Campus radicalization prevention system and method

Abstract

The application relates to the technical field of campus safety management, and discloses a campus overactive behavior prevention system and method, which comprises an Internet of Things sensing system, an intelligent analysis database and an intelligent pushing system; the Internet of Things sensing system is used for collecting multi-source data; the intelligent analysis database is used for completing multi-source data fusion and privacy protection, and comprises a time series database used for storing environment data, a graph database used for constructing a behavior correlation graph, a data processing layer based on federal learning and a privacy protection layer based on differential privacy and a block chain; a multi-modal fusion model is arranged in the intelligent pushing system; the multi-modal fusion model calculates and outputs an overactive behavior risk index based on the multi-source data, and triggers a hierarchical alarm. The application can consider privacy protection, real-time early warning and accurate identification, can effectively solve the problems of concealment, delay and privacy disputes in campus overactive behavior prevention and control, and provides a standardized solution for constructing a zero overactive behavior, high-energy-efficiency intelligent campus.

Description

Technical Field

[0001] This invention relates to the field of campus safety management technology, specifically to a system and method for preventing excessive behavior on campus. Background Technology

[0002] The frequent occurrence of extreme behaviors on campuses is seriously impacting students' physical and mental health. Traditional methods for preventing extreme behavior on campus mainly construct a comprehensive prevention and control system from four dimensions: education, systems, environment, and intervention. Examples include anti-extreme behavior thematic education and monitoring and reporting systems. These measures have achieved some results. While traditional preventative measures have played a role in controlling extreme behavior on campuses, their limitations are becoming increasingly apparent in the face of the current severe reality of frequent and egregious incidents. These limitations are mainly reflected in significant problems with the identification, early warning, and storage of extreme behaviors.

[0003] Specifically: First, there's the issue of the concealment of incident locations. Many violent incidents occur in areas difficult to cover by surveillance cameras, such as restrooms and stairwells, making timely detection difficult. A report from the National Center for Education Statistics (NCES) indicates that up to 70% of violent incidents on school campuses occur in blind spots, such as restrooms and stairwells. Second, there's the issue of the timeliness of behavior recognition and capture. Traditional methods have significant delays in monitoring and capturing violent behavior, failing to achieve rapid response. The average intervention delay for concealed violent behavior is as long as 3-6 months (Yale University study), significantly increasing the risk of psychological problems such as depression and self-harm by 300%. Third, there's the conflict between privacy protection and information access. While expanding the surveillance scope can improve prevention, it inevitably increases the risk of privacy breaches. Fourth, there's the issue of the accuracy of behavioral warnings. Existing technologies struggle to provide accurate warnings, leading to delays in the detection of violent behavior. For example, most existing systems rely on single-modal data (such as video surveillance), resulting in high false alarm rates and an inability to cover multi-dimensional scenarios. The aforementioned problems directly result in a serious lack of means to monitor, capture, and issue early warnings for extreme behaviors on campus, making it difficult to effectively curb such behaviors and severely hindering the advancement of campus safety construction. It is urgent to solve these problems through innovative technological means. Summary of the Invention

[0004] This invention aims to provide a system and method for preventing aggressive behavior on campus, which can balance privacy protection, real-time early warning and accurate identification. It can effectively solve the problems of concealment, delay and privacy disputes in the prevention and control of aggressive behavior on campus, and provide a standardized solution for building a smart campus with zero aggressive behavior and high efficiency.

[0005] To achieve the above objectives, the present invention provides the following basic solution.

[0006] Option 1

[0007] The campus aggressive behavior prevention system includes an Internet of Things (IoT) sensing system, an intelligent analysis database, and an intelligent push system. The IoT sensing system is used to collect multi-source data, including environmental data and biometric data. The environmental data includes carbon concentration and temperature and humidity, while the biometric data includes biological temperature, biological movement, and voiceprint data.

[0008] The intelligent analysis database is used to complete multi-source data fusion and privacy protection, including a time-series database for storing environmental data, a graph database for constructing behavioral correlation graphs, a data processing layer based on federated learning, and a privacy protection layer based on differential privacy and blockchain.

[0009] The intelligent push system is equipped with a multimodal fusion model, which calculates and outputs an overreaction risk index based on multi-source data and triggers tiered alarms.

