computing nodes to preprocess and fuse the status data to obtain a standardized data stream;

[0066] Data is collected through a multi-source sensing network deployed throughout the city. This sensing network includes environmental sensors, lighting equipment sensors, pedestrian flow detectors, and advertising screen status monitors. Environmental sensors primarily collect environmental parameters such as light intensity, temperature, humidity, and air quality; lighting equipment sensors collect parameters such as lamp power, brightness, color temperature, and operating hours; pedestrian flow detectors collect data such as pedestrian count, flow direction, and density; and advertising screen status monitors collect information such as the screen's resolution, brightness, and content switching frequency. This multi-source, heterogeneous data is transmitted via wired or wireless networks to edge computing nodes distributed throughout the city.

[0067] After receiving heterogeneous data from multiple sources, edge computing nodes first clean the data to remove outliers, duplicates, and missing values. They then perform data normalization, converting data of varying dimensions into a unified numerical range. Next, they synchronize data from different devices in time and space to ensure consistency across both temporal and spatial dimensions. Finally, edge computing nodes perform preliminary fusion and feature extraction on the processed data to generate a standardized data stream. This edge computing approach not only reduces data transmission volume and alleviates the computing burden on the cloud platform, but also improves the system's response to local events.

[0068] Standardized data streams are transmitted from edge computing nodes to the cloud platform's data center for further in-depth processing and storage. The cloud platform leverages big data technology to analyze historical data and uncover potential patterns and trends. Simultaneously, the cloud platform builds a knowledge graph that structuredly represents entities such as devices, resources, environments, and users, as well as their relationships, providing knowledge support for subsequent intelligent decision-making. This multi-layered data processing architecture can effectively handle heterogeneous data at a city-scale, providing a solid data foundation for intelligent lighting advertising management.

[0069] Step S2: Based on the standardized data stream, the lighting equipment and advertising resources are bucketed using locality-sensitive hashing, and optimized through iterative local search to obtain resource grouping results;

[0070] During the specific implementation process, the system first constructs feature vectors for the standardized data stream. Through dimensionality reduction and feature extraction technology, high-dimensional heterogeneous data is converted into low-dimensional feature vectors, which not only reduces the computational complexity but also retains the key information of the data. Then, based on these feature vectors, the system uses an improved local sensitive hashing (LSH) algorithm to preliminarily bucket the lighting equipment and advertising resources. The core idea of LSH is to design a set of hash functions so that similar objects in the feature space have a higher probability of being mapped to the same "bucket". The present invention improves the traditional LSH algorithm and designs an adaptive hash function family that can better adapt to the uneven distribution characteristics of urban spatial data.

[0071] After completing the initial bucketing, the system uses an iterative local search algorithm to optimize the bucketing results. The core of this iterative local search is a "perturb-improve" cycle: During the improvement phase, each data point is considered for movement from its current cluster to a neighboring cluster. If the move improves the objective function value, the move is accepted. During the perturbation phase, a certain percentage of points are randomly selected and reallocated to escape the local optimum. To achieve near-linear time complexity, the present invention utilizes optimization techniques such as a proximity graph to restrict the search range, incremental computation techniques, and an early stopping strategy. This approach enables the system to efficiently generate resource grouping results, providing a foundation for the subsequent construction of resource mapping relationships.

[0072] Step S3: Based on the resource grouping results, a random regular graph model is constructed, the graph structure is optimized using a Markov Chain Monte Carlo algorithm, a mapping relationship between lighting equipment and advertising resources is extracted, and a resource allocation strategy is generated;

[0073] Based on the resource grouping results, a multi-layer heterogeneous graph structure is constructed. This graph consists of lighting device nodes, advertising resource nodes, and environmental scene nodes. The edges between nodes represent physical proximity, functional complementarity, and visual influence. The graph structure is then subjected to random regularization, including degree distribution regularization, edge weight randomization, and global structural balancing. This random regularization method, which goes beyond uniqueness, enables the model to better reflect the uncertainty and variability of real-world environments.

[0074] Next, the system applies a Markov Chain Monte Carlo algorithm with fast mixing properties to optimize the graph structure. The algorithm first defines an energy function, comprehensively considering multiple factors such as resource allocation balance, control system responsiveness, visual coordination, and energy efficiency. It then employs techniques such as a block-parallel MCMC strategy, an adaptive proposal distribution, and a temperature scheduling mechanism to improve the algorithm's convergence speed and optimization effectiveness. The algorithm's fast mixing properties are primarily achieved through a non-local proposal mechanism, which enables MCMC to complete "long-distance jumps" in state space in a single step, significantly improving mixing speed.

[0075] After optimization, the system extracts the mapping relationship between lighting equipment and advertising resources from a stable graph structure. This process utilizes techniques such as community detection, bipartite graph matching, and hyperedge mapping, enabling it to handle complex one-to-one, one-to-many, and many-to-many mapping relationships. Furthermore, the system incorporates a time-window-based mapping relationship update mechanism and an event-triggered update mechanism to ensure that resource allocation strategies can dynamically adapt to environmental changes.

[0076] Step S4, generating a collaborative control strategy through a multi-objective optimization algorithm based on the resource allocation strategy, combined with user behavior data and status data of the environmental scenario;

[0077] A multi-objective function was constructed, encompassing dimensions such as energy efficiency, visual comfort, advertising display effectiveness, and environmental impact. The energy efficiency objective evaluated the ratio of system energy consumption to service effectiveness; the visual comfort objective, based on the human visual perception model, assessed the combined visual effects of lighting and advertising; the advertising display effectiveness objective evaluated the visibility and information transmission efficiency of advertising content; and the environmental impact objective assessed the system's impact on the ecological environment.

[0078] The system also collects real-time environmental data, such as light intensity, weather conditions, and crowd density, through a network of environmental sensors. Using video analysis, the system also captures user behavior data, including dwell time, gaze direction, and attention distribution in front of the advertising screen. This environmental and user data undergoes spatiotemporal correlation analysis and is converted into scene feature vectors, which serve as important input for control decisions.

[0079] Based on multi-objective functions and scenario feature vectors, the system applies an improved multi-objective evolutionary algorithm (MOEA) to search for a Pareto-optimal set of solutions. From these non-dominated solutions, the system selects the most suitable one based on the current environmental state and management strategy and transforms it into a specific collaborative control strategy. This approach not only considers the efficiency and effectiveness of resource utilization but also dynamically adjusts control strategies based on real-time conditions, enabling intelligent and personalized management of urban lighting advertising.

[0080] Step S5: Based on the collaborative control strategy, the lighting equipment and advertising resources are controlled; at the same time, the operating status data and display effect data of both the lighting equipment and the advertising resources are collected, and the operating status data and the display effect data are fed back to the multi-source perception network as optimization inputs to realize smart city lighting advertising management.

