Campus safety monitoring system

Abstract

The invention discloses a campus safety monitoring system, which relates to the technical field of safety monitoring, and can more comprehensively identify potential safety hazards by performing fusion analysis on environmental data including temperature T, humidity H, air quality index AQI and noise level L and monitoring video stream data V. Compared with a traditional system depending on video monitoring, fusion of environmental data provides more contextual information. Through introduction of a security event score and a regression model, the system can predict future security threats according to historical data and current environment changes. The predication function greatly improves the perspectiveness of campus safety management, potential risks can be found in time, the possibility and severity of events can be evaluated, and emergency measures can be taken in advance. When the system finds potential security threats through analysis, the system can automatically trigger alarm and inform management personnel in the school of security and fire protection.

Description

Technical Field

[0001] The present invention relates to the technical field of security monitoring, and specifically to a campus security monitoring system. Background Art

[0002] With the continuous development of the educational environment and campus scale, traditional monitoring systems are gradually unable to meet the requirements of real-time early warning and accurate response to complex security threats. Therefore, fusing and analyzing environmental data including temperature, humidity, air quality, and noise with existing monitoring video data has become a new direction for improving campus security management levels. Through this fusion technology, potential security hazards can be identified more accurately, such as extreme weather, environmental pollution, or harmful gas leakage.

[0003] Current campus security monitoring systems usually focus on real-time video monitoring and event alarms, and mostly pay attention to common physical security risks, such as fires, thefts, violent behaviors, etc. However, most of these systems ignore the collection and comprehensive analysis of environmental data and fail to consider the potential impact of environmental factors such as temperature, humidity, air quality, and noise on campus security. Traditional monitoring systems often react slowly when environmental abnormal events occur and cannot timely identify security hazards caused by weather changes or environmental pollution.

[0004] Due to the lack of comprehensive analysis of environmental data in existing campus security monitoring systems and mostly relying on single video monitoring, when facing complex security threats, the system reacts slowly and it is difficult to timely identify potential risks. In the case of deteriorating air quality, traditional security systems usually can only handle emergencies in video images and ignore health problems caused by air pollution, such as toxic gas leakage or breathing crises brought by smoggy weather.

[0005] Traditional monitoring systems often cannot foresee these potential security threats. Therefore, the lack of fusion analysis of environmental data and video monitoring data leads to insufficient campus security prevention and control capabilities, unable to timely discover and respond to potential security hazards, and may seriously lead to health crises or property losses. Summary of the Invention

[0006] In view of the deficiencies of the prior art, the present invention provides a campus security monitoring system, which solves the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention is realized through the following technical solutions: A campus security monitoring system includes a data collection and preprocessing module, an environmental data analysis module, a monitoring video and environmental data fusion analysis module, a security event prediction and evaluation module, an alarm and response module, and a result optimization and feedback module;

[0008] The data acquisition and preprocessing module is responsible for collecting data from multiple sensors and monitoring cameras, including environmental data ED and video stream data V, and performing preprocessing to obtain the dataset W;

[0009] The environmental data analysis module analyzes the environmental data ED in the dataset W, analyzes the change trend of the environmental data ED, judges abnormal changes, and evaluates potential safety hazards;

[0010] The monitoring video and environmental data fusion analysis module performs comprehensive analysis on the dataset W, analyzes the dynamic changes in the monitoring screen through image recognition technology, combines with the changes in environmental data, and calculates and obtains the safety event score S;

[0011] The safety event prediction and evaluation module predicts future safety threat events based on the safety event score S and safety hazards; through a regression model, it predicts future safety risks and issues risk warnings;

[0012] The alarm and response module automatically triggers an alarm according to the safety threat event and risk warning, notifies the school management personnel, and initiates emergency response measures;

[0013] The result optimization and feedback module adjusts the parameters of the regression model again according to the emergency response measures and feedback information, and makes judgment and warning.

[0014] Preferably, the data acquisition and preprocessing module includes a data acquisition unit and a data preprocessing unit;

[0015] The data acquisition unit collects environmental data ED and video stream data V through a variety of environmental sensors and monitoring cameras. Among them, the environmental data includes temperature T, humidity H, air quality index AQI, and noise level L. The variety of environmental sensors include temperature sensors, humidity sensors, air quality sensors, and noise sensors. Temperature T is collected through a temperature sensor, humidity H is collected through a humidity sensor, air quality index AQI is collected through an air quality sensor, noise level L is collected through a noise sensor, and video stream data V is obtained through a monitoring camera;

[0016] The data preprocessing unit cleans and standardizes the collected data to obtain the dataset W;

[0017] Among them, data cleaning includes filling the data using interpolation method and mean filling method; data standardization processing includes using Min-Max normalization to normalize the data to the same scale range to eliminate the influence brought by different data dimensions.

[0018] Preferably, the environmental data analysis module includes a trend analysis unit and an abnormal detection and safety hazard evaluation unit;

[0019] The trend analysis unit is responsible for performing time series analysis on the environmental data ED, evaluating the change trend of the environmental data ED over time, and identifying the normal and abnormal fluctuations of the environmental data ED by calculating the change rates of each parameter in the environmental data ED, including the temperature change rate ΔT, the humidity change rate ΔH, the air quality index change rate ΔAQI, and the noise level change rate ΔL;

[0020] The temperature change rate ΔT is obtained by the ratio of the difference between the temperature at time t and the temperature at time t-1 to the temperature at time t-1;

[0021] The humidity change rate ΔH is obtained by the ratio of the difference between the humidity at time t and the humidity at time t-1 to the humidity at time t-1;

[0022] The air quality index change rate ΔAQI is obtained by the ratio of the difference between the air quality index at time t and the air quality index at time t-1 to the air quality index at time t-1;

[0023] The noise level change rate ΔL is obtained by the ratio of the difference between the noise level at time t and the noise level at time t-1 to the noise level at time t-1;

[0024] The anomaly detection and safety hazard assessment unit detects and evaluates the obtained temperature change rate ΔT, humidity change rate ΔH, air quality index change rate ΔAQI, and noise level change rate ΔL, and compares them with the temperature threshold ZT, humidity threshold ZH, air quality threshold ZAQI, and noise threshold ZL to judge the abnormal situation of the environmental data ED and evaluate potential safety hazards, including fire risk, health risk, and group conflict and riot risk;

[0025] When the temperature change rate ΔT > the temperature threshold ZT, it indicates that the temperature is abnormal, indicating fire risk and high temperature risk;

[0026] When the humidity change rate ΔH > the humidity threshold ZH, it indicates that the humidity is abnormal, indicating the risk of rainfall;

[0027] When the air quality index change rate ΔAQI > the air quality threshold ZAQI, it indicates that the air quality index is abnormal, indicating health risk;

[0028] When the noise level change rate ΔL > the noise threshold ZL, it indicates that the noise level is abnormal, indicating the risk of group conflict and riot.