[0010] Furthermore, the IoT sensing system includes an environmental monitoring unit and a biometric sensing unit; the environmental monitoring unit includes a CO2 sensor for capturing changes in carbon concentration within the sensing area, LoRa temperature and humidity nodes and a thermal imaging camera for capturing changes in environmental temperature and humidity and biological temperature within the sensing area, respectively; the biometric sensing unit includes an optical camera and a millimeter-wave radar for collecting biological motion information within the sensing area, and a directional microphone array for collecting voiceprint information.

[0011] Furthermore, the environmental monitoring unit identifies target actions based on biological motion information using the YOLOv7 algorithm, and extracts target semantics based on voiceprint information using the BERT model; the target actions include pushing, blocking, and pulling actions; the target semantics include abusive semantics.

[0012] Furthermore, the data processing layer implements localized data processing at edge nodes through a federated learning framework, and combines a population dynamics model to quantify the air mass retention index and metabolic rate mutation threshold based on carbon concentration.

[0013] Furthermore, the multimodal fusion model is an LSTM-DNN hybrid model; the LSTM layer predicts the CO2 concentration baseline and dynamically calibrates the threshold; when dynamically calibrating the threshold, a basic threshold is first generated based on the venue capacity and ventilation efficiency, and then the threshold is calibrated based on real-time data; the DNN layer fuses multimodal features and outputs an overreaction risk index.

[0014] Furthermore, the IoT sensing system adopts a modular structure, with the environmental monitoring unit, biometric sensing unit, and edge computing unit forming an integrated module for easy deployment.

[0015] Furthermore, the placement of the integrated modules is determined by combining atmospheric diffusion models.

[0016] Furthermore, the integration module collects multi-source data, preprocesses the multi-source data based on the edge computing unit, and sends it to the regional gateway after AES-256 encryption in fragments. The regional gateway aggregates data from multiple nodes and verifies the integrity of the fragments. Through federated learning, it aggregates and updates the data locally and uploads it to the cloud after compression. The cloud decrypts the data and calls the intelligent analysis database and intelligent push system to store and analyze the data.

[0017] Furthermore, the risk index for excessive behavior ranges from 0 to 1; the intelligent push system is linked with the campus's internal information system; the information system includes an LCD screen, broadcast system, buzzer, and security terminal.

[0018] The tiered alarm includes: when the risk index of aggressive behavior is between 0.6 and 0.8, a level 1 alarm is triggered, which activates the LCD screen warning and directional broadcast; when the risk index of aggressive behavior is greater than 0.8, a level 2 alarm is triggered, which links the buzzer and 5G NR technology to push the alarm information to the security terminal with low latency, and simultaneously adjusts the fresh air system to manage carbon concentration.

[0019] Option 2

[0020] The method for preventing aggressive behavior on campus involves applying the aggressive behavior prevention system described in Solution 1 to prevent aggressive behavior on campus; it includes the following steps:

[0021] The system employs an Internet of Things (IoT) sensing system to collect multi-source data, including environmental data and biometric data.

[0022] Multi-source data fusion and privacy protection are accomplished by an intelligent analysis database;

[0023] The multimodal fusion model of the intelligent push system calculates and outputs an overreaction risk index based on multi-source data, and triggers tiered alerts.

[0024] The working principle and advantages of this invention are as follows:

[0025] This invention relates to a system and method for preventing aggressive behavior on campus, which balances privacy protection, real-time early warning, and accurate identification. It effectively addresses the issues of concealment, delay, and privacy disputes in preventing aggressive behavior on campus, providing a standardized solution for building a smart campus characterized by zero aggressive behavior and high efficiency. The key points are:

[0026] First, this solution innovatively proposes a new dimension for detecting aggressive behavior based on carbon monitoring and temperature changes. Utilizing subtle changes in human body temperature and environmental carbon dioxide concentrations as biomarkers for detecting aggressive behavior, and supplemented by voiceprint analysis, optical imaging, thermal imaging, and motion recognition technologies, combined with artificial intelligence algorithms, this system can monitor the situation on campus in real time. By analyzing students' body temperature and behavioral patterns, it can predict and identify potential aggressive behaviors, enabling early warning and intervention, thereby improving campus safety management and protecting students' physical and mental health.

