Meeting place people flow analysis, statistics and prediction system

Abstract

The invention discloses a meeting place people flow analysis, statistics and prediction system, and relates to the technical field of intelligent people flow management, and the system comprises a face recognition verification terminal, a people flow statistics terminal, a prediction control terminal, an early warning prompt module, an emergency control module, a data storage module and a visualization module. Statistical errors caused by repeated entering and exiting are avoided, the number of people in the museum is calibrated in real time, the problems of identity confusion and insufficient precision existing in traditional manual statistics and single sensor counting are solved, the accuracy and reliability of people flow statistics are improved, and a high-quality data basis is provided for subsequent prediction and control. Historical activity data, real-time reservation information and environment variable factors are fused, a prediction model is constructed to pre-judge the peak period and the peak scale of people flow in advance, the management and control lagging risk caused by a traditional afterward response mechanism is reduced, and intelligent conversion from passive interception to active dredging is achieved in combination with a gate flow limiting and shunting guide strategy.

Description

Technical Field

[0001] This invention relates to the field of intelligent crowd management technology, and in particular to a venue crowd analysis, statistics and prediction system. Background Technology

[0002] A meeting venue is a specific place where meeting organizers and participants discuss and research issues, exchange information, communicate ideas, and achieve meeting transaction objectives; the flow of people refers to the total number of people passing through a specific area within a certain period of time, which is affected by a variety of factors.

[0003] Currently, due to the various dynamic factors in the process of managing the flow of people in the venue, traditional methods that rely on manual counting and basic sensing equipment cannot accurately identify the identity of people when conducting real-time statistics and predictions of the flow of people in the venue, resulting in large statistical errors in the repeated entry and exit behavior.

[0004] Therefore, a meeting venue crowd flow analysis, statistics and prediction system is proposed to solve the above problems. Summary of the Invention

[0005] The main objective of this invention is to provide a venue crowd flow analysis and prediction system to solve the problems mentioned in the background, such as the inability of manual counting and basic sensing equipment to accurately associate personnel identities and the large statistical errors in repeated entry and exit behaviors.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: a venue crowd analysis, statistics and prediction system, the system including a face recognition verification terminal, a crowd statistics terminal and a prediction control terminal, wherein the face recognition verification terminal, the crowd statistics terminal and the prediction control terminal are jointly provided with an early warning module; The facial recognition verification terminal is used to acquire personnel image data in real time through biometric collection devices deployed at the entrance and exit of the venue, and compare and verify it with the pre-stored identity database to distinguish personnel identity categories and assign unique identification IDs, thereby avoiding statistical errors caused by repeated entry and exit. The people flow statistics terminal is used to calculate the real-time number of people in the venue based on entrance verification data, exit verification data and seat occupancy status data inside the venue, and generate classified statistical reports according to personnel identity categories and entry and exit time periods; The predictive control terminal is used to integrate historical pedestrian flow data, real-time reservation data and environmental variable factors, predict peak pedestrian flow periods and peak numbers through machine learning algorithms, and dynamically trigger hierarchical traffic control strategies based on preset thresholds. The early warning module is used to issue visual and audio alarms via LED display and voice broadcaster when the flow threshold is reached or when the flow is predicted to exceed the limit.

[0007] Preferably, the face recognition verification terminal includes an identity registration module, a real-time verification module, and an ID management module; The identity registration module is used to pre-store the facial feature templates of teachers and students in the campus database and to provide a temporary registration channel for visitors. The real-time verification module is used to match the collected facial images with feature templates. When the similarity is ≥95%, the identity is determined to be legitimate; otherwise, passage is denied. The ID management module generates a unique identification code for each verified person and records the first entry timestamp. Repeated entry and exit behaviors with the same identification code are marked as associated events.

[0008] Preferably, the people flow statistics terminal includes a multi-source calibration module and a classification statistics module; The multi-source calibration module monitors the occupancy status of seats in the venue in real time through a pressure sensor array and calculates the real-time number of people in the venue: N. current =N in -N out +C adjust ; Where N current To show the real-time number of people in the venue, N in N represents the cumulative number of people who have passed the entrance verification. out C represents the cumulative number of people who have departed after passing export verification. adjust Calibration factor; When the deviation between the total number of occupied seats and the number of people verified is ≥5%, the population statistics will be adjusted based on the seat data. The classification and statistics module generates structured reports according to the following dimensions: Personnel category breakdown: the number and proportion of students, teachers, and visitors; Time period distribution: Inbound and outbound traffic during the critical periods of 30 minutes before the start of the event and 15 minutes after the start of the event; Duration of stay: The average duration of stay is calculated based on the entry and exit timestamps.