[0029] Preferably, the monitoring video and environmental data fusion analysis module includes an image recognition and event detection unit and an environmental data fusion and safety score calculation unit;

[0030] The image recognition and event detection unit is responsible for real-time analysis of the video stream data V, detecting dynamic changes and potential security events in the video by using image recognition technology, and generating a security event feature vector EV, including personnel density Db, abnormal behavior identifier Ab, and behavior duration Tb;

[0031] The personnel density Db is obtained by the ratio of the number of people N appearing in the monitoring screen to the area of the monitoring screen;

[0032] The abnormal behavior identifier Ab identifies abnormal behaviors in the video, including violence, falling, and running fast, through a long short-term memory network, and outputs a binary identifier, where 0 indicates no abnormality and 1 indicates an abnormal behavior;

[0033] The behavior duration Tb is obtained by monitoring the duration of the occurrence of abnormal behaviors.

[0034] Preferably, the environmental data fusion and security score calculation unit combines the security event feature vector EV with the environmental data ED, calculates a comprehensive security event score S through a fusion algorithm, and evaluates the impact of environmental conditions on potential safety hazards when a security event occurs, including the impact of high temperature and poor air quality on the behavior of the crowd;

[0035] The security event score S is calculated by the weighted average method;

[0036] The risk level of the security event is divided according to the security event score S;

[0037] The risk levels are divided into the following three grades:

[0038] When 0 < security event score S < 0.3, it represents the first risk level;

[0039] When 0.3 ≤ security event score S < 0.7, it represents the second risk level;

[0040] When 0.7 ≤ security event score S < 1, it represents the third risk level.

[0041] Preferably, the security event prediction and evaluation module includes a historical data processing and feature extraction unit and a regression prediction and risk evaluation unit;

[0042] The historical data processing and feature extraction unit extracts features from the security event score S and potential safety hazards to obtain a feature set LF, including the change trend ΔS of the security score, the frequency Efr of sudden security events, temperature T, air quality index AQI, noise level L, and the correlation between environmental data and security events;

[0043] The change trend ΔS of the safety score is obtained by the ratio of the difference between the safety event score at time t and the safety event score at time t-1 to the safety event score at time t-1;

[0044] The frequency Efr of the sudden safety event is obtained by the ratio of the safety events that occurred within time t to the total events that occurred within time t;

[0045] The associations between the environmental data and the safety events include the associations between temperature and safety events, humidity and safety events, air quality and safety events, and noise and safety events.

[0046] Extreme temperatures may trigger violent behaviors, group conflicts, or mood swings in people. A sharp change in temperature may trigger health problems such as heatstroke, collapse, or respiratory diseases, especially in an environment without air conditioning or ventilation;

[0047] Excessive humidity may cause malfunctions of electronic devices on campus, including surveillance cameras, access control systems, laboratory equipment, etc., which may then lead to system failures or facility damages. Excessive or too low humidity may affect people's mental states, resulting in low mood or anxiety, etc., thereby increasing the occurrence of violent behaviors or sudden health events;

[0048] Poor air quality, especially when the pollutant concentration is too high, may pose a threat to the health of teachers and students, leading to respiratory problems or disease outbreaks. The increase in such health hazards may also trigger medical emergencies. When the air pollution is severe, it may cause teachers and students to reduce outdoor activities and gather people in indoor spaces, thereby increasing the risk of accidents caused by crowded people or uncomfortable environments;

[0049] Too high noise levels may affect people's moods, trigger restlessness and stress, and increase the risk of conflicts and violent behaviors, especially in noisy places such as school dormitories and cafeterias. In a high-noise environment, teachers and students may be distracted, increasing the probability of accidents;

[0050] Preferably, the regression prediction and risk assessment unit inputs the obtained feature set LF into a regression model to predict future safety risks, predicts the future event score SFu, and evaluates the future safety risk level through the future event score SFu;

[0051] The future event score SFu is obtained by calculation through the regression model;

[0052] The future safety risk level is obtained by matching in the following manner:

[0053] When 0 < Future Event Score SFu < 0.3, it represents the first risk level, low risk. Regularly monitor environmental data, including temperature T, humidity H, and Air Quality Index AQI, and carry out regular safety publicity and education activities to remind students and teaching staff to stay vigilant.

[0054] When 0.3 ≤ Future Event Score SFu < 0.7, it represents the second risk level, medium risk. Adjust the monitoring accuracy and frequency, adjust the patrol frequency of security personnel, and conduct emergency safety drills regularly.

[0055] When 0.7 ≤ Future Event Score SFu < 1, it represents the third risk level, high risk. Activate the emergency safety response mechanism, increase the security personnel configuration on campus, strengthen the patrol of key areas, and intervene in possible safety incidents in a timely manner; consider setting up temporary security measures in high-risk areas; if the prediction shows extremely high security risks, large-scale early warnings and evacuations should be considered, especially in the face of possible disasters such as fires and earthquakes.

[0056] Preferably, the alarm and response module includes an alarm trigger unit and an emergency response activation unit.

[0057] The alarm trigger unit determines the activation of the alarm program based on the result of the Future Event Score SFu.

[0058] When 0.7 ≤ Future Event Score SFu < 1, activate the alarm program and notify security personnel and the management department.

[0059] The emergency response activation unit notifies relevant school administrators according to the alarm program and activates emergency response measures, and transmits the event information to relevant personnel by means of text messages, APP notifications, and alarms.

[0060] According to the event types including fires, explosions, and violent incidents, activate the corresponding emergency plans; for example, a fire incident will activate the fire protection plan, and a violent incident will activate the security and medical plans, and calculate and obtain the response time Tre.

