AI exhibition hall control method and system based on crowd thermal distribution monitoring

Abstract

The present invention discloses an AI exhibition hall control method and system based on crowd thermal distribution monitoring. This method constructs a fully covered perception network, uses thermal imaging cameras, Wi-Fi probes, and infrared sensors to acquire crowd density distribution data, behavioral data, and environmental data, and generates a thermal distribution map. The method intuitively displays crowd distribution characteristics by fusing and annotating discrete density points into a continuous thermal field dataset. A convolutional neural network and a spatiotemporal graph neural network are combined to effectively model the spatiotemporal correlation of crowd flow. Based on prediction results 6-8 minutes in advance, a closed loop of prediction, adjustment, and verification is achieved. By dynamically adjusting the control strategy based on the predicted crowd distribution trend and business objectives, the method can dynamically adjust the guided route or interactive form according to the crowd density. The method can also adjust the exhibit layout in advance to reduce the number of round trips for visitors and improve the overall flow efficiency, thereby enhancing the display effect of the AI exhibition hall.

Description

Technical Field

[0001] The present invention belongs to the field of AI exhibition hall control technology, and specifically relates to an AI exhibition hall control method and system based on crowd thermal distribution monitoring. Background Art

[0002] AI exhibition hall operations involve multiple aspects such as equipment control, security monitoring, and energy management. AI can realize intelligent linkage of equipment, automatically generate energy consumption reports, and even reduce equipment failures and labor costs through predictive maintenance.

[0003] The current method of controlling AI exhibition halls is mainly based on AI data analysis and decision-making, using cameras and sensors to collect audience behavior data. AI can analyze audience interests and optimize exhibit layout or explanation content. For example, if most visitors stay in front of a certain exhibit for a short time, the system can automatically adjust the lighting or interactive form of the exhibition area.

[0004] However, since AI analysis models mainly rely on training data when analyzing and making decisions, if the exhibition hall environment is very different from the training scenario, it may lead to misjudgment, resulting in poor real-time adjustment of the exhibition hall's guided route or interactive form, thereby affecting the display effect of the AI exhibition hall. For example, in extreme crowd conditions, sensors may fail due to data overload, making it difficult to accurately obtain the distribution of exhibition hall audiences and timely adjust the guided route or interactive form. Therefore, we need to propose an AI exhibition hall control method and system based on crowd thermal distribution monitoring to solve the above-mentioned problems, so that it can dynamically adjust the guided route or interactive form according to the crowd density, thereby improving the display effect of the AI exhibition hall. Summary of the Invention

[0005] The purpose of the present invention is to provide an AI exhibition hall control method and system based on crowd thermal distribution monitoring, which can dynamically adjust the guided route or interactive form according to the crowd density, improve the display effect of the AI exhibition hall, and solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] The AI exhibition hall control method based on crowd thermal distribution monitoring includes the following steps:

[0008] S1. Install multi-source sensor equipment on the ceiling, walls, or floor of the exhibition hall to build a fully covered perception network. The multi-source sensor equipment includes thermal imaging cameras, infrared sensors, WiFi probes, RFID tag readers, and environmental sensors.

[0009] S2, using multi-source sensor equipment to capture crowd density distribution data, behavior data, and environmental data in real time and generate a thermal distribution map;

[0010] S3. Filter the real-time data obtained by multi-source sensor equipment, then fuse the thermal distribution data, behavioral data, and environmental data, and mark the thermal characteristics of each area to form a continuous thermal field data set;

[0011] S4. Build a real-time thermal prediction model by combining convolutional neural networks and spatiotemporal graph neural networks, and predict the crowd distribution trend within 6-8 minutes based on thermal characteristics.

[0012] S5. Dynamically adjust control strategies based on predicted population distribution trends and business objectives;

[0013] S6. Real-time monitoring of the changes in thermal distribution after dynamic adjustment, and optimization of the parameters of the thermal prediction model according to the changes in thermal distribution.

[0014] Preferably, the thermal imaging camera resolution Pixels, deployment spacing ≤ 5 meters; WiFi probe supports Bluetooth 5.0+WiFi6 protocol, positioning error ≤ 1.5 meters; infrared sensor coverage angle ≥ 120°, detection distance ≤ 8 meters.

