Internet of Things-based smart campus security monitoring methods and systems

Abstract

The present application relates to the technical field of intelligent monitoring, in particular to a campus safety monitoring method and system for a smart campus based on the Internet of Things, which collects video monitoring data of each area of a school, calculates the flow dynamic change entropy value of each video monitoring area data in the collection process, divides the key flow area and the non-key flow area in the school according to the entropy value, sets the collection frequency for the key flow area, and collects the image frame when the non-key flow area detects personnel movement. Through the dynamic collection of video monitoring data and the calculation of the flow dynamic change entropy value, the system can intelligently distinguish the key flow area and the non-key flow area in the campus. For these areas, the system can dynamically adjust the video monitoring collection frequency, collect at a higher frequency in the key area, and only collect data when personnel move in the non-key area, thereby optimizing the use of monitoring resources and data storage.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent monitoring technology, and particularly relates to a campus safety monitoring method and system for a smart campus based on the Internet of Things. BACKGROUND

[0002] The field of intelligent monitoring technology includes technology systems that utilize sensors, cameras, and data communication technology to monitor environments in real-time, collect data, and analyze it, thereby achieving functions such as safety management, behavior analysis, and data recording. The core content of this field involves connecting multiple monitoring devices and sensors through Internet of Things technology to achieve real-time transmission and processing of information. Intelligent monitoring is not only applied to homes and public safety, but is also widely used in schools, hospitals, and enterprises, with the goal of improving management efficiency and safety. Core technologies include image recognition, sensor data processing, data fusion, cloud computing, and edge computing, ensuring that the system can efficiently and accurately capture and analyze monitoring data and respond in a timely manner.

[0003] Among them, the campus safety monitoring for a smart campus based on the Internet of Things refers to a safety monitoring system specifically designed for campus environments, which uses Internet of Things technology to interconnect devices and collect a variety of data in real time, including video monitoring, environmental sensor data, access control systems, etc. This patent subject addresses multiple technical matters in campus safety monitoring, including how to connect cameras, sensors, and other devices to the central system through the network, how to collect and transmit these data, and how to monitor and process various types of information collected by sensors and cameras in real time. Through the deployment of Internet of Things devices and information interaction, the monitoring and early warning of various safety incidents within the campus are achieved, and the information is transmitted to the management end for analysis and processing through the data transmission channel.

[0004] Existing monitoring systems typically rely on fixed frequency video data collection and are unable to dynamically adjust monitoring frequency and areas according to actual conditions. Such systems handle monitoring areas roughly and are unable to intelligently distinguish between key areas and non-key areas, resulting in a large amount of unnecessary data collection and transmission, which not only wastes storage space but also increases the pressure on data processing. Moreover, existing technologies often lack detailed processing of real-time behavior analysis, although some abnormal situations can be detected, it is often difficult to predict and warn of the occurrence of abnormal behavior in a timely manner. Existing technologies also lack sufficient prediction of personnel movement paths, resulting in insufficient foresight and countermeasures for potential safety hazards. For example, when personnel gather or run abnormally, the system may not be able to identify this behavior pattern in a timely manner and provide accurate warnings, potentially missing critical prevention opportunities. These shortcomings make existing technologies difficult to meet the efficient and intelligent needs when faced with complex and changing campus safety management. SUMMARY

[0005] To address the technical problems existing in the prior art, embodiments of the present invention provide a campus security monitoring method and system for smart campuses based on the Internet of Things. The technical solution is as follows:

[0006] The IoT-based smart campus security monitoring method includes the following steps:

[0007] S1: Collect video surveillance data from each area of ​​the school, calculate the entropy value of the dynamic change of people flow in each video surveillance area during the data collection process, divide the school into key and non-key people flow areas based on the entropy value, set the collection frequency for key people flow areas, and collect image frames when people move in non-key people flow areas, assign weights according to the entropy value and dynamically adjust the collection frequency of the monitoring screen to generate dynamic monitoring data collection results.