[0061] The response time Tre is obtained by multiplying the ratio of 1 to the Future Event Score SFu and the preliminary judgment time Δti.

[0062] Preferably, the result optimization and feedback module includes a feedback collection and analysis unit and a model adjustment and optimization unit.

[0063] The feedback collection and analysis unit collects feedback information on alarm events, including whether the alarm event is accurately triggered, the effect of response measures, and the actual result of the event, analyzes the collected feedback information, and evaluates the effectiveness of alarm trigger rules and emergency response measures.

[0064] The analysis content includes false alarms and missed alarms, the response time Tre and the effectiveness of emergency measures, as well as the impact of environmental data and event types on security threats.

[0065] Preferably, the model adjustment and optimization unit adjusts the parameters in the regression model according to the feedback information;

[0066] According to the analysis result of the feedback information, improve the sensitivity of alarm triggering, so that the model can more accurately identify risk events in the future, calculate and obtain the parameters βx of the new regression model, and adjust the training regression model;

[0067] Substitute the parameters βx of the new regression model into the formula of the regression model to obtain the new event score SFunew, and make a judgment and early warning.

[0068] The present invention provides a campus security monitoring system, which has the following beneficial effects:

[0069] (1) When the system is running, by fusing and analyzing environmental data including temperature T, humidity H, air quality index AQI and noise level L with the monitoring video stream data V, the system can more comprehensively identify potential security hazards. Compared with traditional systems that rely solely on video monitoring, the integration of environmental data provides more context information. Through the introduction of security event scoring and regression models, the system can predict future security threats based on historical data and current environmental changes. This prediction function greatly improves the foresight of campus security management. It can not only detect potential risks in a timely manner, but also evaluate the likelihood and severity of events and take emergency measures in advance.

[0070] When the system discovers potential security threats through analysis, it can automatically trigger an alarm and notify school administrators including security and fire protection. This automated response mechanism enables security managers to obtain event information in a timely manner and take effective emergency measures quickly, greatly improving the response speed and processing efficiency. At the same time, since the system can give early warnings based on the prediction results, the alarm and response are not limited to sudden events, but can also make preventive preparations in advance to reduce losses and injuries.

[0071] (2) Calculate the safety event score S by the weighted average method and divide the risk levels according to the score. Dividing the risks into three levels can more clearly indicate the criticality of the events. According to the score, the system can respond in a timely manner, prioritize the handling of high-risk events, and ensure the effective allocation of limited resources. When calculating the safety event score, the system avoids the influence of different data scale differences on the score result through standardization processing, and further improves the accuracy and consistency of the score by using the peak and valley normalization method. This standardization processing enables the system to adapt to different environments and situations and avoids errors caused by changes in the data volume. This adaptive mechanism enables the system to adjust according to different campus environments and seasonal changes during long-term use, and always maintain high accuracy and sensitivity.

[0072] By conducting real-time evaluation and risk classification on the safety event score S, the system can provide clear indication of the severity of the events for security managers, enabling them to make more efficient decisions.

[0073] (3) By predicting the future event score SFu through a regression model, the system can give early warnings in advance and initiate corresponding intervention measures in a timely manner according to different risk levels. This predictive warning system not only enhances the ability to predict unexpected events but also can initiate relevant emergency response plans in a timely manner, greatly shortening the event response time and improving the effect of emergency handling. Especially before high-risk events occur, the system can provide early warnings, thus providing sufficient time for the school authorities to prepare emergency resources and coordinate response teams to avoid the expansion of losses.

[0074] By continuously collecting and analyzing safety event data, the system constantly adjusts and optimizes the prediction accuracy of the regression model. This adaptive ability enables the system to continuously improve during long-term operation, dynamically optimize safety events under different environmental conditions and seasonal changes, thereby improving the accuracy and timeliness of prediction. At the same time, the system can also adjust the feature extraction and risk assessment strategies according to the feedback information to ensure that it always maintains a high level of security prevention and control in the complex and changeable campus environment.

[0075] (4) Through the false alarm rate evaluation formula, the system can effectively identify false-triggered alarm situations and calculate the false alarm rate FAR. After evaluating the alarm accuracy, the system will adjust the parameters Δβ of the regression model according to the feedback information to improve the alarm accuracy. This method effectively reduces the false alarm rate, avoids unnecessary intervention and waste of resources, ensures that the alarm system only issues alarms when really needed, and thus improves the overall credibility of the campus security system.

[0076] Through the analysis of feedback information, the system can adjust the parameter βx in the regression model to further improve the prediction accuracy of event scoring. The parameter update of the new regression model enables the system to be optimized according to actual data and feedback in the future, ensuring more accurate prediction of potential risks. As the system runs deeper, it can continuously self-adjust and optimize to adapt to new types of security events that may occur in the campus environment, thereby continuously improving its intelligence level. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 It is a schematic diagram of the block diagram process of a campus security monitoring system according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0078] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0079] Embodiment 1

[0080] The present invention provides a campus security monitoring system. Please refer to Figure 1 , which includes a data collection and preprocessing module, an environmental data analysis module, a monitoring video and environmental data fusion analysis module, a security event prediction and evaluation module, an alarm and response module, and a result optimization and feedback module;

[0081] The data collection and preprocessing module is responsible for collecting data from multiple sensors and monitoring cameras, including environmental data ED and video stream data V, and performing preprocessing to obtain a data set W;

[0082] The environmental data analysis module analyzes the environmental data ED in the data set W, analyzes the change trend of the environmental data ED, judges abnormal changes, and evaluates potential security hazards;

[0083] The monitoring video and environmental data fusion analysis module comprehensively analyzes the data set W, analyzes the dynamic changes in the monitoring screen through image recognition technology, and combines the changes in environmental data to calculate and obtain a security event score S;

[0084] The security event prediction and evaluation module predicts future security threat events based on the security event score S and security hazards; through a regression model, it predicts future security risks and issues risk warnings;

[0085] The alarm and response module automatically triggers an alarm according to the security threat event and risk warning, notifies the school management personnel, and initiates emergency response measures;

[0086] The result optimization and feedback module re - adjusts the parameters of the regression model according to the emergency response measures and feedback information, and makes judgment and early warning.