[0015] Preferably, the thermal distribution map generation process is as follows:

[0016] A1. Divide the thermal camera's field of view into multiple pixel blocks, calculate the number of human targets within each pixel block, and calculate the density value based on the actual area corresponding to each pixel block;

[0017] A2. Smoothing the discrete density points to generate a continuous density field;

[0018] A3. Map density values to thermal levels from 0 to 100.

[0019] A4. Use warm and cold gradients to color-code high-density and low-density distributions, and output thermal distribution maps through GIS rendering.

[0020] Preferably, the process of establishing the continuous thermal field data set is as follows:

[0021] B1. Identify and eliminate data that deviates from the normal range through outlier detection algorithms;

[0022] B2. Use linear interpolation to fill in missing data and use a low-pass filter to filter the data;

[0023] B3. Fusion of thermal distribution data, behavioral data, and environmental data based on timestamps to obtain a visual record containing multi-dimensional information;

[0024] B4. Calculate the crowd density based on the number of people and area within the area, use it as the basic indicator of thermal characteristics, and add a label to each area;

[0025] B5. Normalize the numerical data and organize the normalized data into a structured table or database table structure;

[0026] B6. Add metadata of the dataset to the structured table or database table structure to form a standardized dataset.

[0027] Preferably, in step B3, the fusion process is:

[0028] B31. Synchronize thermal distribution data, behavioral data, and environmental data based on timestamps to make different types of data consistent in time.

[0029] B32. Map thermal distribution data, behavioral data, and environmental data to a unified spatial coordinate system to make the data spatially comparable;

[0030] B33. Map the characteristics of different types of data into color and intensity values in the thermal distribution map, unify the data scale through normalization, and display the data distribution density in color depth on a two-dimensional plane to form a visual record containing multi-dimensional information.

[0031] Preferably, in step S4, the real-time thermal prediction model establishment process is as follows:

[0032] C1. Divide the normalized data into time windows and spatial grids to form input tensors;

[0033] C2, process the input tensor through a multi-layer convolutional neural network;

[0034] C3, calculate the spatial attention weight based on the input tensor processing results;

[0035] C4. Combine the convolution results with time series data and use the gated recurrent algorithm to learn the dependencies of crowd flows in the time dimension.

[0036] C5. Calculate the time attention weight based on the dependency of crowd flow in the time dimension;

[0037] C6. Build a prediction model by combining temporal attention weight, spatial attention weight, spatial features, and temporal features;

[0038] C7. The prediction model is trained using standardized data set data, and the parameters of the prediction model are optimized using the loss function. The optimized prediction model is a real-time thermal prediction model.

[0039] Preferably, the crowd distribution trend prediction process is as follows:

[0040] D1: Real-time collection of multi-source data 6-8 minutes before the current moment, and normalization of the collected data;

[0041] D2. Use the normalized data to predict the thermal prediction model to obtain the prediction results;

[0042] D3. Restore the predicted results to the actual thermal level;

[0043] D4. Convert the predicted heat level into a color heat map, using different colors to represent the crowd density level.

[0044] Preferably, the control content of the control strategy includes control based on exhibit attention, control based on visit fluency, and control of interactive experience; wherein, control based on exhibit attention includes flow optimization of hot exhibits and activation of unpopular areas;

[0045] The control method based on the visit fluency is: when the channel area density value is predicted to be ≥60 people / m 2 , and when the flow speed is less than 0.5m / s, congestion warning and diversion control will be carried out;

[0046] The control method based on interactive experience is: if it is predicted that the number of people queuing in the interactive device area is ≥5 and the waiting time is >10 minutes, dynamic scheduling of the interactive device will be performed.

[0047] Preferably, the process of optimizing the parameters of the thermal prediction model is as follows:

[0048] E1. Collect actual thermal data every two minutes using a thermal imaging camera, compare it with the predicted thermal map, and calculate the error value for each grid;

[0049] E2. Add the latest real-world data to the standardized dataset and use incremental learning to update the parameters of the convolutional neural network and spatiotemporal graph neural network.