[0027] Furthermore, traditional monitoring of aggressive behavior on campus often focuses on direct observation of students' behavior and speech. This direct observation method is highly subjective and prone to delays. This system, however, takes a data-driven approach, introducing data such as carbon concentration and temperature as monitoring indicators. Changes in carbon concentration are closely related to the gathering and activities of people. By quantifying this value, it can indirectly reflect whether there are abnormal gatherings and intense activities on campus, providing a new perspective for the prevention of aggressive behavior on campus. Simultaneously, abnormal changes in body temperature may also indicate intense physical conflict (such as adrenaline surges causing body temperature fluctuations). By integrating and analyzing these characteristics, this system can objectively analyze and promptly capture the risk of aggressive behavior.

[0028] Secondly, the IoT sensing system in this solution adopts a modular structure, which can be easily deployed in various corners of the campus, effectively solving the sensing problem in blind spots such as toilets and stairwells, achieving full-scene coverage and facilitating linkage with existing campus monitoring systems. Combined with an intelligent analysis database and intelligent push system, this solution can achieve a complete closed loop of "sensing-analysis-response," comprehensively preventing excessive behavior on campus.

[0029] Third, this solution effectively addresses privacy disputes in the prevention and control of aggressive behavior on campus, fully protecting the privacy rights of students and faculty. Specifically, in the data processing phase, this solution incorporates a data processing layer based on federated learning and a privacy protection layer based on differential privacy and blockchain. Federated learning allows model training without sharing the original data, protecting data privacy. Differential privacy technology protects individual privacy by adding noise to the data, while blockchain technology ensures the immutability and traceability of the data. This multi-dimensional privacy protection approach meets the needs of data processing and analysis while fully respecting the privacy rights of students and faculty, effectively resolving privacy disputes in the prevention and control of aggressive behavior on campus. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the system structure of the campus excessive behavior prevention system and method of the present invention, according to Embodiment 1.

[0031] Figure 2 This is a schematic diagram illustrating the technical implementation process of the intelligent analysis database in Embodiment 1 of the campus excessive behavior prevention system and method of the present invention. Detailed Implementation

[0032] The following detailed explanation illustrates the specific implementation methods:

[0033] Example 1

[0034] The basic implementation examples are as follows: Figure 1 As shown: The campus excessive behavior prevention system includes an Internet of Things sensing system, an intelligent analysis database, and an intelligent push system.

[0035] The IoT sensing system is used to collect multi-source data; the multi-source data includes environmental data and biological characteristic data; the environmental data includes carbon concentration and temperature and humidity; the biological characteristic data includes biological temperature, biological movement and voiceprint data.

[0036] Specifically, the IoT sensing system includes an environmental monitoring unit and a biometric sensing unit.

[0037] The environmental monitoring unit includes a CO2 sensor for capturing changes in carbon concentration within the sensing area, LoRa temperature and humidity nodes for capturing changes in ambient temperature and humidity and biological temperature within the sensing area, and a thermal imaging camera. In this embodiment, a high-precision CO2 sensor driven by an STM32 microcontroller is selected, with a detection range of 0–5000 ppm and an error of ±2%; a thermal imaging camera in the 8–14 μm far-infrared band is selected to accurately collect environmental information.

[0038] Among these technologies, CO2 sensors can monitor changes in CO2 concentration gradients in enclosed spaces (such as restrooms and stairwells) in real time, capturing situations where respiratory rates surge (e.g., concentrations exceeding 1000 ppm) due to group gatherings or conflicts, thus aiding in the early identification of risks of aggressive behavior on campus. Furthermore, by integrating thermal imaging cameras, infrared spectroscopy can analyze student body temperature fluctuations (e.g., an adrenaline surge can cause a 0.5-1°C temperature rise), identifying abnormal physiological responses. This also contributes to the early identification of risks of aggressive behavior on campus and helps in the timely detection of potential health risks to students.

[0039] The biometric sensing unit includes an optical camera and millimeter-wave radar for collecting biometric motion information within the sensing area, and a directional microphone array for collecting voiceprint information. In this embodiment, a 2-megapixel optical camera is selected to clearly capture motion information.