[0087] In this embodiment, the system can more comprehensively identify potential safety hazards by fusing and analyzing environmental data including temperature T, humidity H, air quality index AQI, and noise level L with surveillance video stream data V. Compared with traditional systems that rely solely on video surveillance, the integration of environmental data provides more context information. Through the introduction of security event scoring and regression models, the system can predict future security threats based on historical data and current environmental changes. This prediction function greatly improves the foresight of campus security management. It can not only detect potential risks in a timely manner but also evaluate the likelihood and severity of events and take emergency measures in advance.

[0088] When the system discovers potential security threats through analysis, it can automatically trigger an alarm and notify school administrators including security and fire protection. This automated response mechanism enables security managers to obtain event information in a timely manner and take effective emergency measures quickly, greatly improving the response speed and handling efficiency. At the same time, since the system can give early warnings based on prediction results, the alarm and response are not limited to sudden events but can also make preventive preparations in advance to reduce losses and injuries.

[0089] This system has a result optimization and feedback mechanism. By collecting feedback information on alarm events and adjusting the parameters of the regression model according to this feedback, the system can achieve continuous optimization. This means that the system can not only learn and adapt by itself during actual operation but also adjust its strategies according to new types of security threats and changing environmental conditions, avoiding false alarms or missed alarms of the system. Through this feedback adjustment mechanism, the accuracy and response speed of the system are continuously improved to ensure that it can adapt to the ever - changing campus security needs.

[0090] The comprehensive analysis combining environmental data and monitoring data can not only detect potential threats brought about by environmental changes in real - time but also greatly enhance the overall security prevention and control ability of the campus through accurate prediction and assessment. Through effective risk early warning and emergency response, negative consequences such as property losses and casualties caused by sudden security incidents are avoided. The system can effectively reduce security hazards, improve the overall security level of the campus, and provide valuable decision - making support for management to optimize resource allocation.

[0091] Embodiment 2

[0092] This embodiment is an explanatory description based on Embodiment 1. Please refer to Figure 1 , specifically: The data acquisition and pre - processing module includes a data acquisition unit and a data pre - processing unit;

[0093] The data acquisition unit collects environmental data ED and video stream data V through a variety of environmental sensors and monitoring cameras. The environmental data includes temperature T, humidity H, air quality index AQI, and noise level L. The variety of environmental sensors includes temperature sensors, humidity sensors, air quality sensors, and noise sensors. The temperature T is collected through a temperature sensor, the humidity H is collected through a humidity sensor, the air quality index AQI is collected through an air quality sensor, the noise level L is collected through a noise sensor, and the video stream data V is obtained through a monitoring camera;

[0094] The data preprocessing unit cleans and standardizes the collected data to obtain a data set W;

[0095] Among them, data cleaning includes filling in the data using interpolation and mean filling methods; data standardization processing includes normalizing the data to the same scale range using Min-Max normalization.

[0096] The environmental data analysis module includes a trend analysis unit and an anomaly detection and safety hazard assessment unit;

[0097] The trend analysis unit is responsible for performing time series analysis on the environmental data ED, evaluating the change trend of the environmental data ED over time, and identifying the normal fluctuations and abnormal fluctuations of the environmental data ED by calculating the change rates of each parameter in the environmental data ED, including the temperature change rate ΔT, the humidity change rate ΔH, the air quality index change rate ΔAQI, and the noise level change rate ΔL;

[0098] The temperature change rate ΔT is obtained by the ratio of the difference between the temperature at time t and the temperature at time t - 1 and the temperature at time t - 1;

[0099] The humidity change rate ΔH is obtained by the ratio of the difference between the humidity at time t and the humidity at time t - 1 and the humidity at time t - 1;

[0100] The air quality index change rate ΔAQI is obtained by the ratio of the difference between the air quality index at time t and the air quality index at time t - 1 and the air quality index at time t - 1;

[0101] The noise level change rate ΔL is obtained by the ratio of the difference between the noise level at time t and the noise level at time t - 1 and the noise level at time t - 1;

[0102] The anomaly detection and safety hazard assessment unit detects and evaluates the obtained temperature change rate ΔT, humidity change rate ΔH, air quality index change rate ΔAQI, and noise level change rate ΔL, compares them with the temperature threshold ZT, humidity threshold ZH, air quality threshold ZAQI, and noise threshold ZL, determines the anomalies of the environmental data ED, and evaluates potential safety hazards, including fire risk, health risk, and group conflict and riot risk;

[0103] The temperature threshold ZT, humidity threshold ZH, air quality threshold ZAQI, and noise threshold ZL are obtained by the standard deviation method, and the obtaining formulas are as follows:

[0104]

[0105] In the formula, ZEDi represents the threshold of the i-th parameter of the environmental data ED, including the temperature threshold ZT, humidity threshold ZH, air quality threshold ZAQI, and noise threshold ZL, EDi represents the i-th parameter of the environmental data ED, μEDi represents the mean value of the i-th parameter of the environmental data ED, and σ represents the standard deviation of the i-th parameter of the environmental data ED;

[0106] When the temperature change rate ΔT > the temperature threshold ZT, it indicates that the temperature is abnormal, prompting fire risk and high temperature risk;

[0107] When the humidity change rate ΔH > the humidity threshold ZH, it indicates that the humidity is abnormal, prompting rainfall risk;

[0108] When the air quality index change rate ΔAQI > the air quality threshold ZAQI, it indicates that the air quality index is abnormal, prompting health risk;

[0109] When the noise level change rate ΔL > the noise threshold ZL, it indicates that the noise level is abnormal, prompting group conflict and riot risk

[0110] In this embodiment, by integrating multiple environmental sensors and surveillance cameras, the system can achieve real-time collection of multi-dimensional environmental data. This multi-sensor fusion method enables the monitoring system to not only rely on video data but also capture the potential impact of environmental changes on safety in real time. The system adopts data cleaning and standardization processing to ensure that the collected environmental data is not interfered by noise and the data errors are effectively corrected, thereby improving the accuracy of subsequent analysis. The data cleaning method can effectively fill in the data missing caused by equipment failures or other reasons and avoid negative impacts on the analysis results. The data standardization processing maps different types of environmental data to the same scale range to ensure that the influence of different parameters can be evenly processed during the analysis process, improving the consistency and reliability of data processing.