[0050] E3. If the spatial attention weight of a certain area is continuously low but the actual flow of people is dense, increase the weight of the regional features in the spatial attention calculation;

[0051] If the prediction error of a certain period in the time series is large, the time attention weight of this period is enhanced;

[0052] E4. Use the sliding window method to verify the optimized model using the data of the past hour. If the prediction error is reduced to within 5%, save the new model; otherwise, repeat E1-E5.

[0053] Based on the AI exhibition hall control method based on crowd thermal distribution monitoring described above, the present invention also provides an AI exhibition hall control system based on crowd thermal distribution monitoring, including: a data acquisition module, which acquires the original data stream in the exhibition hall in real time through multi-source sensor equipment, and the original data stream includes the thermal coordinates, residence time, movement speed and environmental parameters of the crowd;

[0054] The heat map generation module converts the sensor data into a visual heat distribution for real-time monitoring and analysis;

[0055] The data processing and integration module cleans, completes, and integrates multi-source data to form a standardized data set of continuous thermal fields;

[0056] The real-time prediction module uses a combination of convolutional neural networks and spatiotemporal graph neural networks to establish a real-time thermal prediction module, predicting crowd distribution trends 6-8 minutes in advance;

[0057] The execution control module dynamically adjusts exhibition hall resources based on forecast results and business goals. Exhibition hall resources include optimizing exhibit attention, flow, and interactive experience.

[0058] The feedback optimization module is used to build a closed loop of monitoring and optimization, monitor the changes in thermal distribution after dynamic adjustment in real time, and optimize the parameters of the thermal prediction model according to the changes in thermal distribution.

[0059] The AI exhibition hall control method and system based on crowd thermal distribution monitoring proposed in this invention have the following advantages over the existing technology:

[0060] 1. The present invention uses thermal imaging cameras, WiFi probes, and infrared sensors to acquire crowd density distribution data, behavioral data, and environmental data, and generates a thermal distribution map. By fusing and annotating discrete density points into a continuous thermal field dataset, the present invention intuitively displays crowd distribution characteristics. It utilizes a combination of convolutional neural networks and spatiotemporal graph neural networks to effectively model the spatiotemporal correlation of crowd flows. Based on prediction results 6-8 minutes in advance, it implements a closed loop of prediction-adjustment-verification. By dynamically adjusting control strategies based on predicted crowd distribution trends and business objectives, it is possible to dynamically adjust guided routes or interactive formats based on crowd density. It can also adjust exhibit layouts in advance, reducing the number of round trips for visitors and improving overall flow efficiency, thereby enhancing the display effect of the AI exhibition hall.

[0061] 2. The present invention realizes the full-chain intelligence from data collection to business implementation through the collaboration of data acquisition module, heat map generation module, data processing and integration module, real-time prediction module, execution control module and feedback optimization module. The modules are linked by data flow and combined with deep learning and dynamic control. It not only meets the real-time monitoring needs of the exhibition hall, but also dynamically adjusts the guided route or interactive form according to the density of the crowd, thereby improving the display effect of the AI exhibition hall. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 A flowchart of an AI exhibition hall control method according to an embodiment of the present invention is shown;

[0063] Figure 2 A flowchart of a thermal distribution diagram generation process according to an embodiment of the present invention is shown;

[0064] Figure 3 A flowchart of establishing a data set of a continuous thermal field according to an embodiment of the present invention is shown;

[0065] Figure 4 A flowchart of establishing a real-time thermal prediction model according to an embodiment of the present invention is shown;

[0066] Figure 5 A system block diagram of an AI exhibition hall control system according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0067] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. The specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0068] The present invention provides Figure 1-4 The AI exhibition hall control method based on crowd thermal distribution monitoring shown includes the following steps:

[0069] S1. Install multi-source sensor equipment on the ceiling, walls, or floor of the exhibition hall to build a fully covered perception network. The multi-source sensor equipment includes thermal imaging cameras, infrared sensors, WiFi probes, RFID tag readers, and environmental sensors.

[0070] Among them, the thermal imaging camera resolution Pixels, deployment spacing ≤ 5 meters to ensure the accuracy of the heat map grid The WiFi probe supports Bluetooth 5.0 and WiFi 6 protocols, with a positioning error of ≤1.5 meters. The infrared sensor covers an angle of ≥120° and a detection distance of ≤8 meters. The environmental sensor monitors air quality in real time, triggering a ventilation warning when the threshold exceeds 800ppm.