[0040] The environmental monitoring unit uses the YOLOv7 algorithm, combined with skeletal keypoints, to identify target actions based on biological motion information. It also uses the BERT model to extract target semantics based on voiceprint information. The target actions include pushing, blocking, and pulling actions; the target semantics include abusive and threatening semantics. Furthermore, when analyzing voiceprint information, the BERT model only extracts metadata such as tone, speech rate, and keyword frequency to avoid the risks associated with storing raw audio.

[0041] This solution is designed to completely avoid the privacy risks of facial recognition and the risk of personal privacy leakage, achieving contactless behavior recognition and privacy-preserving voiceprint analysis.

[0042] The IoT sensing system adopts a modular structure, with the environmental monitoring unit, biometric sensing unit, and edge computing unit forming an integrated module for easy deployment. After the environmental monitoring unit and biometric sensing unit collect multi-source data, the edge computing unit preprocesses the data before transmitting it to other systems. This preprocessing includes using an adaptive Kalman filter algorithm to suppress data noise.

[0043] When deploying the integrated modules, the placement points are selected in conjunction with an atmospheric diffusion model. Specifically, for open spaces such as playgrounds and corridors, a Gaussian plume model suitable for steady-state diffusion simulation is used for placement selection; for complex architectural spaces such as stairwells and toilets, a CFD (Computational Fluid Dynamics) model suitable for unsteady-state diffusion simulation is used for placement selection. The input parameters of the model include the three-dimensional structure of the building (such as a BIM model or LiDAR scan data), the intensity of the CO2 emission source (in this embodiment, it is estimated based on population density, such as 0.005 L / s CO2 exhaled by each person), wind speed, and wind direction. Taking the CFD model as an example, the CFD model outputs a diffusion characteristic distribution map of the target area (such as toilets and corridors) based on the input parameters. Then, a genetic algorithm is used to select the optimal placement combination with the goal of minimizing the monitoring blind zone, such as deploying one integrated module every 10 meters in the corridor; and placing two integrated modules diagonally in the toilet cubicles, covering the breathing zone height.

[0044] The intelligent analysis database is used to complete multi-source data fusion and privacy protection, including a time-series database for storing environmental data, a graph database for constructing behavioral correlation graphs, a data processing layer based on federated learning, and a privacy protection layer based on differential privacy and blockchain, such as... Figure 2 As shown.

[0045] Specifically, in this embodiment, the time series database can be the existing InfluxDB or TimescaleDB. Such databases can be specifically optimized for time series data, support efficient writing, compressed storage and fast querying, and are suitable for real-time monitoring scenarios.

[0046] The graph database used is Neo4j, which constructs a behavioral association graph through nodes, relationships, and attributes. Node types can include: students (containing attributes such as student ID, grade, and class); behaviors such as "pushing," "blocking," and "insulting"; locations such as "toilet stall A3" and "second-floor corridor corner"; times such as timestamps accurate to the minute; and devices, containing the associated integrated module ID (e.g., camera C01 and CO2 sensor S12 within integrated module x). Relationship types can include: participation (student → behavior), recording whether a student initiates or participates in a specific behavior; occurrence (behavior → location), marking the location where the behavior occurred; detection (device → behavior), associating sensor data with the behavioral event; and time association (behavior → time), marking the specific time the behavior occurred.

[0047] Furthermore, when the environmental monitoring unit identifies target actions, target semantics, and environmental anomalies (i.e., CO2 concentration exceeding a threshold or change gradient exceeding a threshold), it also triggers real-time event injection (automatically inserting newly detected behaviors into the behavior association graph of the graph database). In addition, during the operation of the graph database, student IDs are encrypted and stored, and behavioral data is decoupled from student IDs; only non-sensitive attributes such as behavior type, time, and location are retained to comply with the requirements of the Personal Information Protection Law.

[0048] By using the behavioral association graph of the graph database, this campus security system can not only trace historical events, but also predict conflict risks in real time, providing a scientific basis for proactive intervention.

[0049] The data processing layer implements localized data processing at edge nodes through a federated learning framework, and combines a population dynamics model to quantify the air mass retention index and metabolic rate mutation threshold based on carbon concentration.

[0050] The air mass retention index AI is used to reflect the air flow within a confined space.