[0009] Preferably, the prediction control terminal includes a data fusion module, an algorithm prediction module, and a dynamic execution module; The data fusion module is used to integrate three types of input sources: Historical data: Peak number of participants, peak time periods, and participant composition for similar activities over the past 12 months; Real-time data: Number of registered participants, number of attendees already present, and remaining time for the event; Environmental variables: date type, competitive activity identifier, weather index; The algorithm prediction module uses a random forest regression model and outputs two core prediction results: P peak =w h ·H ratio +w r·R ratio +w e ·E score ; Where P peak To predict the percentage of peak capacity, w h H is a weighting factor for historical data. avg H is the historical peak number of people. ratio The historical peak ratio, w r R is a weighting factor for real-time data. ratio w is the ratio of the number of reservations to the total capacity. e E is the environmental variable weighting factor. score Environmental rating, ranging from 0 to 1; The weight factors are obtained through training on historical datasets. The training method is as follows: Multiple linear regression was used to fit the actual peak number of participants and predicted values ​​for similar activities over the past 12 months. The weight allocation was iteratively optimized with the objective function of minimizing the root mean square error, ultimately yielding the desired result. h The value range is [0.4, 0.6], w r For [0.3, 0.5], w e The expression is [0.1, 0.3] and satisfies w h +w r +w e =1; The output includes: Peak hours: time windows accurate to ±5 minutes; Maximum peak capacity: expressed as a percentage of rated capacity; The dynamic execution module links the hardware devices according to the prediction results: when the predicted peak exceeds 80% of the rated capacity, an early warning is activated; when it exceeds 95%, flow limiting control is triggered.

[0010] Preferably, the hierarchical control strategy of the dynamic execution module includes: Level 1 warning: When the number of people in the venue at any given time is greater than or equal to 70% of the rated capacity, the LED screen will display "Less than 30% of seats remain" and a voice prompt will be activated. Level 2 flow control: When the number of people in the venue at any given time is greater than or equal to 90% of the rated capacity, the turnstiles will switch to flow control mode, allowing only those with special authorization codes to enter, including staff and VIPs; Traffic diversion guidance: When the predicted peak exceeds 100% of the rated capacity, traffic diversion suggestions are automatically pushed to the management end, including increasing the number of guides and opening online live broadcast channels.

[0011] Preferably, the system further includes an emergency control module, which, when triggered by the fire sensor and the manual emergency stop button, performs the following actions: forcibly opening all exit gates and locking the entrance passage; The entire venue will be equipped with voice-guided evacuation instructions, broadcasting the shortest escape route. Send real-time location heatmaps and the number of people stranded to the safety management platform.

[0012] Preferably, the system includes a data storage module, comprising: Encrypted storage unit: The facial feature template is encrypted with AES-256 and only the comparison result and timestamp are stored; Tiered storage unit: Local servers retain frequently accessed data within 3 months, while cloud platforms retain historical data for more than 1 year; Operation log unit: Records the time points of all control commands, threshold modifications, and abnormal events.

[0013] Preferably, the system includes a visualization module, providing: Real-time dynamic heat map: The seating area is rendered in three colors: red (≥80% occupancy), yellow (50%-80%), and green (<50%). Forecast trend curve: Overlays historical peak curves with the current forecast curve; Multi-terminal support: The management computer interface synchronizes data with the mobile APP, and supports exporting CSV format reports.

[0014] Preferably, the system hardware layer includes: Biometric data acquisition terminals: deployed at main entrances, side entrances, and emergency exits, integrating near-infrared liveness detection cameras; Smart turnstiles: physically linked with facial recognition terminals, supporting three operating modes: Verification and access mode: The gate opens after successful facial recognition; Traffic limiting mode: Only responds to the identification codes of people who have already left the venue; Emergency release mode: Automatic gate opening upon power failure; Distributed sensor network: A piezoresistive sensor is embedded under each seat, with a sampling frequency of 1-5Hz. This frequency range is set based on the typical performance of the piezoresistive sensor and the actual needs of people flow statistics. The sensor data is processed by sliding window mean filtering.