[0071] S2, using multi-source sensor equipment to capture crowd density distribution data, behavioral data, and environmental data in real time, and generate a thermal distribution map based on the crowd density distribution data;

[0072] The process of generating a heat map is as follows:

[0073] A1. Divide the thermal camera's field of view into multiple pixel blocks, calculate the number of human targets in each pixel block, and calculate the density value based on the actual area corresponding to each pixel block. The density value calculation formula is:

[0074] ,in, is the density value of each pixel block, is the number of human targets in each pixel block, is the row index, is the column index, The actual area corresponding to a single pixel block;

[0075] A2. Smooth the discrete density points to generate a continuous density field. The smoothing formula is:

[0076] ,in, For coordinates The density estimate at , is the total number of sample points, is the bandwidth parameter used to control the degree of smoothing, For the The coordinates of the sample points, is the Gaussian kernel function;

[0077] A3. Map the density value to the thermal level of 0-100. The mapping formula is:

[0078] ,in, For coordinates The thermal level value at For coordinates The crowd density value at and are the global maximum crowd density value and the global minimum crowd density value respectively;

[0079] A4. Use a warm-cold gradient to color-code high-density and low-density distributions, and output a thermal distribution map after GIS rendering. GIS data is managed by layers (such as basemap, hotspot layer, and dynamic layer). Each layer corresponds to a specific type of geographic feature. When rendering, the layers are drawn by calling a display list (such as OpenGL's display list), and only the changed parts are updated to improve efficiency.

[0080] S3. Filter the real-time data obtained by multi-source sensor equipment, then fuse the thermal distribution data, behavioral data, and environmental data, and mark the thermal characteristics of each area to form a continuous thermal field data set;

[0081] The process of establishing a continuous thermal field dataset is as follows:

[0082] B1. Identify and eliminate data that deviates from the normal range through outlier detection algorithms;

[0083] B2. Use linear interpolation to fill in missing data to ensure data integrity and continuity, and use a low-pass filter to filter the data to remove high-frequency noise;

[0084] B3. Fusion of thermal distribution data, behavioral data, and environmental data based on timestamps to obtain a visual record containing multi-dimensional information;

[0085] The fusion process is:

[0086] B31. Synchronize thermal distribution data, behavioral data, and environmental data based on timestamps to make different types of data consistent in time.

[0087] B32. Map thermal distribution data, behavioral data, and environmental data to a unified spatial coordinate system to make the data spatially comparable;

[0088] B33. Map the characteristics of different types of data into color and intensity values in the heat distribution map. Use normalization to unify the data scale and display the data distribution density using color depth on a two-dimensional plane, forming a visual record containing multi-dimensional information.

[0089] B4. Calculate crowd density based on the number of people and area within the area as a basic indicator of thermal characteristics. Label each area. The labels added include but are not limited to thermal level, area type, and timestamp information.

[0090] B5. Normalize the numerical data and organize the normalized data into a structured table or database table structure;

[0091] B6. Add metadata of the data set to the structured table or database table structure to form a standardized data set. The metadata includes data source, acquisition time, sensor type, etc., to enhance the comprehensibility and traceability of the data.

[0092] S4. Build a real-time thermal prediction model by combining convolutional neural networks and spatiotemporal graph neural networks, and predict the crowd distribution trend within 6-8 minutes based on thermal characteristics.

[0093] The process of establishing a real-time thermal prediction model is as follows:

[0094] C1. Divide the normalized data into time windows and spatial grids to form the input tensor X. The input tensor formula is:

[0095] , where X is the input tensor value, is the spatial grid size, T is the time step of the time window, C is the number of feature channels, is the set of real numbers;

[0096] C2. Process the input tensor X through a multi-layer convolutional neural network. The formula of the multi-layer convolutional neural network is:

[0097] ,in, For the Layer spatial feature map, For the The convolution kernel weights of the layer, is the activation function, is the convolution operation, For the The bias term of the layer, For the Spatial feature map of the layer;

[0098] C3. Calculate the spatial attention weight based on the processing result of the input tensor X. The calculation formula of the spatial attention weight is:

[0099] ,in, For the The characteristics of the spatial location, For all spatial positions The characteristics of the location, is the spatial attention weight, Used for The exponential function that performs exponential operation on the result of the calculation, For the Features of spatial locations The calculated attention score measures the model's attention to this location feature. A higher score means the model considers this location feature more important for focusing on key areas of the exhibition hall (such as popular exhibits).