[0081] The instruction parsing engine converts high-level control strategies into device-level control commands. The instruction parsing engine, comprising three modules: strategy parsing, parameter mapping, and instruction generation, accurately translates abstract control strategies into a set of device-executable instructions. These instructions are reliably distributed to each terminal device via a distributed message queue system.

[0082] Upon receiving control commands, the intelligent LED lighting system adjusts the brightness, color temperature, and illumination angle of the lamps accordingly, achieving precise control of the lighting environment. The digital advertising screen's content management system adjusts parameters such as the display plan, brightness, and animation effects of the advertising content. A collaborative mechanism has been established between the two systems to ensure that the lighting effects and advertising display work together to create an optimal visual experience.

[0083] At the same time, the system's built-in monitoring module continuously collects data on lighting equipment energy consumption and advertising display effectiveness. After initial processing by edge computing nodes, this data is transmitted to the cloud platform for in-depth analysis. The analysis results are used to evaluate the effectiveness of control strategies, identify potential issues, and then feed back to the system's data collection module as optimization input, forming a complete closed-loop optimization mechanism. This adaptive execution and feedback mechanism enables the system to continuously learn and improve, enhancing the intelligence level and service quality of urban lighting advertising management.

[0084] In one embodiment of the present invention, in step S1, status data of urban lighting equipment, advertising resources, and environmental scenes are collected through a multi-source perception network, and the status data are preprocessed and integrated using edge computing nodes to obtain a standardized data stream, including:

[0085] Step S1.1, collecting status data of the lighting device, the advertising resources, and the environmental scene, and transmitting the status data to the edge computing node via a preset data acquisition protocol; wherein the status data includes power parameters, brightness parameters, size parameters and resolution parameters of the advertising screen, and operating data of the environmental sensor;

[0086] Step S1.2, based on the state data, performing data preprocessing and spatiotemporal synchronization at the edge computing node to generate a data stream;

[0087] Step S1.3, performing data cleaning and feature extraction on the data stream to obtain the standardized data stream.

[0088] In step S1.1, a multi-source sensing network is deployed on existing city infrastructure. It includes environmental sensors, lighting sensors, pedestrian flow detectors, and advertising screen status monitors to collect real-time data related to city lighting and advertising displays. These sensors are connected to pre-configured edge computing nodes via wired or wireless connections. Standard IoT communication protocols (such as MQTT and CoAP) are used to transmit data to the edge nodes for preliminary processing.

[0089] In step S1.2, edge computing nodes are deployed in key areas of the city to preprocess and synchronize the raw data from different sensors in time and space. Preprocessing includes data format conversion, outlier detection, and preliminary data fusion. This synchronization ensures the consistency of data from different sensors in both time and space, providing a foundation for subsequent analysis.

[0090] In step S1.3, the data stream processed by edge computing is transmitted to the cloud platform data center, where further data cleaning (such as noise removal and missing value processing) and feature extraction (such as temporal features, spatial features, and device features) are performed. Through this series of processing, the original heterogeneous data is converted into a standardized data stream, providing a unified data foundation for subsequent analysis.

[0091] In one embodiment of the present invention, Figure 2 As shown, in step S2, local sensitive hashing is used to bucket the lighting equipment and advertising resources, and optimization is performed through iterative local search to obtain resource grouping results, including:

[0092] Step S2.1, constructing a feature vector based on the standardized data stream using dimensionality reduction and feature extraction techniques;

[0093] In step S2.1, after receiving the standardized data stream from step S1, the system first constructs a feature vector. This process aims to convert high-dimensional, heterogeneous data into a low-dimensional feature representation, reducing computational effort while preserving the data's key information. This is achieved using a two-stage feature processing approach: the first stage performs feature selection and normalization, and the second stage uses a deep autoencoder for nonlinear feature extraction.

[0094] During the feature selection and normalization phase, the system first analyzes the information content and relevance of each dimension of data. Using principal component analysis (PCA), the system identifies the principal components that contribute most to data variation; using non-negative matrix factorization (NMF), the system extracts key patterns in the data. These techniques help the system select the most representative feature dimensions from the raw high-dimensional data. Furthermore, the system normalizes data of different dimensions, mapping different types of data, such as position coordinates, power parameters, and brightness values, to a unified numerical range to facilitate subsequent processing.

[0095] During the deep autoencoder stage, a multi-layer encoder-decoder network structure is constructed. The encoder, consisting of multiple dimensionality reduction layers, progressively compresses the input data into low-dimensional latent representations; the decoder attempts to reconstruct the original input from these latent representations. By minimizing the reconstruction error, the network learns a nonlinear representation of the data. To maintain the similarity of geographically close devices in feature space, the system adds a spatial sensitivity constraint to the autoencoder's loss function: L = L_reconstruction + λ·L_space, where L_space measures the consistency between feature space distance and actual geographic distance.

[0096] This feature vector construction method not only effectively reduces data dimensionality and subsequent computational effort, but also preserves key information and spatial relationships within the data, providing high-quality input for locality-sensitive hashing and clustering analysis. Ultimately, 32- to 128-dimensional feature vectors were generated for each lighting fixture and advertising asset. These vectors capture the physical attributes and functional characteristics of the device while preserving its spatial distribution.

[0097] Step S2.2: Initially bucketing the lighting devices and advertising resources using a locality-sensitive hash function based on the feature vector.

[0098] In step S2.2, the system uses a modified Locality Sensitive Hashing (LSH) algorithm to initially bucket the lighting fixtures and advertising resources based on the feature vectors constructed in step S2.1. The core idea of the LSH algorithm is to construct a set of hash functions so that points close to each other in the feature space are much more likely to be mapped to the same "bucket" than points farther apart, thereby achieving rapid and coarse data classification.

[0099] To address the uneven distribution of urban spatial data, the present invention designs a family of adaptive hash functions. The basic hash function uses a random projection method: h(v) = floor((v·r+b) / w), where v is the eigenvector, r is the random vector, b is the random offset, and w is the quantization width. By observing the data distribution characteristics, the system dynamically adjusts the parameter w, making it larger in data-sparse areas (increasing sensitivity) and smaller in data-dense areas (reducing hash conflicts). This adaptive adjustment enables the hash function to better adapt to the uneven distribution of urban data.

[0100] To further improve hash quality, the system employs a multi-level LSH strategy, constructing L hash tables containing k hash functions. By adjusting the values of L and k, the system is able to strike a balance between recall and precision. In practice, the system first uses a smaller k value and a larger L value for coarse-grained bucketing. Within each bucket, it then uses a larger k value and a smaller L value for fine-grained bucketing, forming a hierarchical grouping structure.