[0008] S2: Based on the dynamic monitoring data acquisition results, cluster the real-time acquired video monitoring images, classify each behavioral characteristic of personnel on campus during the clustering process into categories, including walking, gathering, and running, and generate behavioral characteristic classification results;

[0009] S3: Based on the behavioral feature classification results, extract local feature points from the video surveillance image frames corresponding to each type of behavioral feature, including the starting point, turning point, and ending point of movement of the target personnel on campus. Refer to the local feature markers to mark the personnel located in the abnormal activity area on campus, and generate abnormal personnel identification results.

[0010] S4: Referring to the abnormal personnel identification results, extract all movement paths of the abnormal personnel within the school, predict the future movement paths of the abnormal personnel, calculate the trajectory residual between the current movement path and the future movement path, identify abnormal paths that exceed the preset residual threshold, and generate abnormal personnel path detection results.

[0011] S5: Based on the abnormal personnel path detection results, determine whether the abnormal personnel will continue to move towards the abnormal activity area in the future time period, and issue a campus safety warning through alarms or real-time notifications, generating campus safety monitoring results.

[0012] The present invention is improved in that the dynamic monitoring data acquisition results include regional entropy values, key and non-key pedestrian flow area division, acquisition frequency setting, and image frame acquisition adjustment; the behavioral feature classification results include walking, gathering, and running behavior categories; the abnormal personnel identification results include local feature points, abnormal activity area markers, and abnormal personnel behavior; the abnormal personnel path detection results include the current path, predicted path, and trajectory residual; and the campus security monitoring results include abnormal path warnings, abnormal personnel path reports, and real-time early warning notifications.

[0013] The present invention improves upon this by collecting video surveillance data from each area of ​​the school, calculating the dynamic entropy value of the data in each video surveillance area during the collection process, dividing the school into key pedestrian flow areas and non-key pedestrian flow areas based on the entropy value, setting a collection frequency for key pedestrian flow areas, and collecting image frames when people move in non-key pedestrian flow areas, allocating weights according to the entropy value and dynamically adjusting the collection frequency of the surveillance images, and generating dynamic monitoring data collection results. The specific steps are as follows:

[0014] S101: Collect video surveillance data from each area of ​​the school, and perform quantitative analysis of changes in personnel activity within the area by calculating the entropy value of dynamic changes in pedestrian flow, and generate personnel activity change analysis results;

[0015] S102: Divide each area according to the analysis results of the changes in personnel activity, compare the magnitude of the entropy value of the dynamic changes in the flow of people with the preset entropy threshold, determine the key flow areas and non-key flow areas, and generate the area division results;

[0016] S103: Based on the area division results, set the collection frequency for key pedestrian areas, and monitor the movement of people in non-key pedestrian areas. Collect image frames of moving areas in real time, and dynamically adjust the collection frequency according to data fluctuations to generate dynamic monitoring data collection results.

[0017] The present invention is improved by specifying that the entropy value of the dynamic change of pedestrian flow in each area is calculated according to the formula:

[0018]

[0019] Calculate the dynamically changing entropy value H(t);

[0020] Where H(t) represents the dynamic entropy value at time t, p i (t) represents the probability of the i-th behavior occurring in the region at time t, and n represents the total number of all possible behaviors in the region;

[0021] The weights are assigned based on the entropy value, and the data collection frequency is adjusted accordingly, using the following formula:

[0022]

[0023] Where r(t) is the sampling frequency at time t, w(t) is the weight calculated based on the dynamically changing entropy value, and f(t) is an additional time-sensitive parameter.

[0024] The present invention improves upon the above-mentioned dynamic monitoring data acquisition results by performing clustering processing on the real-time acquired video monitoring images, classifying each behavioral characteristic of personnel on campus into categories during the clustering process, including walking, gathering, and running. The specific steps for generating behavioral characteristic classification results are as follows:

[0025] S201: Extract the dynamic targets from the real-time video surveillance images in the dynamic monitoring data acquisition results, perform background segmentation, filter and mark the areas associated with personnel, and generate personnel area images;

[0026] S202: Based on the personnel area image, analyze the motion features in the image, identify each behavior pattern, extract the motion trajectory of each pattern and calibrate it, determine the walking, gathering and running behavior types, and generate a behavior feature classification.

[0027] S203: Classify each behavior according to the behavioral characteristics, obtain the occurrence frequency and duration of each behavior, and perform classification and labeling to generate behavioral characteristic classification results.