[0015] Preferably, the linkage logic between the early warning module and the hardware layer is as follows: When the prediction control terminal issues an early warning command, the entrance LED screen is activated to display the predicted number of remaining seats; When the pedestrian flow statistics terminal detects regional clustering and the population density in a single block is ≥2 people / ㎡, the voice broadcaster in that area will be triggered to play a diversion prompt. Emergency control commands take precedence over other commands, directly covering the control of the turnstiles and the content of announcements.

[0016] The present invention has the following beneficial effects: 1. In this invention, when conducting real-time statistics of people flow in a venue, a unique identification ID of each person is associated through an identity verification mechanism to avoid statistical errors caused by repeated entry and exit. Based on the multi-source data fusion calibration, the number of people in the venue in real time can be rectified, which can solve the problems of identity confusion and insufficient accuracy in traditional manual statistics and single sensor counting, improve the accuracy and reliability of people flow statistics, and provide a high-quality data foundation for subsequent prediction and control.

[0017] 2. In this invention, when managing the dynamic flow of people in a venue, a predictive model is constructed by integrating historical event data, real-time reservation information and environmental variable factors to predict peak periods and peak scales of people in advance. This enables the system to proactively trigger tiered warnings before excessive crowding, reducing the risk of delayed control caused by traditional post-event response mechanisms. Combined with gate flow restriction and diversion guidance strategies, an intelligent transformation from passive interception to proactive guidance is achieved.

[0018] 3. In this invention, when dealing with sudden changes in crowd flow in a venue, the system generates a regional heat map in real time and identifies risk points of crowd gathering through the coordinated response of the visual dynamic monitoring and emergency control module. This enables the system to quickly initiate evacuation guidance and channel control in emergency situations, avoiding safety hazards caused by delays in human decision-making and improving the safety redundancy and emergency response efficiency of large venues. Attached Figure Description

[0019] Figure 1 This is a diagram illustrating the architecture of a meeting venue crowd flow analysis, statistics, and prediction system according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Specific embodiment: A venue crowd flow analysis, statistics and prediction system, the system includes a face recognition verification terminal, a crowd flow statistics terminal and a prediction control terminal, the face recognition verification terminal, the crowd flow statistics terminal and the prediction control terminal are jointly equipped with an early warning module; the face recognition verification terminal is used to acquire personnel image data in real time through biometric collection devices deployed at the entrance and exit of the venue, and compare and verify it with a pre-stored identity database, distinguish personnel identity categories and assign unique identification IDs, so as to avoid statistical errors caused by repeated entry and exit; The people flow statistics terminal is used to calculate the real-time number of people in the venue based on entrance verification data, exit verification data, and seat occupancy status data inside the venue, and to generate classified statistical reports according to personnel identity categories and entry and exit time periods; The predictive control terminal is used to integrate historical traffic flow data, real-time reservation data and environmental variable factors, and predict peak traffic periods and peak numbers through machine learning algorithms, and dynamically trigger tiered traffic control strategies based on preset thresholds. The early warning module is used to issue visual and audio alarms via LED display and voice broadcaster when the flow threshold is reached or when the limit is predicted to be exceeded.

[0022] The facial recognition verification terminal includes an identity registration module, a real-time verification module, and an ID management module; The identity registration module is used to pre-store the facial feature templates of teachers and students in the campus database and provide a temporary registration channel for visitors; the real-time verification module is used to match the collected facial images with the feature templates. When the similarity is ≥95% and Confidence ≥0.85, the identity is determined to be legitimate; otherwise, passage is denied. The identity verification and anti-counterfeiting formula is: Confidence = α·S texture +β·S thermal ; Where Confidence is the confidence level for liveness detection, α is the texture feature weight, and S... texture For facial micro-movement matching, β is the weight of thermal radiation features, and S thermal This represents the in vivo response value for near-infrared thermal imaging. The ID management module generates a unique identification code for each verified person and records the first entry timestamp. Repeated entry and exit behaviors with the same identification code are marked as associated events.