[0111] Through time series analysis for trend monitoring of environmental data, the system can identify the temperature change rate ΔT, humidity change rate ΔH, air quality index change rate ΔAQI, and noise level change rate ΔL, and evaluate their fluctuation trends at different time points. This analysis can help detect potential abnormal fluctuations in the environment and discover changes that may affect campus safety at an early stage. Abnormalities in the temperature change rate may be precursors of fire risks, abnormal humidity change rates may indicate rainfall risks, and changes in the air quality index may foreshadow health crises. By accurately capturing these changes, the system can respond in a timely manner and issue warnings to reduce the risk of emergencies.

[0112] When the system detects abnormal changes, combined with the thresholds of environmental data, it can be evaluated by the standard deviation method, and can intelligently judge which environmental changes have exceeded the normal fluctuation range and evaluate possible potential safety hazards. This high-precision anomaly detection mechanism enables the system to quickly respond to different types of safety risks, including fires, health risks, and group conflicts, and initiate emergency response plans in a timely manner according to the risk level. Through intelligent evaluation and judgment of different risks, measures can be taken in advance to reduce potential losses and hazards.

[0113] By using the standard deviation method to dynamically adjust the thresholds of environmental data, the system can adapt to different campus environments and seasonal changes, avoiding false alarms or missed alarms caused by fixed threshold settings. This adaptive mechanism enables the system to maintain a high degree of accuracy during long-term operation and is not affected by environmental changes. The system can continuously adjust the threshold settings according to historical data and environmental change situations, so as to more accurately detect and prevent potential safety hazards.

[0114] Embodiment 3

[0115] This embodiment is an explanatory description carried out in Embodiment 2, please refer to Figure 1 , specifically: The monitoring video and environmental data fusion analysis module includes an image recognition and event detection unit and an environmental data fusion and safety score calculation unit;

[0116] The image recognition and event detection unit is responsible for real-time analysis of the video stream data V, detecting dynamic changes and potential safety events in the video by using image recognition technology, and generating a safety event feature vector EV, including personnel density Db, abnormal behavior identifier Ab, and behavior duration Tb;

[0117] The personnel density Db is obtained by the ratio of the number of people N appearing in the monitoring screen to the area of the monitoring screen;

[0118] The abnormal behavior identifier Ab identifies abnormal behaviors in the video, including violence, falls, and rapid running, through a long short-term memory network, and outputs a binary identifier, where 0 indicates no abnormality and 1 indicates an abnormal behavior;

[0119] The behavior duration Tb is obtained by monitoring the duration of the occurrence of abnormal behaviors.

[0120] The environmental data fusion and safety score calculation unit combines the safety event feature vector EV with the environmental data ED, and calculates the comprehensive safety event score S through a fusion algorithm, and evaluates the impact of environmental conditions on potential safety hazards when a safety event occurs, including the impact of high temperature and poor air quality on the behavior of the crowd;

[0121] The safety event score S is calculated by the weighted average method;

[0122] The safety event score S is obtained through the following formula:

[0123]

[0124] In the formula, ω 1 , ω 2 , ω 3 , ω 4 and ω 5 respectively represent the preset weight values of the personnel density Db, the abnormal behavior identifier Ab, the temperature T, the air quality index AQI, and the noise level L, and ω 1 +ω 2 +ω 3 +ω 4 +ω 5 ≤1, Tno represents the normalized value of the temperature, Tmax represents the peak value of the temperature, Tmin represents the valley value of the temperature, AQIno represents the normalized value of the air quality index, AQImax represents the peak value of the air quality index, and AQImin represents the valley value of the air quality index;

[0125] The risk level of the safety event is divided according to the safety event score S;

[0126] The risk levels are divided into the following three grades:

[0127] When 0 < safety event score S < 0.3, it indicates the first risk level;

[0128] When 0.3 ≤ safety event score S < 0.7, it indicates the second risk level;

[0129] When 0.7 ≤ safety event score S < 1, it indicates the third risk level.

[0130] In this embodiment, the system analyzes the video stream data in real time through image recognition technology, and can detect the dynamic changes and potential security events in the video in a timely manner. This real-time nature enables the system to quickly respond to emergencies occurring on campus, including personnel gatherings and abnormal behaviors, and provide key data for subsequent security assessments.

[0131] By generating a security event feature vector containing features such as personnel density Db, abnormal behavior identifier Ab, and behavior duration Tb, the system can comprehensively capture and analyze the dynamic changes in the video. These multi-dimensional features not only provide the basic information of the event, but also identify potential abnormal behaviors, including violence, falls, and rapid running, through deep learning technology, improving the accuracy and comprehensiveness of event analysis. These features provide richer data support for subsequent security score calculation and risk assessment, helping to judge the severity and potential threat of the event.

[0132] The system not only analyzes the dynamic changes in the video, but also fuses the environmental data ED with the security event features. By comprehensively considering the impact of environmental conditions on behaviors, including the impact of high temperature and poor air quality on crowd behaviors, the system can provide a more accurate security event score. This multi-dimensional fusion analysis method enables the system to not only consider the video data itself, but also combine environmental change factors to comprehensively evaluate the risk of the event.

[0133] This system calculates the security event score S through the weighted average method and divides the risk level according to the score. Dividing the risk into three levels can more clearly indicate the critical degree of the event. According to the score, the system can make a timely response, prioritize the handling of high-risk events, and ensure the effective allocation of limited resources. When calculating the security event score, the system avoids the influence of different data scale differences on the score result through standardization processing, and further improves the accuracy and consistency of the score by using the peak and valley normalization method. This standardization processing enables the system to adapt to different environments and situations, avoiding errors caused by changes in the data volume. This adaptive mechanism enables the system to adjust according to different campus environments and seasonal changes during long-term use, always maintaining high accuracy and sensitivity.

[0134] By performing real-time evaluation and risk classification on the security event score S, the system can provide a clear indication of the severity of the event for security managers, enabling managers to make more efficient decisions. The risk level classification can help managers quickly identify high-risk events, prioritize the mobilization of resources for handling, and thus effectively avoid losses caused by misjudgment or response delays.