[0028] The present invention improves upon this by extracting local feature points from the video surveillance image frames corresponding to each type of behavioral feature based on the behavioral feature classification results. These local feature points include the starting point, turning point, and ending point of movement of the target personnel within the school. The specific steps for generating abnormal personnel identification results by referring to these local feature markers and identifying personnel located in areas of abnormal activity within the school are as follows:

[0029] S301: Based on the behavioral feature classification results, extract the movement trajectory of the person from the video surveillance image frame corresponding to each type of behavioral feature, identify the starting point, turning point and ending point of each trajectory, and generate motion feature points;

[0030] S302: Based on the motion feature points, track the movement path of personnel, analyze the start and end positions of the path, mark the location of abnormal activity areas, including wall areas, warehouse areas, closed areas, and areas where power equipment is located, and generate abnormal activity area markers.

[0031] S303: Based on the abnormal activity area markers and with reference to motion feature points, identify and mark personnel located within the abnormal area, complete the marking and recording of abnormal personnel, and generate abnormal personnel identification results.

[0032] The present invention is improved by extracting all movement paths of abnormal personnel within the school based on the abnormal personnel identification results, predicting the future movement paths of abnormal personnel, calculating the trajectory residuals between the current movement path and the future movement path, identifying abnormal paths exceeding a preset residual threshold, and generating abnormal personnel path detection results. The specific steps are as follows:

[0033] S401: Based on the abnormal personnel identification results, extract the movement path of the abnormal personnel within the school, record the node position of each path segment, obtain the time interval and spatial changes of each path node, and generate the movement path of the abnormal personnel.

[0034] S402: Based on the current abnormal personnel movement path, and referring to the current motion mode and video surveillance image environment, predict future trajectory changes and generate future movement path prediction results;

[0035] S403: Based on the predicted future movement path, calculate the trajectory residual between the current path and the predicted path and analyze the path differences, identify abnormal paths that exceed a preset threshold, and generate abnormal personnel path detection results.

[0036] The present invention is improved by using the following formula to calculate the trajectory residual between the predicted trajectory and the actual trajectory:

[0037]

[0038] Calculate the trajectory residual Δx(t) between the predicted trajectory and the actual trajectory;

[0039] Where Δx(t) represents the trajectory residual at time t, x p (t) and y p (t) represents the x and y coordinates of the predicted trajectory at time t, where x is the x and y coordinates. a (t) and y a λ(t) represents the horizontal and vertical coordinates of the actual trajectory at time t, λ(t) is the time-dependent adjustment coefficient, and σ is the dynamic uncertainty parameter of the trajectory point.

[0040] The present invention is improved in that, based on the abnormal personnel path detection results, it determines whether the abnormal personnel will continue to move towards the abnormal activity area in the future time period, and issues a campus security early warning through alarms or real-time notifications. The specific steps for generating campus security monitoring results are as follows:

[0041] S501: Based on the abnormal personnel path detection results, analyze the current path and abnormal activity area of ​​the abnormal personnel, and determine whether to continue approaching the target area based on the distance between the path and the target area and the direction of movement, and generate a path movement judgment result;

[0042] S502: Based on the path movement judgment result, assess the activities of the abnormal person in the future time period, estimate the time when the abnormal person arrives at the abnormal activity area, predict the future path change trend of the abnormal person, and generate future movement assessment results.

[0043] S503: Based on the future mobility assessment results, issue early warning information to security management personnel by triggering alarms or real-time notifications, and generate campus security monitoring results.

[0044] A smart campus security monitoring system based on the Internet of Things (IoT) includes:

[0045] The data acquisition module collects video surveillance data from each area of ​​the school, calculates the dynamic change entropy value of the data in each area, divides the school area into key pedestrian flow areas and non-key pedestrian flow areas based on the entropy value, sets the acquisition frequency for different areas, and collects image frames in non-key pedestrian flow areas by moving people, and assigns weights based on the dynamic change entropy value, dynamically adjusts the acquisition frequency, and generates dynamic monitoring data acquisition results.

[0046] Based on the dynamic monitoring data collection results, the behavior feature recognition module clusters real-time video images to identify the behavior features of people on campus, classifying them into walking, gathering, and running behavior types. By analyzing the category of each behavior feature, the module generates behavior feature classification results.