[0101] In addition, an automatic threshold adjustment mechanism based on local density is introduced. By calculating the distribution density of points in the feature space, the system automatically adjusts the parameters of the hash function, avoiding the performance degradation problem of traditional LSH when dealing with unevenly distributed data. This adaptive mechanism enables the system to obtain better initial bucketing results under different urban areas and different densities of device distributions.

[0102] With this improved LSH algorithm, the preliminary data bucketing can be completed within the time complexity of O(n). The output initial grouping results retain the local proximity relationship of the data, providing a good starting point for subsequent iterative local search optimization. This preliminary grouping can transform the O(n2) complexity problem that may originally require global comparison into a nearly linear time problem that only needs to be optimized within a local range.

[0103] Step S2.3: Perform iterative local search optimization on the initial bucketing results to obtain the resource grouping results.

[0104] In step S2.3, the system performs iterative local search optimization on the initial bucketing results of step S2.2 to further improve the clustering quality. Traditional clustering optimization usually adopts a global search strategy with high computational complexity; while simple greedy optimization is prone to falling into local optimal solutions. The present invention designs an iterative local search algorithm with a time complexity of O(nlog n), effectively controlling the computational complexity while ensuring the clustering quality.

[0105] The core of iterative local search is the "perturbation - improvement" loop strategy. In the improvement stage, the system considers moving each data point from the current cluster to a neighboring cluster. If this move can improve the clustering objective function value, the move is accepted. The clustering objective function comprehensively considers multiple factors such as intra - cluster similarity, spatial continuity, energy load balance, and advertising coverage efficiency. To avoid falling into local optimality, in the perturbation stage, the system randomly selects a certain proportion of points (usually 5 - 10%) for re - allocation, creating an opportunity to jump out of the local optimum.

[0106] The nearly linear time complexity of the algorithm is mainly achieved through three technical innovations: Step one, constructing a proximity graph using the LSH results of step S2.2, restricting the search range of each point within its K - nearest neighbors (K << n), significantly reducing the number of comparison operations; Step two, adopting an incremental calculation technique, only updating the affected part of the objective function each time a point moves, avoiding global recalculation; Step three, introducing an early stopping strategy, terminating the iteration in advance when the improvement in consecutive rounds is below a preset threshold.

[0107] To verify the effectiveness of the clustering results, a multi-index adaptive clustering validation method was also used. This method constructs a comprehensive evaluation index system, including internal indicators (such as the silhouette coefficient, Davies-Bouldin index, and Calinski-Harabasz index) and external indicators (such as energy load balance and advertising coverage uniformity). At the same time, the system designed a parameter adaptive adjustment mechanism based on Bayesian optimization, which can automatically search for the optimal clustering parameter combination. In addition, the system also introduces a constraint verification mechanism based on domain knowledge to ensure that the clustering results meet actual engineering requirements such as urban planning zoning.

[0108] Through this multi-dimensional, adaptive clustering verification and optimization method, a fully verified and optimized resource grouping result is finally output, providing a high-quality grouping foundation for the subsequent construction of random regular graph models.

[0109] In step S2.1, feature vectors are constructed for the standardized data stream from step S1. This process employs a two-stage feature processing approach: first, feature selection and normalization are performed, using principal component analysis (PCA) and non-negative matrix factorization (NMF) techniques to identify and retain the most informative feature dimensions. A deep autoencoder network is then applied to compress the high-dimensional input into a low-dimensional latent representation. To ensure that geographically close devices are similar in feature space, a location-based regularization term is added to the autoencoder's loss function. The final output is a low-dimensional feature vector (typically 32-128 dimensions) for each lighting device and advertising asset.

[0110] In step S2.2, based on the feature vector constructed in step S2.1, an improved locality-sensitive hashing (LSH) algorithm is designed for initial data bucketing. The core of the improved LSH algorithm is to construct a set of hash functions such that points close to each other in the feature space are much more likely to be mapped to the same "bucket" than points farther apart. To address the uneven distribution of urban spatial data, an adaptive family of hash functions is designed. By dynamically adjusting the parameter w, the algorithm increases sensitivity in data-sparse areas and reduces hash conflicts in data-dense areas. A multi-level LSH strategy is also employed to construct L hash tables containing k hash functions. By adjusting the values of L and k, a balance is achieved between recall and precision.

[0111] In step S2.3, based on the initial bucketing result of step S2.2, an iterative local search optimization is performed on the initial bucketing result to obtain a final resource grouping result.

[0112] In one embodiment of the present invention, in step S2.3, iterative local search optimization is performed on the initial bucketing result to obtain the resource grouping result, including:

[0113] Step S2.3.1, using each bucket of the initial bucketing result as an initial cluster, constructing a clustering objective function; wherein the clustering objective function includes intra-cluster similarity, spatial continuity, energy load balance, and advertising coverage efficiency;

[0114] Step S2.3.2, based on the clustering objective function, performing a local search on the data points within each of the initial clusters, and updating the clustering objective function by an incremental calculation technique;

[0115] Step S2.3.3: When it is determined that the clustering objective function converges, the resource grouping result is obtained.

[0116] In step S2.3.1, based on the initial bucketing results, a clustering objective function suitable for urban lighting and advertising resources is constructed. This function comprehensively considers multiple factors such as intra-cluster similarity, spatial continuity, energy load balance, and advertising coverage efficiency. It is formally expressed as: F(C) = α·F

[0117] (C)+β·F(C)+γ·F(C)+δ·F(C), where C represents the clustering result, F1 to F4 represent the sub-objective functions respectively, and α, β, γ, and δ are weight coefficients.

[0118] In step S2.3.2, an iterative local search algorithm is used to optimize the clustering. The core of this algorithm lies in a "perturb-improve" cyclical strategy: During the improvement phase, each data point is considered for movement from its current cluster to a neighboring cluster. If the move improves the objective function, the move is accepted. During the perturbation phase, a certain percentage of points are randomly selected for reallocation, creating opportunities to escape from local optima. To achieve near-linear time complexity, three technical innovations are employed: LSH results are used to construct a proximity graph, limiting the search range of each point to its K nearest neighbors; incremental computation techniques are employed, with each point move only updating the affected portion of the objective function; and an early stopping strategy is introduced, terminating iterations prematurely when the improvement over multiple consecutive rounds falls below a preset threshold.

[0119] In step S2.3.3, if the change in the clustering objective function falls below the preset threshold multiple times in a row, the algorithm is considered to have converged, and the final resource grouping result is output. To verify the effectiveness of the clustering results, a multi-metric adaptive clustering validation method is also used. This includes internal metrics such as the Silhouette Coefficient, Davies-Bouldin Index, and Calinski-Harabasz Index, as well as external metrics such as energy load balance and advertising coverage uniformity.