[0135] Embodiment 4

[0136] This embodiment is an explanatory description carried out in Embodiment 3. Please refer to Figure 1, specifically: the security event prediction and evaluation module includes a historical data processing and feature extraction unit and a regression prediction and risk assessment unit;

[0137] The historical data processing and feature extraction unit extracts features from the security event score S and security hazards to obtain a feature set LF, including the change trend ΔS of the security score, the frequency Efr of sudden security events, temperature T, air quality index AQI, noise level L, and the correlation between environmental data and security events;

[0138] The change trend ΔS of the security score is obtained by the ratio of the difference between the security event score at time t and the security event score at time t - 1 to the security event score at time t - 1;

[0139] The frequency Efr of sudden security events is obtained by the ratio of the security events that occurred within time t to the total events that occurred within time t;

[0140] The correlation between environmental data and security events includes the correlation between temperature and security events, the correlation between humidity and security events, the correlation between air quality and security events, and the correlation between noise and security events.

[0141] The regression prediction and risk assessment unit inputs the obtained feature set LF into a regression model to predict future security risks, predicts the future event score SFu, and evaluates the future security risk level through the future event score SFu;

[0142] The future event score SFu is obtained by calculating through the regression model;

[0143] The future event score SFu is obtained by the following formula:

[0144]

[0145] In the formula, β0 represents the parameter of the regression model, n represents the total number of features in the feature set LF, LFi represents the i-th feature in the feature set LF, βi represents the regression model coefficient of the i-th feature in the feature set LF, and C represents the error term of the regression model;

[0146] The future security risk level is obtained by the following matching method:

[0147] When 0 < future event score SFu < 0.3, it indicates the first risk level, low risk. Regularly monitor environmental data, including temperature T, humidity H, and air quality index AQI, and carry out regular safety publicity and education activities to remind students and teaching staff to stay vigilant;

[0148] When 0.3 ≤ Future Event Score SFu < 0.7, it indicates the second risk level, medium risk. Adjust the monitoring accuracy and frequency, adjust the patrol frequency of security personnel, and conduct emergency safety drills regularly.

[0149] When 0.7 ≤ Future Event Score SFu < 1, it indicates the third risk level, high risk. Activate the emergency safety response mechanism, increase the security personnel allocation on campus, strengthen the patrol of key areas, and intervene in possible security incidents in a timely manner; Consider setting up temporary security measures in high-risk areas; If the prediction shows extremely high security risks, large-scale early warnings and evacuations should be considered, especially in the face of possible disasters such as fires and earthquakes.

[0150] In this embodiment, the system extracts multiple security-related features through the historical data processing and feature extraction unit, including the change trend of security scores, the frequency of sudden security incidents, the correlation between environmental data and security incidents, etc. This feature extraction method integrates the data of historical security incidents and environmental factors, effectively capturing potential security hazards and risk patterns. Through the analysis of these features, the system can more accurately evaluate the possibility of future security incidents, thereby improving the accuracy of security incident prediction.

[0151] The system predicts future security risks based on the extracted feature set through the regression prediction and risk assessment unit. The regression model can predict the future security incident score according to historical data and existing features, thereby effectively evaluating the future security risk level. The introduction of this prediction model enables the system to predict possible future security incidents based on historical trends and environmental changes, take preventive measures in advance, and avoid the occurrence of emergencies or reduce losses. This data-based prediction of security risk assessment has stronger foresight and initiative compared to the traditional passive reactive security management mode.

[0152] During the feature extraction process, the system considers the correlation between environmental factors such as temperature T, humidity H, air quality index AQI, and noise level L and security incidents. This makes the risk assessment not only depend on the occurrence frequency and change trend of security incidents themselves, but also comprehensively considers the potential impact of environmental changes on the occurrence of incidents. Through this multi-dimensional data fusion, the system can more accurately evaluate the impact of environmental factors on the occurrence of security incidents. For example, high temperature may exacerbate the fire risk, and abnormal noise may be related to group conflicts. This method improves the accuracy of security incident prediction, especially when dealing with complex environments and multiple risk factors, and can better capture the microscopic changes in the occurrence of incidents.

[0153] The system classifies security risks into three levels through future event scoring, and each level corresponds to different security response measures. Low-risk events can be monitored regularly and security publicity can be carried out. Medium-risk events require enhanced security patrols and security drills. High-risk events need to activate the emergency response mechanism for personnel evacuation or increased security efforts. This classification of risk levels can not only help campus management accurately identify the severity of events, but also ensure the reasonable allocation of resources, concentrating emergency resources in high-risk areas or for handling emergencies, and avoiding waste of resources.

[0154] Embodiment 5

[0155] This embodiment is an explanatory description based on Embodiment 4. Please refer to Figure 1 , specifically: The alarm and response module includes an alarm trigger unit and an emergency response activation unit;

[0156] The alarm trigger unit determines the activation of the alarm program according to the result of the future event score SFu;

[0157] When 0.7 ≤ future event score SFu < 1, the alarm program is activated, and the security personnel and the management department are notified;

[0158] The emergency response activation unit notifies relevant campus management personnel according to the alarm program and activates the emergency response measures, and transmits the event information to relevant personnel by means of text messages, APP notifications and alarms;

[0159] According to the event types including fire, explosion and violent incidents, corresponding emergency plans are activated; and the response time Tre is calculated and obtained;

[0160] The response time Tre is obtained by multiplying the ratio of 1 to the future event score SFu and the preliminary judgment time Δti.

[0161] The result optimization and feedback module includes a feedback collection and analysis unit and a model adjustment and optimization unit;

[0162] The feedback collection and analysis unit collects the feedback information of alarm events, including whether the alarm event is accurately triggered, the effect of the response measures and the actual result of the event, analyzes the collected feedback information, and evaluates the effectiveness of the alarm trigger rules and emergency response measures;

[0163] The analysis content includes false alarms and missed alarms, the relationship between the response time Tre and the effectiveness of emergency measures, and the impact of environmental data and event types on security threats.