[0047] Based on the behavioral feature classification results, the abnormal personnel identification module extracts local feature points from the video surveillance image frames, identifies the target personnel's starting point, turning point, and ending point, and marks the abnormal activity area. Based on the marking information, it performs abnormal personnel identification and generates abnormal personnel identification results.

[0048] Based on the abnormal personnel identification results, the path detection module extracts the complete movement path of the abnormal personnel within the campus, uses relevant data on movement trajectory and behavior patterns to predict the path, calculates the trajectory residual between the current path and the predicted path, identifies abnormal paths that exceed a preset residual threshold, and generates abnormal personnel path detection results.

[0049] Based on the abnormal personnel path detection results, the security early warning module determines whether the abnormal personnel continue to move towards the abnormal activity area. Through the real-time monitoring and alarm system, it issues alarms or notifications to campus security management personnel in real time and generates campus security monitoring results.

[0050] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0051] In this invention, by dynamically acquiring video surveillance data and calculating the entropy value of dynamic changes in pedestrian flow, the system can intelligently distinguish between key and non-key pedestrian flow areas within the campus. For these areas, the system can dynamically adjust the video surveillance acquisition frequency, acquiring data at a higher frequency in key areas and only collecting data when people are moving in non-key areas, thereby optimizing the use of monitoring resources and data storage. This dynamic adjustment mechanism significantly improves monitoring efficiency, reduces unnecessary redundant data storage and transmission, and ensures timely and comprehensive monitoring of key areas. Through cluster analysis of behavioral characteristics, the system can identify different types of activities, such as walking, gathering, and running, and accurately identify abnormal behaviors. By extracting key points such as starting points, turning points, and ending points, and combining this with future path prediction, the system can identify potential abnormal behaviors or abnormal personnel paths in advance and issue real-time warnings. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a flowchart of the method of the present invention;

[0054] Figure 2 This is a detailed flowchart of step S1 of the present invention;

[0055] Figure 3 This is a detailed flowchart of step S2 of the present invention;

[0056] Figure 4 This is a detailed flowchart of step S3 of the present invention;

[0057] Figure 5 This is a detailed flowchart of step S4 of the present invention;

[0058] Figure 6 This is a detailed flowchart of step S5 of the present invention;

[0059] Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0060] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0061] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0062] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent.

[0063] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0064] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0065] Please see Figure 1 This invention provides a campus security monitoring method for smart campuses based on the Internet of Things, comprising the following steps:

[0066] S1: Collect video surveillance data from each area of ​​the school, calculate the entropy value of the dynamic change of people flow in each video surveillance area during the data collection process, divide the school into key and non-key people flow areas based on the entropy value, set the collection frequency for key people flow areas, and collect image frames when people move in non-key people flow areas, assign weights according to the entropy value and dynamically adjust the collection frequency of the monitoring screen to generate dynamic monitoring data collection results.

[0067] S2: Based on the dynamic monitoring data acquisition results, cluster the real-time acquired video monitoring images, classify each behavioral characteristic of the personnel on campus into categories, including walking, gathering, and running, and generate behavioral characteristic classification results;

[0068] S3: Based on the behavioral feature classification results, extract local feature points from the video surveillance image frames corresponding to each type of behavioral feature, including the starting point, turning point, and ending point of movement of the target personnel on campus. Refer to the local feature markers to mark the personnel located in the abnormal activity area on campus, and generate abnormal personnel identification results.

[0069] S4: Based on the abnormal personnel identification results, extract all movement paths of abnormal personnel within the school, predict the future movement paths of abnormal personnel, calculate the trajectory residual between the current movement path and the future movement path, identify abnormal paths that exceed the preset residual threshold, and generate abnormal personnel path detection results.

[0070] S5: Based on the abnormal personnel path detection results, determine whether the abnormal personnel will continue to move towards the abnormal activity area in the future time period, and issue campus safety warnings through alarms or real-time notifications, generating campus safety monitoring results.