[0023] The pedestrian flow statistics terminal includes a multi-source calibration module and a classification statistics module; The multi-source calibration module monitors the occupancy status of seats in the venue in real time through a pressure sensor array and calculates the real-time number of people in the venue: N current =N in -N out +C adjust ; Where N current To show the real-time number of people in the venue, N in N represents the cumulative number of people who have passed the entrance verification. out C represents the cumulative number of people who have departed after passing export verification. adjust Calibration factor; When the deviation between the total number of occupied seats and the number of people verified is ≥5%, the population statistics will be adjusted based on the seat data. The classification and statistics module generates structured reports based on the following dimensions: Personnel category breakdown: the number and proportion of students, teachers, and visitors; Time period distribution: Inbound and outbound traffic during the critical periods of 30 minutes before the start of the event and 15 minutes after the start of the event; Duration of stay: The average duration of stay is calculated based on the entry and exit timestamps.

[0024] The predictive control module includes a data fusion module, an algorithm prediction module, and a dynamic execution module. The data fusion module is used to integrate three types of input sources: Historical data: Peak number of participants, peak time periods, and participant composition for similar activities over the past 12 months; Real-time data: Number of registered participants, number of attendees already present, and remaining time for the event; Environment variables: I. Date type: Weekday = 0.7, Weekend = 1.0, Holiday = 1.0; II. Competitive Activity Identifier: The identifier = 1.0 when there are concurrent competitive activities, otherwise = 0.8; The competitive event identifier refers to the number of large-scale events held at other venues within a 1-kilometer radius during the same period. A high score coefficient is triggered when the number is ≥3. III. Weather Index: Index = 0.8 when rainfall > 10mm, otherwise = 1.0; Environmental rating E score The calculation formula is: Where K is a normalization constant of 1.5, determined by fitting historical extreme values; The algorithm prediction module uses a random forest regression model and outputs two core prediction results: P peak =w h ·H ratio +w r ·R ratio +w e ·E score ; Where P peak To predict the percentage of peak capacity, w h H is a weighting factor for historical data. avg H is the historical peak number of people. ratio The historical peak ratio, w r R is a weighting factor for real-time data. ratio w is the ratio of the number of reservations to the total capacity. e E is the environmental variable weighting factor. score Environmental rating, ranging from 0 to 1; The weight factors are obtained through training on historical datasets. The training method is as follows: Multiple linear regression was used to fit the actual peak number of participants and predicted values ​​for similar activities over the past 12 months. The weight allocation was iteratively optimized with the objective function of minimizing the root mean square error, ultimately yielding the desired result. h The value range is [0.4, 0.6], w r For [0.3, 0.5], w e The expression is [0.1, 0.3] and satisfies w h +w r +w e =1; The output includes: Peak hours: time windows accurate to ±5 minutes; Maximum peak capacity: expressed as a percentage of rated capacity; The dynamic execution module coordinates with hardware devices based on the prediction results: when the predicted peak exceeds 80% of the rated capacity, an early warning is activated; when it exceeds 95%, current limiting control is triggered.

[0025] The hierarchical control strategy for the dynamic execution module includes: T warn =0.7C max ; T limit =0.9C max ; T divert =C max ; Where T warn The threshold for Level 1 warning is T. limit T is the secondary current limiting threshold. divert C is the threshold for initiating traffic splitting. max This is the current rated capacity of the venue; Level 1 Warning: When the number of people in the venue at any given time is greater than or equal to T warn When the number of people in the venue is greater than 90% of the rated capacity, the turnstiles will switch to flow control mode, allowing only those with special authorization codes to enter, including staff and VIPs. Triage guidance: When the predicted peak value is ≥ C max Automatically push traffic diversion suggestions to the management end, including increasing the number of guides and opening online live streaming channels.

[0026] The system also includes an emergency control module that executes actions triggered by fire sensors and the manual emergency stop button: Force all exit gates to open and lock the entrance passage; The entire venue will be equipped with voice-guided evacuation instructions, broadcasting the shortest escape route. Send real-time location heatmaps and the number of people stranded to the safety management platform; The formula for generating a heatmap is: Where H value W represents the regional thermal value. i Let D be the detection weight of the i-th sensor. i This is the distance between the sensing unit and the geometric center of the target area.