[0100] C4. Combine the convolution results with the time series data and use the gated loop algorithm to learn the dependency of crowd flow in the time dimension. The formula of the gated loop algorithm is:

[0101] ,

[0102] in, is the hidden state at time t, Input data for time t, To perform graph convolution operations on graph structure data for feature extraction and transformation, It is the adjacency matrix in the graph structure, which describes the connection relationship and weight between the nodes in the graph. A gate loop algorithm that controls the flow of information through a gating mechanism. is the hidden state at time t-1;

[0103] C5. Calculate the time attention weight based on the dependency of crowd flow in the time dimension. The calculation formula for time attention weight is:

[0104] ,in, is the feature of the t-th time step, Center all time steps The features of the time step, is the temporal attention weight, Used for The exponential function that performs exponential operation on the result of the calculation, For the Features of time steps The calculated attention score;

[0105] C6. Combine the temporal attention weight, spatial attention weight, spatial features, and temporal features to establish a prediction model. The prediction model formula is:

[0106] ,in, is the predicted heat map, is the spatial attention weight, is an element-by-element multiplication that integrates the weighted spatial and temporal features. is the spatial feature matrix, is the temporal attention weight, is the time feature matrix;

[0107] C7. The prediction model is trained using standardized data set data, and the parameters of the prediction model are optimized using the loss function. The optimized prediction model is a real-time thermal prediction model.

[0108] The process of predicting population distribution trends is as follows:

[0109] D1: Real-time collection of multi-source data 6-8 minutes before the current moment, and normalization of the collected data;

[0110] D2. Use the normalized data to predict the thermal prediction model to obtain the prediction results;

[0111] D3. Restore the predicted results to the actual thermal level. The thermal level restoration algorithm formula is:

[0112] ,in, is the actual thermal level, is the normalized thermal grade, is the global maximum crowd density value, is the global minimum crowd density value;

[0113] D4. Convert the predicted heat level into a color heat map, using different colors to represent the crowd density level;

[0114] S5. Dynamically adjust control strategies based on predicted crowd distribution trends and business objectives, where the business objectives are exhibit attention, visitor flow, and interactive experience;

[0115] The control strategy includes control based on exhibit attention, control based on visitor fluency, and control of interactive experience. Control based on exhibit attention includes optimizing traffic flow to popular exhibits and activating unpopular areas. The exhibit attention control method is as follows: if the heat level of a certain area is predicted to be ≥80% and the average dwell time is >15 minutes, adjustments will be made to optimize traffic flow to popular exhibits, triggering operations such as enhanced display effects, intelligent explanation scheduling, and space expansion. The enhanced display effect is achieved by adjusting the lighting brightness in the area (increase by 30%) through the IoT system and increasing the frequency of holographic projection or AR interactions (+2 times per hour) to attract more attention.

[0116] Intelligent tour dispatching automatically dispatches mobile tour guide robots to the area (or triggers voice broadcasts) to supplement manual tours, while also pushing detailed exhibit information to visitors' mobile phones via WiFi probes.

[0117] Space expansion involves temporarily removing non-critical exhibits from the periphery and widening aisles (e.g., adjusting the exhibition area layout through electric partitions) to alleviate congestion.

[0118] If the predicted heat level in a certain area is ≤20% and the average dwell time is less than 5 minutes, adjustments will be made to activate unpopular areas, triggering content updates and route guidance. Content updates automatically switch the display format of exhibits (such as rotating booths and dynamic screen content) or launch interactive tasks (such as scanning QR codes and answering questions to win souvenirs). Route guidance guides visitors to the area through the exhibition hall LED screen, ground projection arrows, or app push notifications.