[0071] The dynamic monitoring data collection results include regional entropy values, the division of key and non-key pedestrian flow areas, the setting of collection frequency, and the adjustment of image frame collection. The behavioral feature classification results include walking, gathering, and running behavior categories. The abnormal personnel identification results include local feature points, abnormal activity area markers, and abnormal personnel behavior. The abnormal personnel path detection results include the current path, predicted path, and trajectory residual. The campus security monitoring results include abnormal path warnings, abnormal personnel path reports, and real-time early warning notifications.

[0072] Please see Figure 2 The specific steps for collecting video surveillance data from each area of ​​the school, calculating the dynamic entropy value of the data in each video surveillance area during the collection process, dividing the school into key and non-key pedestrian flow areas based on the entropy value, setting the collection frequency for key pedestrian flow areas, and collecting image frames when people move in non-key pedestrian flow areas, assigning weights according to the entropy value, and dynamically adjusting the collection frequency of the monitoring images to generate dynamic monitoring data collection results are as follows:

[0073] S101: Collect video surveillance data from each area of ​​the school, and perform quantitative analysis of changes in personnel activity within the area by calculating the entropy value of dynamic changes in pedestrian flow, and generate personnel activity change analysis results;

[0074] Referring to the dynamic entropy value of pedestrian flow changes in each area, according to the formula:

[0075]

[0076] Calculate the dynamically changing entropy value H(t);

[0077] Here, H(t) represents the dynamic entropy value at time t, indicating the randomness of the behavior occurring within the region at time t. It is calculated by analyzing the probability of each behavior in the video, specifically by calculating the probability distribution of all behaviors in the video, thus deriving the entropy value. A higher entropy value indicates more complex or uncertain behavior in that region. i(t) represents the probability of the i-th behavior occurring within the region at time t, indicating the probability of the i-th behavior occurring at time t. This value is obtained by analyzing surveillance video and classifying behaviors. For example, assuming that within a certain period, surveillance video analysis identifies three behaviors: stationary, moving, and abnormal behavior, the monitoring system will count the frequency of each behavior and use it as the probability of that behavior. n represents the total number of all possible behaviors within the region. It is the number of behavior categories, determined by the number of behavior categories set in the surveillance video analysis. For example, if there are three behavior categories within a region (such as stationary, walking, and running), then n = 3. Weights are allocated based on entropy values, and the data collection frequency is adjusted using the following formula:

[0078]

[0079] Where r(t) is the acquisition frequency at time t, and w(t) is the weight calculated based on the dynamically changing entropy value. This weight reflects the magnitude of the impact of adjusting the video data acquisition frequency at that moment. It is implemented through a weighted strategy, with the weight typically set based on the entropy value H(t). Specifically, the calculation method involves dynamically adjusting the weight of the region by comparing the entropy values ​​of historical data and real-time data. For example, if the entropy value of a region is large, indicating significant changes in that region, the weight w(t) should be increased to increase the data acquisition frequency. f(t) is an additional time-sensitive parameter designed to adjust the acquisition frequency flexibly according to different time periods. This time-sensitive parameter adjusts the acquisition frequency flexibly based on specific time periods (such as day and night, weekdays and weekends). This parameter can be calculated through the relationship between historical behavioral data and time periods. For example, the system can analyze historical data to determine that certain areas have a higher probability of abnormal behavior occurring at night; therefore, f(t) can be set to a larger value during nighttime periods to increase the data acquisition frequency. This formula allows for the automatic adjustment of the monitoring frequency based on changes in regional entropy and the effects of time, thereby optimizing the video data acquisition process and improving system efficiency.

[0080] For example, if the probability of a behavior occurring in a certain time region is: p1 = 0.4 (stationary), p2 = 0.3 (moving), p3 = 0.3 (abnormal behavior), then calculate the entropy value at that time:

[0081] H(t)=-[0.4log(0.4)+0.3log(0.3)+0.3log(0.3)]≈1.029;

[0082] If the weight w(t) = 2 and the time-sensitive parameter f(t) = 1.5, then the sampling frequency is calculated as follows:

[0083]

[0084] The results show that, based on the comprehensive calculation of entropy, weight, and time-sensitive parameters, the system can adjust the acquisition frequency, making the video acquisition process more efficient. Furthermore, it can flexibly adjust the data acquisition frequency according to different time periods, thereby improving the accuracy and real-time performance of monitoring.