[0164] The evaluation formula for feedback data is:

[0165] FeedbackScore = f(Accuracy, Tre, EventOutcome);

[0166] Wherein, Accuracy represents the accuracy of the alarm, indicating whether it truly reflects the existence of a security threat after the alarm is triggered; Tre represents the response time, indicating the time difference from the alarm trigger to the start of the emergency response; EventOutcome represents the event handling result, indicating whether the event is resolved in a timely and effective manner, for example: successful handling, delayed response, unable to handle;

[0167] The alarm trigger error evaluation formula is:

[0168]

[0169] Wherein, FAR represents the false alarm rate, FA represents the number of false-triggered alarms, and TA represents the total number of triggered alarms;

[0170] The model adjustment and optimization unit adjusts the parameters in the regression model according to the feedback information;

[0171] According to the analysis result of the feedback information, improve the sensitivity of the alarm trigger, so that the model can more accurately identify risk events in the future, calculate and obtain the parameters βx of the new regression model, and adjust the training regression model;

[0172] Substitute the parameters βx of the new regression model into the formula of the regression model to obtain the new event score SFunew, and make a judgment and early warning.

[0173] The parameters βx of the new regression model are obtained through the following formula:

[0174] βx = βx + Δβ;

[0175] Wherein, Δβ represents the parameter difference;

[0176] The parameter difference Δβ is obtained through the following formula:

[0177] Δβ = SFu - 0.7;

[0178] The new event score SFunew is obtained through the following formula:

[0179]

[0180] In this embodiment, by introducing the future event score SFu, the system can automatically start the alarm program when the event score reaches a specific threshold, 0.7 ≤ SFu < 1. After the alarm trigger unit starts the alarm program, the system can promptly transmit the event information to relevant school administrators through multiple methods such as text messages, APP notifications, and alarms. This multi-channel notification mechanism can ensure that relevant personnel can receive the early warning information in a timely manner when an event occurs, avoiding the problems of information lag or untimely transmission.

[0181] By calculating the response time Tre, the system can accurately quantify the time difference from alarm trigger to the start of the emergency response and evaluate the timeliness of the emergency response. The calculation method of the response time Tre combines the event score and the preliminary judgment time Δti, enabling it to prepare in advance and optimize the response time based on the prediction of future events. By evaluating the response time, it is possible to identify and improve the links with overly long response times, thereby enhancing the overall emergency efficiency of the system.

[0182] Through the false alarm rate evaluation formula, the system can effectively identify false-triggered alarm situations and calculate the false alarm rate FAR. After evaluating the alarm accuracy, the system will adjust the parameters Δβ of the regression model according to the feedback information to improve the alarm accuracy. This method effectively reduces the false alarm rate, avoids unnecessary interventions and resource waste, and ensures that the alarm system only issues alarms when truly necessary, thereby enhancing the overall credibility of the campus security system.

[0183] Through the analysis of the feedback information, the system can adjust the parameter βx in the regression model to further improve the prediction accuracy of the event score. The parameter update of the new regression model enables the system to be optimized according to actual data and feedback in the future, ensuring more accurate prediction of potential risks. As the system operates deeper, it can continuously self-adjust and optimize to adapt to new types of security events that may occur in the campus environment, thereby continuously enhancing its intelligent level.

[0184] The system starts different emergency response plans according to the types of alarm events, including fires, explosions, and violent incidents, and conducts corresponding event handling. This flexible emergency response mechanism ensures that targeted handling can be quickly carried out for different types of events, avoiding the risk of event deterioration caused by improper handling. Through this flexible and efficient emergency response plan, the system can quickly switch different emergency measures, improving the timeliness and accuracy of event handling.

[0185] Through the real-time feedback and adjustment mechanism, the system can continuously optimize its alarm trigger mechanism, emergency response strategy, and regression model, making the overall campus security prevention and control work more predictive and responsive. Through continuous correction and optimization, the system can better handle various sudden security events, enhance the comprehensive ability of campus security management, and thus effectively reduce potential risks and hazards.

[0186] Although embodiments of the present invention have been shown and described, those of ordinary skill in the art will understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A campus security monitoring system, characterized in that: It includes data acquisition and preprocessing module, environmental data analysis module, monitoring video and environmental data fusion analysis module, security event prediction and evaluation module, alarm and response module and result optimization and feedback module; The data acquisition and preprocessing module is responsible for collecting data from multiple sensors and surveillance cameras, including environmental data ED and video stream data V, and performing preprocessing to obtain a data set W; The environmental data analysis module analyzes the environmental data ED in the data set W, analyzes the change trend of the environmental data ED, determines abnormal changes, and evaluates potential safety hazards; The monitoring video and environmental data fusion analysis module performs a comprehensive analysis on the data set W, analyzes the dynamic changes in the monitoring screen through image recognition technology, and calculates and obtains the security event score S in combination with the changes in the environmental data; The security event prediction and assessment module predicts future security threat events based on the security event score S and security risks; Predict future security risks and issue risk warnings through regression models; The alarm and response module automatically triggers alarms based on security threat events and risk warnings, notifies campus administrators, and initiates emergency response measures; The result optimization and feedback module readjusts the parameters of the regression model according to the emergency response measures and feedback information, and makes judgments and early warnings.

2. A campus security monitoring system according to claim 1, characterized in that: The data acquisition and preprocessing module includes a data acquisition unit and a data preprocessing unit; The data acquisition unit collects environmental data ED and video stream data V through a variety of environmental sensors and monitoring cameras, wherein the environmental data includes temperature T, humidity H, air quality index AQI and noise level L, and the various environmental sensors include temperature sensors, humidity sensors, air quality sensors and noise sensors. The temperature T is collected by the temperature sensor, the humidity H is collected by the humidity sensor, the air quality index AQI is collected by the air quality sensor, the noise level L is collected by the noise sensor, and the video stream data V is obtained by the monitoring camera; The data preprocessing unit cleans and standardizes the collected data to obtain a data set W; Data cleaning includes filling the data using interpolation and mean filling methods; data standardization includes normalizing the data to the same scale range using Min-Max normalization.