[0119] The control method based on the visit fluency is: when the channel area density value is predicted to be ≥60 people / m 2 , and the flow speed is less than 0.5m / s, congestion warning and diversion control will be carried out, that is, physical diversion, digital guidance and flow control will be triggered. Physical diversion is to open emergency channels (such as automatic unlocking of electric doors), monitor channel flow through infrared sensors, and dynamically adjust the frequency of gate opening and closing; digital guidance is to mark red congested areas through in-hall broadcasts and APP real-time heat maps, and recommend detour routes (based on RFID tags to track visitor locations and plan the shortest path); flow control is to use RFID / face recognition to count the real-time number of people in the museum at the entrance. If the number exceeds 80% of the maximum capacity, the time-sharing reservation mechanism will be activated (such as limiting the speed of new visitors entering);

[0120] The interactive experience-based control method is as follows: if the number of people queuing in the interactive device area (such as the VR experience area) is predicted to be ≥5 and the waiting time is >10 minutes, dynamic scheduling of interactive equipment is carried out, triggering equipment expansion, time management and exclusion visualization operations; equipment expansion involves remotely starting a backup interactive terminal (such as a dormant VR device) and pushing a notification via the app that "device has been added and can be experienced immediately"; time management involves automatically shortening the duration of a single experience (such as from 8 minutes to 5 minutes) while increasing the number of sessions (such as 3 sessions per hour); queue visualization involves setting up an electronic screen in the area to display the real-time number of queue members and estimated waiting time, reducing visitor anxiety;

[0121] S6. Real-time monitoring of the changes in thermal distribution after dynamic adjustment, and optimization of the parameters of the thermal prediction model according to the changes in thermal distribution.

[0122] The process of parameter optimization of the thermal prediction model is as follows:

[0123] E1. Collect actual thermal data every two minutes using a thermal imaging camera, compare it with the predicted thermal map, and calculate the error value for each grid:

[0124]

[0125] If the regional average error is greater than 15%, the model optimization process is triggered;

[0126] E2. Add the latest actual data (including the adjusted control strategy labels) to the standardized dataset and use incremental learning to update the parameters of the convolutional neural network and spatiotemporal graph neural network;

[0127] E3. If the spatial attention weight of a certain area is continuously low but the actual flow of people is dense, increase the weight of the regional features in the spatial attention calculation, such as adjusting the bandwidth parameter of the Gaussian kernel function. ;

[0128] If the prediction error of a certain period in the time series is large, the temporal attention weight of this period is enhanced, such as adjusting the forget gate parameters of the gated recurrent unit;

[0129] E4. Use the sliding window method to verify the optimized model using the data of the past hour. If the prediction error is reduced to within 5%, save the new model; otherwise, repeat E1-E5.

[0130] Crowd density distribution data, behavioral data and environmental data are obtained through thermal imaging cameras, WiFi probes and infrared sensors, and a thermal distribution map is generated. By fusing and labeling discrete density points into a continuous thermal field data set, the crowd distribution characteristics are intuitively displayed. The convolutional neural network and the spatiotemporal graph neural network are combined to effectively model the spatiotemporal correlation of crowd flow. Based on the prediction results 6-8 minutes in advance, a closed loop of prediction-adjustment-verification is achieved. By dynamically adjusting the control strategy according to the predicted crowd distribution trend and business objectives, the guided route or interactive form can be dynamically adjusted according to the crowd density. The exhibit layout can also be adjusted in advance to reduce the number of round trips for visitors and improve the overall flow efficiency, thereby improving the display effect of the AI exhibition hall.

[0131] Based on the above-described AI exhibition hall control method based on crowd thermal distribution monitoring, the present invention also provides an AI exhibition hall control system based on crowd thermal distribution monitoring, such as Figure 5 As shown, it includes a data acquisition module, a heat map generation module, a data processing and integration module, a real-time prediction module, an execution control module, and a feedback optimization module. The data acquisition module uses multi-source sensor equipment to obtain the original data stream in the exhibition hall in real time. The original data stream includes the thermal coordinates of the crowd, the length of stay, the movement speed, and the environmental parameters. The multi-source sensor equipment generates the original data stream by deploying thermal imaging cameras, infrared sensors, WiFi probes, RFID tag readers, and environmental sensors.