[0085] S102: Divide each area according to the analysis results of changes in personnel activity, compare the magnitude of the entropy value of the dynamic changes in the flow of people with the preset entropy threshold, determine the key flow areas and non-key flow areas, and generate the area division results;

[0086] The key criterion for zone division is the dynamic entropy value of each zone. The entropy value of each zone is calculated through analysis of video surveillance data, reflecting the complexity of changes in human behavior within that area. The criticality of a zone is determined based on its entropy value and a preset threshold range. For example, if the entropy value of a zone is 1.5, which may be greater than the threshold of 1.0, then that zone is considered a "critical pedestrian flow zone." According to monitoring data, the dynamic behavior of people in these zones is relatively complex and changes drastically, requiring high-frequency data collection to ensure real-time accuracy. Zones with lower entropy values ​​are identified as "non-critical pedestrian flow zones," allowing for a lower data collection frequency. The thresholds are set through analysis of historical and real-time data to ensure the distinction between critical and non-critical zones.

[0087] S103: Based on the area division results, set the collection frequency for key pedestrian areas, and conduct personnel movement monitoring for non-key pedestrian areas. Collect image frames of moving areas in real time, and dynamically adjust the collection frequency according to data fluctuations to generate dynamic monitoring data collection results.

[0088] Based on the zoning results, the system sets a higher data collection frequency for each key pedestrian flow area to ensure real-time capture of detailed human behavior within that area. These areas include high-traffic locations such as classroom entrances, dormitory corridors, and campus squares. For non-critical areas, such as parking lots or empty corridors in teaching buildings, the system relies on human movement monitoring to collect activity data in real time and adjusts the collection frequency based on data fluctuations. If human activity suddenly increases in a certain area, such as during holidays or specific events, the system will automatically increase the data collection frequency for that area to meet the monitoring needs of emergencies. Detection of data fluctuations typically relies on comparative analysis with historical data. If the current human activity frequency exceeds a certain baseline value, the collection frequency will increase accordingly, thereby achieving a dynamically optimized monitoring data collection process.

[0089] Please see Figure 3Based on the dynamic monitoring data acquisition results, clustering processing is performed on the real-time acquired video surveillance images to classify each behavioral characteristic of personnel on campus into categories, including walking, gathering, and running. The specific steps for generating behavioral characteristic classification results are as follows:

[0090] S201: Extract dynamic targets from real-time video surveillance images in the dynamic monitoring data acquisition results, perform background segmentation, filter and mark areas associated with personnel, and generate personnel area images;

[0091] The first step is to extract dynamic targets from real-time video surveillance images and perform background segmentation. Background segmentation technology processes the dynamic changes in each frame of the image, using algorithms such as frame differencing or background modeling based on Gaussian mixture models (GMMs) to separate static backgrounds from dynamic targets. The system determines which areas contain significant dynamic targets based on the changing regions within each frame. It then calculates the activity intensity (e.g., changes in grayscale values ​​or pixel counts) of these areas to determine if they belong to areas of human activity. During this process, parameters such as the area of ​​the dynamic target region and the magnitude of grayscale changes are used to filter areas related to human activity, based on preset thresholds. For example, when the change magnitude exceeds a set threshold, the system considers the area a dynamic target region and likely involves human activity, thus generating a human area image. The selection of dynamic targets is based on preset standards, ensuring that the system accurately identifies areas related to human activity during campus security monitoring.

[0092] S202: Based on the personnel area image, analyze the motion features in the image, identify each behavior pattern, extract and calibrate the motion trajectory of each pattern, determine the walking, gathering, and running behavior types, and generate a behavior feature classification.

[0093] Based on images of the personnel area, motion feature analysis identifies motion patterns in the images using optical flow or deep learning-based target detection techniques. By comparing the motion trajectories of targets in consecutive frames, the system calculates the motion features of each behavior pattern and classifies them accordingly. Specifically, walking behavior is characterized by small displacements and relatively stable speeds, gathering behavior is characterized by multiple target points moving and remaining within a specific area, and running behavior is characterized by large displacements and rapid changes. The system calibrates different motion features with behavior types and determines the behavior category using preset speed and displacement thresholds. For example, when the object's displacement exceeds a set step length threshold, the system identifies it as running behavior; while smaller displacements and slower speeds are classified as walking behavior. Through this analysis, the system can extract each type of behavior based on the motion trajectory and generate corresponding behavior feature classifications.