3. A campus security monitoring system according to claim 1, characterized in that: The environmental data analysis module includes a trend analysis unit and an anomaly detection and safety hazard assessment unit; The trend analysis unit is responsible for performing time series analysis on the environmental data ED, evaluating the temporal change trend of the environmental data ED, and identifying normal fluctuations and abnormal fluctuations of the environmental data ED by calculating the change rate of each parameter in the environmental data ED, including the temperature change rate ΔT, the humidity change rate ΔH, the air quality index change rate ΔAQI, and the noise level change rate ΔL; The temperature change rate ΔT is obtained by the ratio of the temperature difference at time t and the temperature at time t-1 to the temperature at time t-1; The humidity change rate ΔH is obtained by the ratio of the difference between the humidity at time t and the humidity at time t-1 and the humidity at time t-1; The air quality index change rate ΔAQI is obtained by the ratio of the difference between the air quality index at time t and the air quality index at time t-1 to the air quality index at time t-1; The noise level change rate ΔL is obtained by the ratio of the difference between the noise level at time t and the noise level at time t-1 to the noise level at time t-1; The anomaly detection and safety hazard assessment unit detects and assesses the acquired temperature change rate ΔT, humidity change rate ΔH, air quality index change rate ΔAQI and noise level change rate ΔL, and compares them with the temperature threshold ZT, humidity threshold ZH, air quality threshold ZAQI and noise threshold ZL, determines the abnormality of the environmental data ED, and assesses potential safety hazards, including fire risks, health risks and group conflicts and riots risks; When the temperature change rate ΔT> the temperature threshold ZT, it means that the temperature is abnormal, indicating fire risk and high temperature risk; When the humidity change rate ΔH>humidity threshold ZH, it means that the humidity is abnormal, indicating the risk of rainfall; When the air quality index change rate ΔAQI> the air quality threshold ZAQI, it means that the air quality index is abnormal, indicating a health risk; When the noise level change rate ΔL>noise threshold ZL, it means that the noise level is abnormal, indicating the risk of group conflict and riots.

4. A campus security monitoring system according to claim 1, characterized in that: The monitoring video and environmental data fusion analysis module includes an image recognition and event detection unit and an environmental data fusion and safety score calculation unit; The image recognition and event detection unit is responsible for real-time analysis of the video stream data V, detecting dynamic changes and potential security events in the video by using image recognition technology, and generating a security event feature vector EV, including personnel density Db, abnormal behavior identifier Ab and behavior duration Tb; The personnel density Db is obtained by the ratio of the number of personnel N appearing in the monitoring screen to the area of ​​the monitoring screen; The abnormal behavior identifier Ab identifies abnormal behaviors in the video, including violence, falling, and fast running, through a long short-term memory network, and outputs a binary identifier; The behavior duration Tb is obtained by monitoring the duration of the abnormal behavior.

5. A campus security monitoring system according to claim 4, characterized in that: The environmental data fusion and safety score calculation unit combines the safety event feature vector EV with the environmental data ED, calculates a comprehensive safety event score S through a fusion algorithm, and evaluates the impact of environmental conditions on safety hazards when a safety event occurs, including the impact of high temperature and poor air quality on crowd behavior; The security incident score S is calculated by weighted average method; The risk level of security incidents is divided according to the security incident score S; The risk levels are divided into three levels: When 0<security incident score S<0.3, it indicates the first risk level; When 0.3≤security incident score S<0.7, it indicates the second risk level; When 0.7≤security incident score S<1, it indicates the third risk level.

6. A campus security monitoring system according to claim 5, characterized in that: The security incident prediction and assessment module includes a historical data processing and feature extraction unit and a regression prediction and risk assessment unit; The historical data processing and feature extraction unit extracts features from the safety event score S and the safety hazard to obtain a feature set LF, including the change trend ΔS of the safety score, the frequency Efr of sudden safety events, the temperature T, the air quality index AQI and the noise level L, and the association between environmental data and safety events; The change trend ΔS of the safety score is obtained by the difference between the safety event score at time t and the safety event score at time t-1 and the ratio of the safety event score at time t-1; The frequency Efr of the sudden safety event is obtained by the ratio of the safety events occurring within time t to the total events occurring within time t; The association between the environmental data and security events includes the association between temperature and security events, the association between humidity and security events, the association between air quality and security events, and the association between noise and security events.

7. A campus security monitoring system according to claim 6, characterized in that: The regression prediction and risk assessment unit inputs the acquired feature set LF into the regression model to predict future security risks, predict future event scores SFu, and assess future security risk levels through future event scores SFu; The future event score SFu is obtained by calculating the regression model; The future security risk level is obtained by matching: When 0<future event score SFu<0.3, it indicates the first risk level, and environmental data, including temperature T, humidity H and air quality index AQI, are regularly monitored, and regular safety publicity and education activities are carried out to remind students and faculty to remain vigilant; When 0.3≤Future Event SFu<0.7, it indicates the second risk level, and the monitoring accuracy and frequency should be adjusted, the patrol frequency of security personnel should be adjusted, and emergency safety drills should be conducted regularly; When 0.7≤Future Event SFu<1, it indicates the third risk level and the emergency safety response mechanism is activated.

8. A campus security monitoring system according to claim 7, characterized in that: The alarm and response module includes an alarm triggering unit and an emergency response starting unit; The alarm triggering unit determines the start of the alarm procedure according to the result of the future event score SFu; When 0.7≤Future Event SFu<1, the alarm procedure is initiated and the security personnel and management department are notified; The emergency response initiation unit notifies relevant school management personnel and initiates emergency response measures according to the alarm procedure, and transmits event information to relevant personnel via SMS, APP notification and alarm; According to the event type, including fire, explosion and violent incidents, the corresponding emergency plan is activated and the response time Tre is calculated; The response time Tre is obtained by multiplying the ratio of 1 to the future event score SFu and the initial judgment time Δti.

9. A campus security monitoring system according to claim 1, characterized in that: The result optimization and feedback module includes a feedback collection and analysis unit and a model adjustment and optimization unit; The feedback collection and analysis unit collects feedback information of the alarm event, including whether the alarm event is accurately triggered, the effect of the response measures and the actual result of the event, analyzes the collected feedback information, and evaluates the effect of the alarm triggering rules and the emergency response measures; The analysis includes the situations of false positives and false negatives, the effectiveness of response time Tre and emergency measures, and the impact of environmental data and event types on security threats.

10. A campus security monitoring system according to claim 9, characterized in that: The model adjustment and optimization unit adjusts the parameters in the regression model according to the feedback information; According to the analysis results of the feedback information, the parameters βx of the new regression model are calculated and the training regression model is adjusted; The parameter βx of the new regression model is introduced into the formula of the regression model to obtain the new event score SFunew, and make a judgment and early warning.

Images