[0027] The system has a data storage module, including: Encrypted storage unit: The facial feature template is encrypted with AES-256 and only the comparison result and timestamp are stored; Tiered storage unit: Local servers retain frequently accessed data within 3 months, while cloud platforms retain historical data for more than 1 year; Operation log unit: Records the time points of all control commands, threshold modifications, and abnormal events.

[0028] The system includes a visualization module, providing: Real-time dynamic heatmap: Press Color render The rules use three colors to render the seating area: red (≥80%), yellow (50%-80%), and green (<50%). Color render To render colors, H value H is the real-time thermal value. max This represents the current maximum theoretical heat value of the venue. Forecast trend curve: Overlays historical peak curves with the current forecast curve; Multi-terminal support: The management computer interface synchronizes data with the mobile APP, and supports exporting CSV format reports.

[0029] The system hardware layer includes: Biometric data acquisition terminal: deployed at the main entrance, side entrances, and emergency exits, integrating near-infrared liveness detection cameras; Intelligent gate unit: Physically linked with facial recognition terminal, supporting three working modes: Verification and access mode: The gate opens after successful facial recognition; Traffic limiting mode: Only responds to the identification codes of people who have already left the venue; In flow control mode, the RFID reader built into the gate verifies the electronic identification code worn by the personnel in real time. When the departure record associated with the identification code is detected in the local cache database, the gate opening command is triggered. The special authorization code is dynamically generated by the management terminal, with a validity period of 30 minutes. It is pushed to the user's mobile terminal in the form of a QR code. The gate scanner verifies the encrypted digital signature and then allows passage. Emergency release mode: the gate opens automatically when the power is off. Distributed sensor network: A piezoresistive sensor is embedded under each seat, with a sampling frequency of 1-5Hz. This frequency range is set based on the typical performance of the piezoresistive sensor and the actual needs of people flow statistics. The sensor data is processed by sliding window mean filtering.

[0030] The linkage logic between the early warning module and the hardware layer is as follows: When the prediction control terminal issues an early warning command, the entrance LED screen is activated to display the predicted number of remaining seats; When the pedestrian flow statistics terminal detects regional clustering and the population density in a single block is ≥2 people / ㎡, the voice broadcaster in that area will be triggered to play a diversion prompt. Regional clustering: Risk zone N represents the regional risk status. zone For the real-time number of people in the target area, A zone ρ is the area of ​​the region. safe The safe density threshold; Emergency control commands take precedence over other commands, directly covering the control of the turnstiles and the content of announcements.

[0031] The system operates as follows: Step 1: Identity Verification and Data Collection The system deploys biometric recognition terminals at each entrance to the venue, capturing facial images of attendees using near-infrared liveness detection cameras. The terminals compare the acquired facial features with a pre-stored identity database in real time. When the similarity reaches a set threshold, the identity is deemed legitimate, and a unique identification ID is assigned to the verified individual. For visitors, the system provides a temporary registration channel, allowing administrators to enter facial information and authorize temporary access. This process simultaneously records timestamps accurate to milliseconds, establishing a foundational data chain for subsequent statistical analysis.

[0032] Step Two: Multi-Source Pedestrian Flow Statistics and Calibration Based on the number of people entering the venue generated from entrance verification data and the number of people leaving the venue generated from exit verification data, the system calculates the initial number of people in the venue. Simultaneously, an array of pressure sensors embedded under the seats monitors seat occupancy in real time. When the deviation between the sensor data and the verification statistics exceeds a tolerance threshold, the system automatically and dynamically calibrates the flow statistics based on the physical sensor data. The calibrated data is then structured and categorized by personnel type and time period distribution, generating real-time statistical reports.

[0033] Step 3: Integrating Prediction and Early Warning Trigger The system accesses historical databases, real-time reservation data, and environmental variables, inputting them into a machine learning prediction model. After analyzing data correlations, the model outputs two core results: peak flow window and peak passenger volume prediction. When the predicted value exceeds a preset capacity ratio, the system activates an early warning mechanism, dynamically displaying the remaining seat number on the entrance LED screen and providing voice prompts for passenger flow management.