[0094] S203: Classify each behavior according to its behavioral characteristics, obtain the frequency and duration of each behavior, classify and label them, and generate behavioral characteristic classification results;

[0095] Based on behavioral characteristic classification, the system further analyzes the frequency and duration of each type of behavior. The frequency of a behavior is calculated by counting the number of times a particular behavior type occurs within a specific time period, while the duration is obtained by the difference between the start and end times of each behavior. For example, the frequency of walking behavior can be determined by counting the number of times a walking trajectory appears within 24 hours, while the duration of walking behavior is calculated based on the start and end times of each walking trajectory. After behavior classification, the system calibrates the frequency and duration of each behavior and performs further analysis based on this data. In actual monitoring, the system sets reasonable thresholds to determine which behaviors are high-frequency, long-duration, and which are low-frequency, short-duration. The goal of this process is to provide foundational data for subsequent system optimization and behavior prediction, ensuring that the campus security monitoring system can achieve real-time monitoring and alarm triggering of key areas without increasing redundant data collection.

[0096] Please see Figure 4 Based on the behavioral feature classification results, local feature points are extracted from the video surveillance image frames corresponding to each type of behavioral feature, including the starting point, turning point, and ending point of movement of the target personnel on campus. These local feature points are then used to mark personnel located in areas of abnormal activity on campus. The specific steps for generating abnormal personnel identification results are as follows:

[0097] S301: Based on the behavioral feature classification results, extract the movement trajectory of the person from the video surveillance image frame corresponding to each type of behavioral feature, identify the starting point, turning point and ending point of each trajectory, and generate motion feature points;

[0098] The process of extracting video surveillance image frames corresponding to each type of behavioral feature and generating motion feature points relies on the comparison of consecutive frame images. Through image processing techniques, such as optical flow and background modeling, the system can identify and extract dynamic targets in the images. The trajectory of each dynamic target is derived by calculating the motion changes between image frames, determining the starting point, turning point, and ending point of the target's movement in space. For the identification of motion trajectory points, the system determines when to identify a starting point, turning point, or ending point based on a set speed threshold and motion range. For example, a pixel distance threshold is set; when the target displacement exceeds this threshold, it is considered a new feature point. During this process, the system tracks and calibrates each target according to the maximum displacement threshold range set within the area, generating corresponding motion feature points to ensure accurate identification of each motion trajectory.

[0099] S302: Based on motion feature points, track the movement path of personnel, analyze the start and end positions of the path, mark the location of abnormal activity areas, including walled areas, warehouse areas, enclosed areas, and areas where power equipment is located, and generate abnormal activity area markers.

[0100] Based on motion feature points, personnel movement path tracking determines the start and end positions of the path by analyzing the distance and time differences between these points. The system determines the validity of a path by setting a maximum acceptable displacement range and a time delay threshold for each path. For example, if the displacement of consecutive feature points on the path exceeds a predetermined threshold of 30 pixels or the time interval exceeds 5 seconds, the system will determine that the path is interrupted and mark it as an abnormal path. For sensitive areas within the campus, such as walled areas, warehouses, and enclosed areas, the system will determine whether these areas have been accessed abnormally based on the path's start and end points. When a person's path overlaps with the location of these sensitive areas, the system can identify the area as an area of ​​abnormal activity and mark it. According to preset standards, the definition of an abnormal area is based on the persistence and frequency of personnel movement trajectories within these areas. For example, when the frequency of personnel appearing in an area exceeds a certain number of times, it is considered abnormal behavior.

[0101] S303: Based on the abnormal activity area markers and referencing motion feature points, identify and mark personnel located within the abnormal area, complete the marking and recording of abnormal personnel, and generate abnormal personnel identification results;

[0102] Based on the marking of abnormal activity areas, the system compares motion feature points with the area marking results to further identify and mark personnel located within the abnormal areas. The system matches the trajectory of personnel in real-time monitoring video with pre-marked abnormal activity areas to confirm whether personnel have entered designated high-risk areas. This process determines whether it constitutes abnormal activity by matching the temporal and spatial range of personnel movement trajectories. For example, if a person's trajectory enters a closed area and remains there for more than 30 seconds, the system will mark that person as an abnormal individual and generate an abnormal individual identification result. Through continuous motion trajectory tracking and comparison with area markings, the system can ensure timely and accurate identification of abnormal individuals, thereby further improving campus security capabilities.