[0051] in, CO2 concentration gradient (ppm / m 3 V can be calculated by interpolating data from adjacent sensors; V refers to the spatial volume (m³). 3 Q can be obtained based on BIM models or measured data; Q refers to the ventilation rate (m³ / s). 3 / h). When AI is less than 0.8, it indicates good air circulation and no risk of air stagnation; when AI is greater than 0.8, it can be determined that there is a risk of air stagnation and there may be group conflict.

[0052] The metabolic rate Q m It is used to reflect the total amount of CO2 exhaled by a group per unit time, and is positively correlated with the number of people and the intensity of activity.

[0053] Q m = n·R·k; where n refers to the number of people, counted via camera in this embodiment; R refers to the average CO2 exhalation rate per person (0.005 L / s under sedentary conditions, 0.01 L / s under active conditions); k refers to the unit conversion factor (1 L = 0.001 m 3 ).

[0054] The metabolic rate mutation threshold is set in conjunction with the multimodal fusion model of the intelligent push system; the metabolic rate mutation specifically refers to: predicting the baseline CO2 concentration under normal conditions using an LSTM layer (e.g., a baseline of 400 ppm in an empty classroom); then calculating the difference ΔCO2 between the real-time CO2 concentration and the baseline; and then analyzing based on the time window method. If the difference ΔCO2 is greater than the metabolic rate mutation threshold within a certain period of time (e.g., within 5 minutes), it proves that a metabolic rate mutation has occurred.

[0055] Optionally, in large-scale events such as sports meets and gatherings, the aforementioned indices (AI and Q) can be used. m Analyzing changes in crowd density and movement trends can help optimize the allocation of security resources and prevent stampede risks and conflicts.

[0056] The intelligent push system is equipped with a multimodal fusion model, which calculates and outputs an overreaction risk index based on multi-source data and triggers tiered alarms.

[0057] The multimodal fusion model is an LSTM-DNN hybrid model; the LSTM layer predicts the CO2 concentration baseline and dynamically calibrates the threshold (specifically, the metabolic rate mutation threshold in this embodiment); during dynamic threshold calibration, a basic threshold is first generated based on the site capacity and ventilation efficiency, and then the threshold is calibrated based on this, combined with real-time data; the DNN layer fuses multimodal features (including multi-source data and air mass retention index AI and metabolic rate Q obtained based on multi-source data). m (e.g., behavioral correlation graphs, etc.) and outputs an overreaction risk index.

[0058] The risk index for the excessive behavior ranges from 0 to 1; the intelligent push system is linked with the campus's internal information system; the information system includes an LCD screen, a broadcast system, a buzzer, and a security terminal;

[0059] The tiered alarm includes: when the risk index of aggressive behavior is between 0.6 and 0.8, a level 1 alarm is triggered, which activates the LCD screen warning and directional broadcast; when the risk index of aggressive behavior is greater than 0.8, a level 2 alarm is triggered, which links the buzzer and 5G NR technology to push the alarm information to the security terminal with low latency, and simultaneously adjusts the fresh air system to manage carbon concentration.

[0060] Furthermore, in the application of this campus overreaction prevention system, the integration module collects multi-source data, preprocesses the multi-source data based on the edge computing unit, and sends the data to the regional gateway after AES-256 encryption. The regional gateway aggregates data from multiple nodes, verifies the integrity of the fragments, and aggregates and updates the data locally through federated learning. After compressing the data, it is uploaded to the cloud. The cloud decrypts the data and calls the intelligent analysis database and intelligent push system to store and analyze the data.

[0061] This embodiment also provides a method for preventing excessive behavior on campus, which uses the excessive behavior prevention system as described in Scheme 1 to prevent excessive behavior on campus; including the following steps:

[0062] The system employs an Internet of Things (IoT) sensing system to collect multi-source data, including environmental data and biometric data.

[0063] Multi-source data fusion and privacy protection are accomplished by an intelligent analysis database;

[0064] The multimodal fusion model of the intelligent push system calculates and outputs an overreaction risk index based on multi-source data, and triggers tiered alerts.

[0065] This embodiment provides a campus excessive behavior prevention system and method that can balance privacy protection, real-time early warning and accurate identification. It can effectively solve the problems of concealment, delay and privacy disputes in the prevention and control of excessive behavior on campus, and provide a standardized solution for building a smart campus with zero excessive behavior and high efficiency.