[0120] In one embodiment of the present invention, in step S3, constructing a random regular graph model includes:

[0121] Step S3.1: Based on the resource grouping result and in combination with the state data of the environment scene, a heterogeneous graph structure of lighting device nodes, advertising resource nodes, and environment scene nodes is constructed;

[0122] Step S3.2: performing degree distribution regularization and edge weight randomization processing on the heterogeneous graph structure to obtain the random regularized graph model.

[0123] In step S3.1, a multi-layer heterogeneous graph structure is constructed based on the resource grouping results output from step S2. This graph structure consists of three basic node types: lighting equipment nodes, advertising resource nodes, and environmental scene nodes. Each type of node has a different set of attributes. For example, lighting equipment nodes contain attributes such as location, power, and illumination radius; advertising resource nodes contain attributes such as size, resolution, and viewing angle; and environmental scene nodes contain attributes such as regional characteristics and pedestrian density. Edges between nodes represent potential associations, including physical proximity, functional complementarity, and visual influence. To improve the computational efficiency of the graph structure, sparse matrix representation and spatial index optimization techniques are used to reduce the edge construction complexity from O(n²) to O(nlog n).

[0124] In step S3.2, the initial heterogeneous graph structure is subjected to random regularization. Unlike traditional regular graphs that require all nodes to have the same degree, this model allows the node degree to vary randomly within a certain range while maintaining the overall degree distribution in line with the preset statistical characteristics. The specific implementation adopts a three-stage random regularization process: the first stage is to regularize the degree distribution. According to the functional characteristics of different types of nodes, the probability model of the degree distribution is set, and the target degree is assigned to each node through Monte Carlo sampling; the second stage is to randomize the edge weights. Appropriate random perturbations are introduced to each edge in the graph according to the type of relationship it represents; the third stage is to perform global structural balance and apply a global optimization algorithm to restore the key structural characteristics of the graph.

[0125] In one embodiment of the present invention, in step S3, the graph structure is optimized using a Markov Chain Monte Carlo algorithm, including:

[0126] Step S3.3, constructing an energy function including resource allocation balance and control system responsiveness based on the random regular graph model;

[0127] Step S3.4, iteratively optimizing the random regular graph model using a block parallel Markov chain Monte Carlo algorithm to obtain an optimized graph structure;

[0128] Step S3.5: Based on the optimized graph structure and in combination with the energy function, the mapping relationship between the device nodes and the resource nodes is extracted to generate the resource configuration strategy.

[0129] In step S3.3, an energy function E(G) is defined to evaluate the quality of the graph structure G. This energy function comprehensively considers multiple aspects such as resource allocation balance, control system responsiveness, visual coordination, and energy efficiency: E(G) = w·E(G) + w·E(G) + w·E(G) + w·E(G), where E1 to E4 correspond to sub-energy functions of different optimization objectives, and w1 to w are weight coefficients.

[0130] In step S3.4, based on the energy function defined in step S3.3, a Markov Chain Monte Carlo (MCMC) algorithm with fast mixing properties is used to optimize the graph structure. To address the challenges of optimizing large-scale graph structures, three technical innovations are introduced: a block-parallel MCMC strategy is used to decompose the large graph into multiple overlapping subgraphs, and MCMC iterations are performed in parallel on each subgraph; an adaptive proposal distribution is designed to dynamically adjust the proposal step size based on the historical acceptance rate; and a temperature scheduling mechanism is introduced, with a high temperature in the initial stage to promote extensive exploration, and a gradual decrease in temperature to achieve refined optimization. The key to the fast mixing feature lies in the design of a non-local proposal mechanism, which combines local proposals (such as unilateral flipping and node attribute fine-tuning) with non-local proposals (such as subgraph reconstruction, pattern permutation, and cross-community exchange). This enables the algorithm to achieve "long-distance jumps" in the state space in a single step, significantly improving the mixing speed.

[0131] In step S3.5, the optimal mapping relationship between lighting devices and advertising resources is extracted based on the optimized graph structure. This process employs a multi-layered resource mapping relationship extraction approach: first, based on the graph's community structure analysis, closely connected device-resource clusters are identified; then, a bipartite graph matching algorithm, which considers node type constraints, is applied to find the optimal mapping relationship; finally, a constraint satisfaction check is performed to ensure the feasibility of resource mapping in actual implementation. To cope with dynamic environmental changes, a time-window-based mapping relationship update mechanism and an event-triggered update mechanism are also designed.

[0132] In one embodiment of the present invention, Figure 3 As shown, in step S4, a collaborative control strategy is generated by a multi-objective optimization algorithm, including:

[0133] Step S4.1, constructing a multi-objective function including energy efficiency and visual comfort based on the resource allocation strategy;

[0134] Based on the resource allocation strategy output in step S3, a multi-objective function for collaborative optimization is constructed. These objective functions are not simple single evaluation indicators, but rather a set of potentially conflicting evaluation dimensions, fully reflecting the complexity of smart city lighting advertising management. The system defines four core objective functions: energy efficiency, visual comfort, advertising display effectiveness, and environmental impact.

[0135] The energy efficiency objective function evaluates the ratio of system energy consumption to service effectiveness. This function takes into account factors such as the power parameters, working hours, and service coverage of lighting equipment and advertising screens, and calculates energy utilization efficiency through a nonlinear model. To improve accuracy, the system also introduces a time factor, setting different weight coefficients for different time periods (such as peak period, off-peak period, and valley period). The visual comfort objective function is based on the human eye visual perception model and evaluates the impact of the comprehensive visual effects of lighting and advertising on human comfort. This function integrates sub-indicators such as color temperature consistency, brightness uniformity, contrast suitability, and visual interference, and performs quantitative evaluation through a neural network model trained with perception experiment data.

[0136] The advertising display effectiveness objective function evaluates the visibility, appeal, and information delivery efficiency of advertising content. This function combines the location characteristics of the advertising screen, surrounding environmental conditions, and historical viewing data to construct an advertising effectiveness prediction model. It also considers the interaction between the advertising content and the surrounding lighting environment, such as the impact of lighting conditions on the perception of advertising color. The environmental impact objective function assesses the system's impact on the ecological environment, focusing primarily on light pollution, energy consumption, and visual disturbance. This function sets evaluation indicators based on environmental standards and urban planning requirements to ensure that system operation does not place an excessive burden on the urban ecological environment.

[0137] These objective functions are combined into a unified multi-objective optimization problem using weight coefficients: F = {f1(x),f2(x),f3(x),f4(x)}, where x represents a vector of decision variables, including lighting parameters and advertising display parameters. This multi-objective function allows for comprehensive evaluation and optimization across multiple dimensions, including efficiency, comfort, advertising effectiveness, and environmental protection.

[0138] Step S4.2, optimizing and solving the multi-objective function to obtain a Pareto optimal solution for resource allocation;

[0139] The constructed multi-objective function is optimized and solved to find a Pareto-optimal set of solutions for resource allocation. Due to inherent conflicts between objective functions (for example, increasing the brightness of advertisements may increase energy consumption and light pollution), traditional single-objective optimization methods struggle to find a solution that simultaneously optimizes all objectives. The concept of Pareto optimality provides a framework for addressing such problems: finding a set of solutions where improving any objective necessarily degrades at least one other objective.