[0132] The heat map generation module converts the sensing data into a visual heat distribution for real-time monitoring and analysis;

[0133] The data processing and integration module cleans, completes and integrates multi-source data to form a standardized data set of continuous thermal field;

[0134] The real-time prediction module uses a combination of convolutional neural networks and spatiotemporal graph neural networks to establish a real-time thermal prediction module, predicting crowd distribution trends 6-8 minutes in advance;

[0135] The execution control module dynamically adjusts exhibition hall resources based on prediction results and business goals. Exhibition hall resources include optimizing exhibit attention, flow, and interactive experience.

[0136] The feedback optimization module is used to build a closed loop of monitoring and optimization, monitor the changes in thermal distribution after dynamic adjustment in real time, and optimize the parameters of the thermal prediction model according to the changes in thermal distribution;

[0137] Through the collaboration of data collection module, heat map generation module, data processing and integration module, real-time prediction module, execution control module and feedback optimization module, the whole chain of intelligence from data collection to business implementation is realized. The modules are linked by data flow and combined with deep learning and dynamic control. It not only meets the real-time monitoring needs of the exhibition hall, but also dynamically adjusts the guided route or interactive form according to the density of people flow, thereby improving the display effect of the AI exhibition hall.

[0138] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An AI exhibition hall control method based on crowd thermal distribution monitoring is characterized by: The steps include: S1. Install multi-source sensor equipment on the ceiling, walls, or floor of the exhibition hall to build a fully covered perception network. The multi-source sensor equipment includes thermal imaging cameras, infrared sensors, WiFi probes, RFID tag readers, and environmental sensors. S2, using multi-source sensor equipment to capture crowd density distribution data, behavior data, and environmental data in real time and generate a thermal distribution map; S3. Filter the real-time data obtained by multi-source sensor equipment, then fuse the thermal distribution data, behavioral data, and environmental data, and mark the thermal characteristics of each area to form a continuous thermal field data set; S4. Build a real-time thermal prediction model by combining convolutional neural networks and spatiotemporal graph neural networks, and predict the crowd distribution trend within 6-8 minutes based on thermal characteristics. S5. Dynamically adjust control strategies based on predicted population distribution trends and business objectives; S6. Real-time monitoring of the changes in thermal distribution after dynamic adjustment, and optimization of the parameters of the thermal prediction model according to the changes in thermal distribution; The process of establishing a continuous thermal field dataset is as follows: B1. Identify and eliminate data that deviates from the normal range through outlier detection algorithms; B2. Use linear interpolation to fill in missing data and use a low-pass filter to filter the data; B3. Fusion of thermal distribution data, behavioral data, and environmental data based on timestamps to obtain a visual record containing multi-dimensional information; B4. Calculate the crowd density based on the number of people and area within the region, and use crowd density as a basic indicator of thermal characteristics to add a label to each region. B5. Normalize the numerical data and organize the normalized data into a structured table or database table structure; B6. Add metadata of the dataset to the structured table or database table structure to form a standardized dataset; In step B3, the fusion process is: B31. Synchronize thermal distribution data, behavioral data, and environmental data based on timestamps to make different types of data consistent in time. B32. Map thermal distribution data, behavioral data, and environmental data to a unified spatial coordinate system to make the data spatially comparable; B33. Map the characteristics of different types of data into color and intensity values in the thermal distribution map, unify the data scale through normalization, and display the data distribution density in color depth on a two-dimensional plane to form a visual record containing multi-dimensional information.

2. The AI exhibition hall control method based on crowd thermal distribution monitoring according to claim 1 is characterized by: The thermal imaging camera has a resolution of ≥640*480 pixels and a deployment spacing of ≤5 meters; the WiFi probe supports Bluetooth 5.0+WiFi6 protocols, with a positioning error of ≤1.5 meters; the infrared sensor has a coverage angle of ≥120° and a detection distance of ≤8 meters.

3. The AI exhibition hall control method based on crowd thermal distribution monitoring according to claim 2 is characterized by: The process of generating a heat map is as follows: A1. Divide the thermal camera's field of view into multiple pixel blocks, calculate the number of human targets within each pixel block, and calculate the density value based on the actual area corresponding to each pixel block; A2. Smoothing the discrete density points to generate a continuous density field; A3. Map density values to thermal levels from 0 to 100. A4. Use warm and cold gradients to color-code high-density and low-density distributions, and output thermal distribution maps through GIS rendering.