[0066] Example 2

[0067] An improvement to the first embodiment of a campus overreaction prevention system has been made as follows.

[0068] The intelligent push system also includes a group behavior quantification model.

[0069] Specifically, the group behavior quantification model is constructed using CO2 concentration gradient, voiceprint information, and other core parameters, and outputs a behavior risk index. It can predict conflict risks 10-15 minutes in advance, breaking the traditional "post-event tracing" logic.

[0070] The behavioral risk index = α·AI + β·ΔCO2 + γ·f 动作 +δ·P 情感 .

[0071] Where α, β, γ, and δ are weighting coefficients. In this embodiment, supervised learning (labeling historical conflict events) is used to train the model, and gradient descent is used to optimize the weighting coefficients. AI is the air mass retention index; ΔCO2 is the difference between the real-time CO2 concentration and the baseline; f 动作The frequency of target actions (such as pushing, blocking, and pulling) detected per unit of time reflects the level of activity in physical conflict; P 情感 This represents the probability that the emotional tendency of the voice obtained through voiceprint analysis is negative (such as threats or insults), which is the frequency of the target semantics. The higher the value, the greater the risk of conflict.

[0072] The behavioral risk index ranges from 0 to 1; the intelligent push system is linked with the campus's internal information system; the information system includes an LCD screen, broadcast system, buzzer, and security terminal.

[0073] The tiered alarm includes: when the risk index of aggressive behavior is between 0.6 and 0.8, a level 1 alarm is triggered, which activates the LCD screen warning and directional broadcast; when the risk index of aggressive behavior is greater than 0.8, a level 2 alarm is triggered, which links the buzzer and 5G NR technology to push the alarm information to the security terminal with low latency, and simultaneously adjusts the fresh air system to manage carbon concentration.

[0074] The campus excessive behavior prevention system provided in this embodiment can directly output the risk index using a quantitative model, compared with the first embodiment. Compared with the multimodal fusion model constructed in the first embodiment, it has lower construction cost and higher operating efficiency.

[0075] Example 3

[0076] An improvement to the first embodiment of a campus overreaction prevention system has been made as follows.

[0077] The intelligent push system also links with the campus's fresh air system and lighting equipment, and dynamically adjusts their operating status based on CO2 concentration and camera data to regulate energy consumption. This includes: sending a wind speed increase signal to the fresh air system in the corresponding area when the CO2 concentration exceeds the corresponding threshold; sending a wind speed decrease signal to the fresh air system in the corresponding area when the CO2 concentration is below the corresponding threshold; and sending a shutdown signal to the fresh air system and lighting equipment in the corresponding area when the camera shows that there are no people in its sensing area.

[0078] The campus overreaction prevention system provided in this embodiment, compared with the first embodiment, can also be linked with the internal campus system, which helps to improve environmental comfort while reducing energy consumption, and simultaneously achieve the goals of safety prevention and low carbon.

[0079] Example 4

[0080] An improvement to the first embodiment of a campus overreaction prevention system has been made as follows.

[0081] In addition to the IoT sensing system, intelligent analysis database, and intelligent push system, a visualization management system is also included. This visualization management system is used to visually display multi-source data.

[0082] The IoT sensing system, intelligent analysis database, and intelligent push system are integrated into the backend information management system. The backend information management system uses the Spring Boot framework to receive, store, and process data uploaded from the hardware. The visualization management system is located in the frontend interface, which uses the Vue framework to display multi-source data. Optionally, it can also be used to display data such as the number of people, air mass retention index (AI), and metabolic rate (Q) derived from the multi-source data. m Behavioral correlation maps, extreme behavior risk indices, etc.

[0083] Furthermore, preferably, the visualization management system also constructs a 3D digital twin campus using Three.js and renders a risk heat map in real time based on the risk index of excessive behavior.

[0084] This embodiment provides a campus overreaction prevention system. Compared to Embodiment 1, it integrates the front-end and back-end of each system, enabling better integration with existing campus management systems and offering superior visualization capabilities.

[0085] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics of the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent.