[0140] An improved multi-objective evolutionary algorithm (MOEA) was used to solve this complex problem. Specifically, the Non-Dominated Sorting Genetic Algorithm III (NSGA-III) was selected as the underlying algorithm. This algorithm effectively maintains the diversity of solutions in high-dimensional objective spaces through its non-dominated sorting and reference point-assisted selection mechanisms. To adapt to the characteristics of smart city scenarios, three improvements were made to the standard NSGA-III: first, a problem-knowledge-based initial population generation strategy was designed to distribute initial solutions in more promising regions of the search space; second, adaptive crossover and mutation operators were developed to dynamically adjust parameters based on population diversity and convergence status; and finally, a local search strategy was introduced to fine-tune the frontier solutions and improve solution quality.

[0141] To address the computational challenges of city-scale optimization, a distributed computing architecture and GPU acceleration are employed. Using a regional decomposition strategy, the large-scale optimization problem is broken down into multiple subproblems, which are solved in parallel. The results are then consolidated using a boundary coordination mechanism. This divide-and-conquer approach enables the system to solve multi-objective city-scale optimization problems within an acceptable timeframe (typically minutes).

[0142] The algorithm ultimately outputs a set of Pareto-optimal solutions, each representing a resource allocation scheme with a different trade-off between the objective functions. These solutions form the Pareto frontier in the objective space, providing decision makers with a diverse range of options, allowing them to select the most appropriate allocation scheme based on their specific management needs and preferences.

[0143] Step S4.3: Generate the collaborative control strategy based on the Pareto optimal solution.

[0144] Based on the Pareto optimal solution set obtained in step S4.2, and combined with the actual operating environment and management policies, an executable collaborative control strategy is generated. This process does not simply select a Pareto optimal solution, but rather generates specific and feasible control instructions through a complex decision-making mechanism that considers the current system state, management preferences, and execution constraints.

[0145] First, a solution set screening mechanism is applied to select the solution that best suits the current situation from the Pareto front. This screening process combines multi-criteria decision-making methods with context-aware technology. Multi-criteria decision-making methods such as TOPSIS (Ranking Technique of Approximately Ideal Solutions) and AHP (Analytical Hierarchy Process) help the system balance multiple objectives; context-aware technology considers time factors (such as date and time period), spatial factors (such as regional characteristics), and activity factors (such as special events and holidays). For example, during prime time in commercial areas, the system may prioritize advertising display effects; while in residential areas at night, it may prioritize visual comfort and minimizing light pollution.

[0146] Next, decision variable mapping is performed to convert the abstract optimization variables into specific control parameters. This process involves device characteristic mapping (considering the performance parameters and control interfaces of different devices), environmental adaptation mapping (adjusting control parameters based on real-time environmental conditions), and timing planning mapping (designing execution plans for parameters that change over time). For example, the system might map the abstract "lighting intensity 50%" to the specific power value and PWM dimming parameters for a specific LED fixture.

[0147] Finally, a control instruction set is generated, formatting the control parameters into device-executable instructions and attaching the necessary execution timing and safety check logic. This control instruction set not only contains basic operating instructions for the lighting equipment and advertising screens, but also includes coordinated control logic to ensure the operation of the two types of equipment works in harmony, creating a harmonious and unified urban visual experience. For example, when the advertising content changes, the system may synchronously adjust the brightness and color temperature of the surrounding lighting to enhance the advertising effect; or, under specific environmental conditions, dynamically adjust the parameters of the lighting and advertising screens to adapt to the changing environment.

[0148] Through this sophisticated collaborative control strategy generation mechanism, the system can convert theoretical optimization results into practical and executable control solutions, realizing intelligent collaborative management of lighting equipment and advertising resources.

[0149] In step S4.1, a collaborative optimization multi-objective function is constructed based on the resource allocation strategy output in step S3. This multi-objective function comprehensively considers multiple dimensions, including energy efficiency, visual comfort, advertising display effect, and environmental impact. The energy efficiency objective function evaluates the ratio of system energy consumption to service effect; the visual comfort objective function, based on the human visual perception model, evaluates the impact of the combined visual effects of lighting and advertising on human comfort; the advertising display effect objective function evaluates the visibility, attractiveness, and information transmission efficiency of advertising content; and the environmental impact objective function evaluates the degree of impact of the system on the ecological environment (such as light pollution).

[0150] In step S4.2, the multi-objective functions are optimized and solved to find a Pareto-optimal solution set. Due to potential conflicts between objective functions, traditional single-objective optimization methods are not applicable. This solution employs an improved multi-objective evolutionary algorithm (MOEA), such as NSGA-III (Non-dominated Sorting Genetic Algorithm III), to maintain a diverse solution set and find a balance between multiple objectives. A resource allocation constraint processing mechanism is incorporated into the algorithm design to ensure that the generated solution satisfies the physical constraints of the actual system. Furthermore, to improve algorithm efficiency, an adaptive mutation strategy and a parallel computing architecture are employed to keep the solution time for large-scale problems within an acceptable range.

[0151] In step S4.3, an executable collaborative control strategy is generated based on the Pareto-optimal solution set obtained in step S4.2, combined with the actual operating environment and management policies. This process includes three steps: solution set screening, decision variable mapping, and control instruction generation. Solution set screening selects the most suitable solution from the Pareto frontier based on the current system state and decision preferences; decision variable mapping converts abstract optimization variables into specific control parameters; and control instruction generation formats the control parameters into a device-executable instruction set, adding the necessary execution timing and safety check logic.

[0152] In one embodiment of the present invention, in step S4, combining the user behavior data and the state data of the environmental scene includes:

[0153] Step S4.4, collecting data on ambient light intensity, weather conditions, and crowd density to obtain state data of the environmental scene;

[0154] Step S4.5, obtaining user dwell time and attention heat map based on video analysis technology to obtain the user behavior data;

[0155] Step S4.6: Perform spatiotemporal correlation analysis on the state data of the environmental scene and the user behavior data to obtain a scene feature vector.

[0156] Spatiotemporal correlation analysis is performed on environmental scene state data and user behavior data to explore the complex relationships between environmental factors and user behavior. This analysis reveals changing patterns in user behavior under different environmental conditions, helping the system build predictive models and proactively adapt to environmental changes. Spatiotemporal correlation analysis utilizes a variety of data mining techniques, including spatiotemporal autocorrelation analysis, conditional random field models, and deep learning methods.