[0034] Step 4: Dynamic Flow Classification Control Based on the comparison between the real-time number of people in the venue and the preset threshold, the system executes three response strategies: Level 1 warning, Level 2 flow restriction, and flow diversion guidance. All turnstiles support three working modes: verification and access mode for regular gate opening, flow restriction mode for blocking new entrants, and emergency release mode forcibly opening the channel.

[0035] Step 5: Heat Map Monitoring and Emergency Response The system generates a heat map of the entire venue based on distributed sensor network data, and displays the area density intuitively through a three-color rendering mechanism. When the personnel density in a specific area exceeds the safety threshold, the system automatically triggers the voice guidance device in that area to play evacuation prompts. When fire sensors and emergency stop buttons are activated, the system immediately initiates the emergency protocol: forcibly opening all exit turnstiles, locking entrance passages, and broadcasting the optimal escape route through the voice system.

[0036] Step Six: Data Storage and Decision Support All verification records, statistical data, and operation logs employ a tiered encryption storage strategy: facial feature templates are encrypted, retaining only the comparison results; frequently accessed data is stored on a local server; and all historical data is synchronized to the cloud platform. Administrators can view real-time heatmaps, predicted trend curves, and categorized statistical reports through a visualization platform, which supports multi-terminal access and data export, providing a basis for venue resource allocation decisions.

[0037] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A system for analyzing, statistically predicting, and forecasting crowd flow at a meeting venue, characterized in that: The system comprises a face recognition verification end, a people flow statistics end and a prediction control end, and the face recognition verification end, the people flow statistics end and the prediction control end are provided with a warning prompt module; The face recognition verification end is used for acquiring personnel image data in real time through a biological feature acquisition device arranged at the entrance and exit of the venue, and comparing and verifying the acquired personnel image data with a pre-stored identity database, so as to distinguish the identity of the personnel and assign a unique identification ID, thereby avoiding statistical errors caused by repeated entry and exit; The people flow statistics end is used for calculating the number of people in the venue in real time based on the entrance verification data, the exit verification data and the seat occupancy state data in the venue, and generating a classified statistical report according to the identity of the personnel and the entry and exit time period; The prediction control end is used for fusing historical people flow data, real-time reservation data and environmental variable factors, predicting the people flow peak time period and the peak number of people through a machine learning algorithm, and dynamically triggering a hierarchical flow control strategy according to a preset threshold value; The warning prompt module is used for issuing visual and audio alarms when the flow threshold value is reached or the predicted number of people exceeds the limit through an LED display screen and a voice broadcaster.

2. The system according to claim 1, wherein: The face recognition verification end comprises an identity registration module, a real-time verification module and an ID management module; The identity registration module is used for pre-storing the face feature templates of teachers and students in a campus database, and providing a temporary registration channel for external visitors; The real-time verification module is used for performing similarity matching between the collected face image and the feature templates, and determining that the identity is legal when the similarity is greater than or equal to 95%, otherwise, the passage is refused; The ID management module generates a unique identification code for each verified personnel, and records the first entry time stamp, and marks the repeated entry and exit behavior of the same identification code as an associated event.

3. The system of claim 1, wherein: The people flow statistics end comprises a multi-source calibration module and a classified statistical module; The multi-source calibration module monitors the seat occupancy state of the venue in real time through a pressure sensor array, and calculates the number of people in the venue in real time: N current = N in - N out + C adjust ; where N current is the real-time number of people in the museum, N in is the cumulative number of people entering through the entrance, N out is the cumulative number of people leaving through the exit, C adjust is the calibration factor; When the deviation between the total number of seat occupancy and the number of people verified is greater than or equal to 5%, the people flow statistics value is corrected based on the seat data; The classified statistical module generates a structured report according to the following dimensions: Personnel category proportion: the number and proportion of students, teachers and visitors; Time period distribution: the entry and exit flow in the key time period of 30 minutes before the start of the activity and 15 minutes after the start of the activity; Residence time: the average residence time is calculated based on the entry and exit time stamps.