Claims

1. A campus excessive behavior prevention system, characterized in that: It includes an IoT sensing system, an intelligent analysis database, and an intelligent push system; the IoT sensing system is used to collect multi-source data; the multi-source data includes environmental data and biometric data; the environmental data includes carbon concentration and temperature and humidity; the biometric data includes biological temperature, biological movement, and voiceprint data; The intelligent analysis database is used to complete multi-source data fusion and privacy protection, including a time-series database for storing environmental data, a graph database for constructing behavioral correlation graphs, a data processing layer based on federated learning, and a privacy protection layer based on differential privacy and blockchain. The intelligent push system is equipped with a multimodal fusion model, which calculates and outputs an overreaction risk index based on multi-source data and triggers tiered alarms.

2. The campus excessive behavior prevention system according to claim 1, characterized in that, The IoT sensing system includes an environmental monitoring unit and a biometric sensing unit; the environmental monitoring unit includes components for capturing changes in carbon concentration within the sensing area. The sensor includes a LoRa temperature and humidity node and a thermal imaging camera, which are used to capture changes in environmental temperature and humidity and biological temperature within the sensing area, respectively. The biometric sensing unit includes an optical camera and a millimeter-wave radar for collecting biological motion information within the sensing area, as well as a directional microphone array for collecting voiceprint information.

3. The campus excessive behavior prevention system according to claim 2, characterized in that, The environmental monitoring unit identifies target actions based on biological motion information using the YOLOv7 algorithm, and extracts target semantics based on voiceprint information using the BERT model; the target actions include pushing, blocking, and pulling actions; the target semantics include abusive semantics.

4. The campus excessive behavior prevention system according to claim 1, characterized in that, The data processing layer implements localized data processing at edge nodes through a federated learning framework, and combines a population dynamics model to quantify the air mass retention index and metabolic rate mutation threshold based on carbon concentration.

5. The campus excessive behavior prevention system according to claim 1, characterized in that, The multimodal fusion model is an LSTM-DNN hybrid model; predictions are made by the LSTM layer. Concentration baseline, and dynamically calibrate thresholds; When dynamically calibrating the threshold, a basic threshold is first generated based on the venue capacity and ventilation efficiency. Then, based on this, the threshold is calibrated by combining real-time data. Multimodal features are fused by the DNN layer, and an overreaction risk index is output.

6. The campus excessive behavior prevention system according to claim 2, characterized in that, The IoT sensing system adopts a modular structure, with the environmental monitoring unit, biometric sensing unit, and edge computing unit forming an integrated module for easy deployment.

7. The campus excessive behavior prevention system according to claim 6, characterized in that, When deploying the integrated modules, the placement points are selected in conjunction with an atmospheric diffusion model.

8. The campus excessive behavior prevention system according to claim 6, characterized in that, The integration module collects multi-source data, preprocesses the multi-source data based on the edge computing unit, and sends the fragmented data to the regional gateway after AES-256 encryption. The regional gateway aggregates the data from multiple nodes, verifies the integrity of the fragments, and aggregates and updates the data locally through federated learning, and uploads the compressed data to the cloud. Data is decrypted in the cloud and stored and analyzed by calling the intelligent analysis database and intelligent push system.

9. The campus excessive behavior prevention system according to claim 1, characterized in that, The risk index for the excessive behavior ranges from 0 to 1; the intelligent push system is linked with the campus's internal information system; the information system includes an LCD screen, broadcast system, buzzer, and security terminal. The tiered alarm includes: when the risk index of aggressive behavior is 0.6 to 0.8, a level 1 alarm is triggered, and the LCD screen warning and directional broadcast are activated; when the risk index of aggressive behavior is greater than 0.8, a level 2 alarm is triggered, and the buzzer and 5G NR technology are linked to push the alarm information to the security terminal with low latency, and the fresh air system is adjusted simultaneously to manage carbon concentration.

10. Methods for preventing excessive behavior on campus, characterized by: Applying the school aggressive behavior prevention system as described in any one of claims 1-9 to prevent school aggressive behavior includes the following steps: The system employs an Internet of Things (IoT) sensing system to collect multi-source data, including environmental data and biometric data. Multi-source data fusion and privacy protection are accomplished by an intelligent analysis database; The multimodal fusion model of the intelligent push system calculates and outputs an overreaction risk index based on multi-source data, and triggers tiered alerts.

Images