[0157] First, data preprocessing and alignment are performed to ensure the consistency of data from different sources in time and space. This includes time synchronization (adjusting data with different sampling frequencies to a unified time axis), spatial registration (mapping location information in different coordinate systems to a unified reference system), and data interpolation (filling missing values and smoothing outliers). Then, the system applies spatiotemporal autocorrelation analysis methods, such as Moran's I and LISA statistics, to detect spatiotemporal clustering patterns and hot spots in the data. These analyses reveal the temporal and spatial distribution patterns of environmental variables and user behaviors, such as the spatial correlation between foot traffic and lighting brightness, or the temporal correlation between advertising attention and weather conditions.

[0158] Next, an environment-behavior relationship model is constructed to reveal the mechanisms by which environmental factors influence user behavior. The system employs a conditional random field (CRF) model to capture local dependencies, a long short-term memory network (LSTM) to model temporal patterns, and a graph neural network (GNN) to represent spatial interactions. These models can answer complex questions such as "How does lighting brightness change user dwell time?" and "How do advertising content preferences change in rainy weather?"

[0159] Finally, a scenario feature vector is generated, serving as an important input for collaborative optimization decisions. A scenario feature vector is a compact representation of key features of the environment state and user behavior, typically consisting of a 100-200-dimensional numerical vector. The vector design considers the information content, complementarity, and stability of the features, effectively supporting subsequent decision optimization. Through this in-depth spatiotemporal correlation analysis, the system establishes a bridge between the environment, users, and control strategies, enabling the decision-making process to consider both objective environmental conditions and the user's subjective experience.

[0160] In step S4.4, the environmental sensor network collects real-time data on ambient scene status, including ambient light intensity, weather conditions, temperature and humidity, and pedestrian density. After standardization and spatiotemporal synchronization, this data is used to represent the real-time state of the current urban environment. This environmental data has a direct impact on lighting control and advertising display effectiveness. For example, under varying weather and lighting conditions, lighting brightness and advertising contrast need to be adjusted to ensure optimal results.

[0161] In step S4.5, cameras and sensors deployed near the advertising screen use advanced video analysis technology to capture user behavior data, including duration of stay in front of the screen, gaze direction, and attention distribution. This data is processed using a privacy-preserving algorithm, removing personal identification information and retaining only statistical characteristics. Deep learning-based visual algorithms can generate attention heat maps, visually demonstrating which areas and content are most likely to attract user attention.

[0162] In step S4.6, a spatiotemporal correlation analysis is performed on the environmental scene state data and user behavior data to explore the relationship between environmental factors and user behavior. This analysis utilizes spatiotemporal data mining techniques, such as spatiotemporal autocorrelation analysis and conditional random field models, to identify key environmental factors and spatiotemporal patterns that influence user behavior. The analysis results form a scene feature vector, which serves as an important input for subsequent collaborative optimization decisions, enabling the system to dynamically adjust lighting and advertising control strategies based on environmental changes and user responses.

[0163] In one embodiment of the present invention, in step S5, controlling the lighting device and the advertising resources based on the collaborative control strategy includes:

[0164] Step S5.1, converting the collaborative control strategy into brightness parameters of the lighting device and display parameters of the advertising resources;

[0165] The command parsing engine converts the high-level collaborative control strategy output from step S4 into specific device-level control commands. This process is like translating an abstract conductor's intent into specific performance instructions for each instrument in an orchestra, ensuring that all devices execute the overall strategy in a coordinated and consistent manner. The command parsing engine consists of three core modules: a strategy parsing module, a parameter mapping module, and a command generation module.

[0166] The policy parsing module first understands the semantics and intent of the high-level control policy. Control policies are typically expressed as target states (e.g., "Main commercial street area lighting provides a comfortable and bright environment while prominently displaying holiday advertisements") or conversion instructions (e.g., "Gradually increase lighting brightness as sunset while adjusting advertising color saturation"). Using semantic analysis techniques, the policy parsing module decomposes these statements into specific control objectives, constraints, and priority rules. The parameter mapping module then converts the parsed control objectives into operating parameters for specific devices. This process takes into account differences in device characteristics, the impact of environmental conditions, and the requirements for synergy.

[0167] For example, the abstract goal of "comfortable lighting" is mapped to the brightness percentage, color temperature, and illumination angle combination of LED lamps in a specific area; the advertising requirement of "prominent display" is converted into the brightness, contrast, and content refresh rate parameters of the display screen. During the parameter mapping process, the system utilizes a device characteristic database that stores the performance parameters, control ranges, and response curves of each device model, ensuring that the generated parameter values are suitable for the technical capabilities of the specific device.

[0168] Finally, the instruction generation module formats the mapped parameters into an instruction set executable by the device. These instructions comply with the device communication protocol specifications (such as the DALI lighting control protocol, the DMX512 digital multi-channel communication protocol, etc.) and contain the necessary address information, command codes, and parameter values. The system also adds timestamps, execution order, and safety check logic to the instructions to ensure the reliability and security of the control process. The generated instructions are reliably distributed through a distributed message queue system (such as Apache Kafka or RabbitMQ), ensuring that the instructions are correctly delivered to the target device even in the event of network fluctuations or partial node failure.

[0169] Step S5.2, adjusting the brightness of the lighting device and the display content of the advertising resources through the Internet of Things control system;

[0170] The IoT control system receives distributed control commands and precisely controls lighting equipment and advertising resources. This process involves multi-layered control logic and coordination mechanisms, ensuring the system creates a unified and harmonious urban visual experience as intended. The lighting control system utilizes adaptive lighting technology based on environmental awareness, dynamically adjusting the brightness, color temperature, and illumination angle of LED lights based on real-time environmental conditions and control strategies.

[0171] It supports a variety of advanced lighting control functions, such as gradual transition (smoothly adjusting lighting parameters to avoid sudden changes that may cause visual discomfort), scene presets (quickly switching lighting solutions based on different scene requirements), intelligent grouping (logically grouping lamps for precise regional control), and timed scheduling (executing lighting changes according to preset plans). In terms of energy efficiency, the system automatically adjusts lighting intensity based on ambient light conditions, such as reducing artificial lighting output when natural light is sufficient and enhancing lighting effects on cloudy days or at night, to achieve rational energy utilization.

[0172] The advertising content management system adjusts parameters such as the display plan, brightness, contrast, and animation effects of the advertising content based on control instructions. The system supports intelligent content scheduling, such as selecting the most appropriate advertising content based on time of day, weather, and audience characteristics. It also supports dynamic rendering, which can adjust the visual effects of ads in real time to adapt to environmental changes. It also supports interactive display, adjusting content presentation based on audience response or triggering specific interactive modules. The advertising system also features content adaptation capabilities, automatically optimizing content layout and display effects based on the size, resolution, and color characteristics of different display devices.