4. The system of claim 1, wherein: The prediction control end comprises a data fusion module, an algorithm prediction module and a dynamic execution module; The data fusion module is used for integrating three types of input sources: Historical data: the peak number of people, the peak time period and the personnel composition of the same type of activity in the past 12 months; Real-time data: the number of people who have entered the venue, the number of people who have entered the venue and the remaining time; Environmental variables: date type, competitive activity identifier, weather index; The algorithm prediction module adopts a random forest regression model, and outputs two core prediction results: P peak = w h · H ratio + w r · R ratio + w e · E score ; where P peak is the predicted percentage of peak capacity, w h is the historical data weight factor, H avg is the historical peak number, H ratio is the historical peak ratio, w r is the real-time data weight factor, R ratio is the ratio of reservations to total capacity, w e is the environmental variable weight factor, E score is the environmental score, ranging from 0-1; The weight factor is obtained by training the historical data set, and the training method is: A multiple linear regression fit is performed on the actual peak numbers of similar events in the past 12 months versus the predicted values, with the objective function being to minimize the root mean square error, and the weights are iteratively optimized. The final w h is in the interval [0.4, 0.6], w r is in the interval [0.3, 0.5], w e is in the interval [0.1, 0.3] and satisfies w h + w r + w e = 1. The output includes: People flow peak time period: the time window is accurate to ± 5 minutes; Maximum peak number of people: expressed as a percentage of the rated capacity; The dynamic execution module links hardware devices according to the prediction result: when the predicted peak value exceeds 80% of the rated capacity, a warning prompt is activated, and when it exceeds 95%, flow control is triggered.

5. The system of claim 4, wherein: The hierarchical control strategy of the dynamic execution module includes: First-level warning: when the real-time number of people in the library is greater than or equal to 70% of the rated capacity, the LED screen displays "less than 30% of the remaining seats" and a voice prompt is started; Second-level flow control: when the real-time number of people in the library is greater than or equal to 90% of the rated capacity, the gate switches to flow control mode, allowing only personnel with special authorization codes, including staff and VIPs, to enter; Shunt guidance: when the predicted peak value exceeds 100% of the rated capacity, automatically push the shunt suggestion to the management end, including adding guide personnel and opening the online live channel.

6. The system of claim 1, wherein: The system also includes an emergency control module that executes when the fire sensor and manual emergency stop button are triggered: Forcibly open all exit gates and lock the entrance channel; Start the whole venue voice evacuation guidance, broadcast the shortest escape route; Send the real-time positioning heat map and the number of people to the security management platform.

7. The system of claim 1, wherein: The system has a data storage module, including: Encrypted storage unit: AES-256 encryption of face feature templates, only storing comparison results and timestamps; Hierarchical storage unit: local server retains high-frequency access data within 3 months, and cloud platform retains historical data for more than 1 year; Operation log unit: records the time nodes of all control instructions, threshold modifications, and abnormal events.

8. The system of claim 1, wherein: The system also has a visualization module that provides: Real-time dynamic heat map: render seat areas in red for ≥80% occupancy, yellow for 50%-80%, and green for <50%; Prediction trend curve: superimposed display of historical peak curve and current prediction curve; Multi-terminal support: management end computer interface synchronizes mobile APP data, supports CSV format report export.

9. The system of claim 1, wherein: The system hardware layer includes: Biometric feature collection terminal: deployed at main entrances, side entrances, and emergency exits, integrated with near-infrared live detection cameras; Intelligent gate group: physically linked with face recognition terminals, supporting three working modes: Verification passage mode: gate opens after successful face comparison; Flow control mode: only responds to identification codes of people who have left; Emergency release mode: automatically opens the gate when powered off; Distributed sensing network: pressure resistance sensors are buried under each seat, with a sampling frequency of 1-5Hz, which is based on the typical performance of pressure resistance sensors and the actual demand of people flow statistics, and sensor data is processed using sliding window mean filtering.

10. The system of claim 1, wherein: The linkage logic of the warning prompt module and the hardware layer is: When the prediction control end issues a warning instruction, activate the entrance LED screen to display the predicted number of remaining seats; When the people flow statistics end detects regional aggregation, with a single block personnel density ≥2 people / ㎡, trigger the regional voice broadcaster to play the shunt prompt; Emergency control instructions take priority over other instructions and directly override gate control and broadcast content.

Images