4. The AI exhibition hall control method based on crowd thermal distribution monitoring according to claim 3 is characterized by: In step S4, the real-time thermal prediction model establishment process is as follows: C1. Divide the normalized data into time windows and spatial grids to form input tensors; C2, process the input tensor through a multi-layer convolutional neural network; C3, calculate the spatial attention weight based on the input tensor processing results; C4. Combine the convolution results with time series data and use the gated recurrent algorithm to learn the dependencies of crowd flows in the time dimension. C5. Calculate the time attention weight based on the dependency of crowd flow in the time dimension; C6. Build a prediction model by combining temporal attention weight, spatial attention weight, spatial features, and temporal features; C7. The prediction model is trained using standardized data set data, and the parameters of the prediction model are optimized using the loss function. The optimized prediction model is a real-time thermal prediction model.

5. The AI exhibition hall control method based on crowd thermal distribution monitoring according to claim 4 is characterized in that: The process of predicting population distribution trends is as follows: D1: Real-time collection of multi-source data 6-8 minutes before the current moment, and normalization of the collected data; D2. Use the normalized data to predict the thermal prediction model to obtain the prediction results; D3. Restore the predicted results to the actual thermal level; D4. Convert the predicted heat level into a color heat map, using different colors to represent the crowd density level.

6. The AI exhibition hall control method based on crowd thermal distribution monitoring according to claim 5 is characterized by: The control strategy includes control based on exhibit attention, control based on visit flow, and control of interactive experience. Among them, control based on exhibit attention includes traffic optimization of hot exhibits and activation of unpopular areas. The control method based on the visit fluency is: when the channel area density value is predicted to be ≥60 people / m 2 , and when the flow speed is less than 0.5m / s, congestion warning and diversion control will be carried out; The control method based on interactive experience is: if it is predicted that the number of people queuing in the interactive device area is ≥5 and the waiting time is >10 minutes, dynamic scheduling of the interactive device will be performed.

7. The AI exhibition hall control method based on crowd thermal distribution monitoring according to claim 6 is characterized by: The process of parameter optimization of the thermal prediction model is as follows: E1. Collect actual thermal data every two minutes using a thermal imaging camera, compare it with the predicted thermal map, and calculate the error value for each grid. E2. Add the latest real-world data to the standardized dataset and use incremental learning to update the parameters of the convolutional neural network and spatiotemporal graph neural network. E3. If the spatial attention weight of a certain area is continuously low but the actual flow of people is dense, the weight of the regional features in the spatial attention calculation needs to be increased; If the prediction error of a certain period in the time series is large, the time attention weight of this period is enhanced; E4. Use the sliding window method to verify the optimized model using the data of the past hour. If the prediction error is reduced to within 5%, save the new model; otherwise, repeat E1-E3.

8. An AI exhibition hall control system based on crowd thermal distribution monitoring, based on the AI exhibition hall control method based on crowd thermal distribution monitoring according to any one of claims 1 to 7, characterized in that: include: The data acquisition module uses multi-source sensor equipment to obtain the original data stream in the exhibition hall in real time. The original data stream includes the thermal coordinates of the crowd, the length of stay, the movement speed and environmental parameters; The heat map generation module converts the sensor data into a visual heat distribution for real-time monitoring and analysis; The data processing and integration module cleans, completes, and integrates multi-source data to form a standardized data set of continuous thermal fields; The real-time prediction module uses a combination of convolutional neural networks and spatiotemporal graph neural networks to establish a real-time thermal prediction model, predicting crowd distribution trends 6-8 minutes in advance; The execution control module dynamically adjusts exhibition hall resources based on forecast results and business goals. The dynamic adjustment of exhibition hall resources includes optimizing exhibit attention, flow, and interactive experience. The feedback optimization module is used to build a closed loop of monitoring and optimization, monitor the changes in thermal distribution after dynamic adjustment in real time, and optimize the parameters of the thermal prediction model according to the changes in thermal distribution.

Images