[0173] A two-way communication and coordination mechanism has been established between the lighting and advertising systems, ensuring they work together to create an optimal visual experience. For example, when a bright advertisement screen displays content, the surrounding lighting intensity can be appropriately reduced to enhance the visual impact of the advertisement. When the advertisement content changes, the lighting tone can be coordinated to enhance the atmosphere. During special events or emergencies, the two systems can respond synchronously to jointly convey important information. This coordinated control not only enhances advertising effectiveness and the lighting experience, but also optimizes overall energy use and reduces visual distraction.

[0174] Step S5.3: Collect the energy consumption data of the lighting equipment and the display effect data of the advertising resources.

[0175] The built-in monitoring module continuously collects data on lighting equipment energy consumption and advertising display performance. This real-time feedback data is key to system self-assessment and optimization, forming a complete closed loop from control to monitoring and optimization. Energy consumption data is collected through smart meters and the device's built-in power monitoring unit, recording parameters such as voltage, current, power factor, and cumulative power consumption of lighting equipment and advertising screens. The system processes this raw data to calculate energy efficiency indicators (such as lighting power density per square meter and energy consumption per viewer) and detect anomalies (such as sudden changes in device power consumption and early warning of energy efficiency degradation).

[0176] Ad display performance data is collected through various channels, including display device operating parameters (actual brightness, contrast, refresh rate), environmental factors (ambient light interference, viewing angle occlusion), and audience response indicators (number of viewers, dwell time, and interaction rate). The system pays special attention to discrepancies between expected and actual performance, such as the deviation between planned and actual brightness and the ratio of estimated to actual viewers. This discrepancy data helps identify inaccurate assumptions in system models or unexpected changes in the environment.

[0177] Collected data first undergoes preliminary processing at the edge computing node, including data compression (reducing transmission load), anomaly detection (flagging possible faults or errors), and feature extraction (calculating key performance indicators). The processed data is then transmitted to the cloud platform via a secure channel for more in-depth analysis, such as trend analysis (identifying long-term patterns of change), correlation analysis (discovering dependencies between factors), and comparative analysis (comparing with historical data or similar facilities).

[0178] The analysis results are used in multiple areas: performance evaluation (assessing the effectiveness of the current control strategy), problem diagnosis (identifying bottlenecks or defects in the system), model calibration (updating predictive model parameters), and strategy optimization (adjusting control rules and parameters). This feedback data ultimately flows back to the system's data acquisition module as input for the next round of decision optimization, forming a complete closed-loop optimization mechanism. Through this continuous data collection and feedback analysis, the system can continuously learn and improve, adapt to environmental changes, improve resource utilization efficiency, and achieve truly intelligent management.

[0179] In step S5.1, the high-level collaborative control strategy output from step S4 is converted into device-level control instructions through the instruction parsing engine. The engine includes a strategy parsing module, a parameter mapping module, and an instruction generation module. The strategy parsing module understands the semantics of the high-level strategy; the parameter mapping module maps abstract control parameters to the operating parameters of specific devices, such as converting lighting effect requirements into the brightness, color temperature, and power parameters of a specific LED lamp; and the instruction generation module generates standardized control instructions based on the device communication protocol requirements. The generated instructions are reliably distributed through a distributed message queue system (such as Kafka and RabbitMQ) to ensure that the control instructions can be transmitted to each device in a timely and accurate manner.

[0180] In step S5.2, the IoT control system receives the control instructions and precisely controls the lighting equipment and advertising resources. The lighting control system utilizes adaptive lighting technology based on environmental awareness to dynamically adjust the brightness, color temperature, and illumination angle of the LED lights according to real-time environmental conditions and control strategies. The advertising content management system, in turn, adjusts parameters such as the display plan, brightness, contrast, and animation effects of the advertising content based on the control instructions. A collaborative mechanism has been established between the two systems to ensure that the lighting effects and advertising display work together to create an optimal visual experience.

[0181] In step S5.3, the built-in monitoring module continuously collects lighting equipment energy consumption data (such as power, current, and operating hours) and advertising display performance data (such as actual brightness, content switching frequency, and number of viewers). This data is initially processed and aggregated by the edge computing node and then transmitted to the cloud platform for in-depth analysis. The analysis results are used to evaluate the effectiveness of the control strategy, identify potential problems, and provide feedback for strategy optimization. This closed-loop feedback mechanism enables continuous learning and improvement, improving resource utilization efficiency and service quality.

[0182] like Figure 4 As shown, an embodiment of the present invention further provides a smart city lighting advertising management device, comprising:

[0183] The data acquisition module 10 is used to collect status data of urban lighting equipment, advertising resources, and environmental scenes through a multi-source perception network, and pre-process and fuse the status data using edge computing nodes to obtain a standardized data stream;

[0184] A resource grouping module 20 is configured to bucket lighting equipment and advertising resources using locality-sensitive hashing based on the standardized data stream, and optimize the buckets through iterative local search to obtain resource grouping results;

[0185] A resource mapping module 30 is configured to construct a random regular graph model based on the resource grouping results, optimize the graph structure using a Markov chain Monte Carlo algorithm, extract the mapping relationship between lighting equipment and advertising resources, and generate a resource allocation strategy;

[0186] A control decision module 40 is configured to generate a collaborative control strategy based on the resource allocation strategy, combined with user behavior data and state data of the environmental scenario, using a multi-objective optimization algorithm;

[0187] The execution feedback module 50 is used to control the lighting equipment and advertising resources based on the collaborative control strategy; at the same time, the operating status data and display effect data of both the lighting equipment and the advertising resources are collected, and the operating status data and the display effect data are fed back to the multi-source perception network as optimization input to realize smart city lighting advertising management.

[0188] Preferably, the data acquisition module 10 includes a multi-source perception submodule and an edge computing submodule, wherein the multi-source perception submodule is used to collect status data of urban lighting equipment, advertising resources and environmental scenes, and the edge computing submodule is used to preprocess and fuse the status data to obtain a standardized data stream.

[0189] Preferably, the resource grouping module 20 includes a feature vector construction submodule, a local sensitive hash submodule and an iterative local search submodule, wherein the feature vector construction submodule is used to construct a feature vector based on the standardized data stream, the local sensitive hash submodule is used to perform initial bucketing based on the feature vector, and the iterative local search submodule is used to optimize the initial bucketing result to obtain the resource grouping result.

[0190] Preferably, the resource mapping module 30 includes a graph structure construction submodule, a random regularization submodule and an MCMC optimization submodule, wherein the graph structure construction submodule is used to construct a heterogeneous graph structure based on the resource grouping results, the random regularization submodule is used to perform random regularization processing on the heterogeneous graph structure, and the MCMC optimization submodule is used to optimize the graph structure through the Markov chain Monte Carlo algorithm and extract the mapping relationship.

[0191] Preferably, the control decision module 40 includes a multi-objective function construction submodule, an environmental perception data integration submodule, a user behavior analysis submodule and a multi-objective optimization submodule, which are used to generate a collaborative control strategy based on resource allocation strategy and environmental user data.

[0192] Preferably, the execution feedback module 50 includes a control instruction parsing submodule, a device control submodule and a performance evaluation submodule, which are used to execute the control strategy and collect feedback data to form a closed-loop optimization.

[0193] The present invention realizes the collection, processing and fusion of multi-source heterogeneous data by constructing a four-layer architecture system; adopts a method combining improved local sensitive hashing and iterative local search to achieve near-linear time clustering; designs a random regular graph model and a fast hybrid MCMC algorithm to achieve optimal mapping of lighting equipment and advertising resources; develops a multi-objective optimization decision system to achieve coordinated control of lighting and advertising; and designs an adaptive execution and feedback mechanism to form a closed-loop optimization, thereby improving the efficiency and effectiveness of smart city lighting advertising management.

[0194] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. All equivalent structural transformations made by using the contents of the present invention description and drawings under the inventive concept of the present invention, or direct / indirect application in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A smart city lighting advertising management method, characterized in that: include: The state data of urban lighting equipment, advertising resources and environmental scenes are collected through a multi-source perception network, and the state data are pre-processed and integrated using edge computing nodes to obtain a standardized data stream; Based on the standardized data stream, locality-sensitive hashing is used to bucket the lighting equipment and advertising resources, and optimization is performed through iterative local search to obtain resource grouping results; Based on the resource grouping results, a random regular graph model is constructed, the graph structure is optimized using a Markov chain Monte Carlo algorithm, the mapping relationship between lighting equipment and advertising resources is extracted, and a resource allocation strategy is generated; Based on the resource allocation strategy, combined with user behavior data and status data of the environmental scenario, a collaborative control strategy is generated through a multi-objective optimization algorithm; Based on the collaborative control strategy, lighting equipment and advertising resources are controlled; at the same time, the operating status data and display effect data of both the lighting equipment and the advertising resources are collected, and the operating status data and the display effect data are fed back to the multi-source perception network as optimization input to realize smart city lighting advertising management.

2. The method according to claim 1, characterized in that The state data of urban lighting equipment, advertising resources and environmental scenes are collected through the multi-source perception network, and the state data are pre-processed and integrated using the edge computing node to obtain a standardized data stream, including: Collecting status data of the lighting device, the advertising resources, and the environmental scene, and transmitting the status data to the edge computing node through a preset data acquisition protocol; wherein the status data includes power parameters, brightness parameters, size parameters of the advertising screen, resolution parameters, and working data of the environmental sensor; Based on the state data, data preprocessing and spatiotemporal synchronization are performed at the edge computing node to generate a data stream; The data stream is cleaned and features are extracted to obtain the standardized data stream.

3. The method according to claim 1, characterized in that The lighting equipment and advertising resources are bucketed using local sensitive hashing and optimized through iterative local search to obtain resource grouping results, including: Based on the standardized data stream, construct a feature vector using dimension reduction and feature extraction techniques; Based on the feature vector, initially bucketing the lighting devices and advertising resources using a locality-sensitive hash function; Iterative local search optimization is performed on the initial bucketing result to obtain the resource grouping result.

4. The method according to claim 3, characterized in that The iterative local search optimization of the initial bucketing result to obtain the resource grouping result includes: Using each bucket of the initial bucketing result as an initial clustering cluster, a clustering objective function is constructed; wherein the clustering objective function includes intra-cluster similarity, spatial continuity, energy load balance, and advertising coverage efficiency; Based on the clustering objective function, performing local search on the data points within each of the initial clusters, and updating the clustering objective function by using an incremental calculation technique; When it is determined that the clustering objective function converges, the resource grouping result is obtained.

5. The method according to claim 1, wherein The constructing of the random regular graph model includes: Based on the resource grouping result and in combination with the state data of the environmental scene, a heterogeneous graph structure of lighting device nodes, advertising resource nodes and environmental scene nodes is constructed; Degree distribution regularization and edge weight randomization are performed on the heterogeneous graph structure to obtain the random regular graph model.

6. The method according to claim 5, characterized in that The optimization of the graph structure by the Markov Chain Monte Carlo algorithm includes: Based on the random regular graph model, an energy function including resource allocation balance and control system responsiveness is constructed; Iteratively optimizing the random regular graph model using a block parallel Markov chain Monte Carlo algorithm to obtain an optimized graph structure; Based on the optimized graph structure and in combination with the energy function, the mapping relationship between device nodes and resource nodes is extracted to generate the resource configuration strategy.

7. The method according to claim 1, characterized in that The generating of the collaborative control strategy by the multi-objective optimization algorithm includes: Based on the resource allocation strategy, a multi-objective function including energy efficiency and visual comfort is constructed; Optimizing and solving the multi-objective function to obtain a Pareto optimal solution for resource allocation; Based on the Pareto optimal solution, the collaborative control strategy is generated.

8. The method according to claim 1, characterized in that The state data combining the user behavior data and the environmental scenario includes: Collecting data on ambient light intensity, weather conditions, and crowd density to obtain status data of the environmental scene; Obtain user behavior data by obtaining user dwell time and attention heat map based on video analysis technology; Performing spatiotemporal correlation analysis on the state data of the environmental scene and the user behavior data to obtain a scene feature vector.

9. The method according to claim 1, characterized in that The controlling of lighting equipment and advertising resources based on the collaborative control strategy includes: Converting the collaborative control strategy into brightness parameters of the lighting device and display parameters of the advertising resources; Adjusting the brightness of the lighting device and the display content of the advertising resources through an Internet of Things control system; Collect energy consumption data of the lighting equipment and display effect data of the advertising resources.

10. A smart city lighting advertising management device, characterized in that: include: A data acquisition module is used to collect status data of urban lighting equipment, advertising resources, and environmental scenes through a multi-source perception network, and use edge computing nodes to preprocess and fuse the status data to obtain a standardized data stream; A resource grouping module, configured to bucket lighting equipment and advertising resources using locality-sensitive hashing based on the standardized data stream, and optimize through iterative local search to obtain resource grouping results; A resource mapping module is used to construct a random regular graph model based on the resource grouping results, optimize the graph structure using a Markov chain Monte Carlo algorithm, extract the mapping relationship between lighting equipment and advertising resources, and generate a resource allocation strategy; A control decision module, configured to generate a collaborative control strategy based on the resource allocation strategy, combined with user behavior data and status data of the environmental scenario, through a multi-objective optimization algorithm; An execution feedback module is used to control lighting equipment and advertising resources based on the collaborative control strategy; at the same time, the operating status data and display effect data of both the lighting equipment and the advertising resources are collected, and the operating status data and the display effect data are fed back to the multi-source perception network as optimization input to realize smart city lighting advertising management.

Images