[0103] Please see Figure 5 Based on the abnormal personnel identification results, the following are the specific steps for extracting all movement paths of abnormal personnel within the school, predicting their future movement paths, calculating the trajectory residuals between the current and future movement paths, identifying abnormal paths exceeding a preset residual threshold, and generating abnormal personnel path detection results:

[0104] S401: Based on the results of abnormal personnel identification, extract the movement path of the abnormal personnel within the school, record the node position of each path segment, obtain the time interval and spatial changes of each path node, and generate the movement path of the abnormal personnel.

[0105] Based on the results of abnormal person identification, when extracting the movement path of an abnormal person, the system first records the node positions of each path segment. The time interval and spatial changes of each path node are derived by the system from real-time monitoring data. For example, the system records the path step by step according to a set threshold (such as one node per second or every 10 meters of movement). The time interval of each path segment is calculated by the time difference between adjacent nodes, usually using inter-frame differencing to ensure that the position of each node can be accurately captured. In some cases, if the time interval between path nodes is too short (such as less than 1 second), it may be necessary to consider the fast movement speed between nodes and adjust the acquisition frequency to ensure the continuity and accuracy of data acquisition. Through these technical means, the system can generate the complete movement path of an abnormal person and ensure the timeliness and spatial accuracy of the path information.

[0106] S402: Based on the current abnormal personnel movement path, and referring to the current movement pattern and video surveillance image environment, predict future trajectory changes and generate future movement path prediction results;

[0107] Based on the movement paths of individuals exhibiting unusual behavior, the system predicts future trajectory changes by combining current movement patterns with the video surveillance environment. By analyzing the speed and direction of movement at each node along the current path, the system can predict future activities. For example, if a person moves quickly along a path on campus, the system predicts their possible future trajectory based on historical data and current movement trends. If an abnormal change occurs on a path, such as a sudden deceleration or change of direction, the system predicts possible future pauses or path changes by comparing historical trajectories with real-time data. During this process, the system dynamically adjusts the predicted future paths by setting time windows and spatial ranges (e.g., a 5-minute time window, where a path change exceeding a certain threshold is considered a trajectory change). This prediction helps the campus security system respond to potential risks or abnormal behavior in real time, improving the real-time performance and accuracy of monitoring.

[0108] S403: Based on the prediction results of future movement paths, calculate the trajectory residual between the current path and the predicted path and analyze the path differences, identify abnormal paths that exceed a preset threshold, and generate abnormal personnel path detection results.

[0109] The trajectory prediction data is compared with the collected movement trajectories of individuals exhibiting abnormal behavior, according to the formula:

[0110]

[0111] Calculate the trajectory residual Δx(t) between the predicted trajectory and the actual trajectory;

[0112] Where Δx(t) represents the trajectory residual at time t, x p (t) and y p (t) represents the x and y coordinates of the predicted trajectory at time t, where x is the x and y coordinates. a (t) and y a λ(t) represents the x and y coordinates of the actual trajectory at time t, and λ(t) is a time-dependent adjustment coefficient, indicating the flexibility of trajectory prediction adjustments within different time periods. The accuracy of trajectory prediction may vary with the time period. The dynamic adjustment coefficient is used to dynamically adjust the calculation weights, determined by monitoring behavioral pattern fluctuations within a specific time period. During the experiment, behavioral data from different time periods were analyzed to derive the adjustment coefficient for each time period. For example, when the frequency of personnel movement is high within a certain time period, the value of λ(t) may be set to a larger value. σ is the dynamic uncertainty parameter of the trajectory points, reflecting the stability and reliability of trajectory prediction. By introducing the dynamic adjustment coefficient λ(t) and the uncertainty parameter σ, the calculation of trajectory residuals can be further optimized, and the differences in prediction accuracy within different time periods can be considered. p (t) and y p (t) represent the x and y coordinates of the trajectory points predicted based on the movement patterns of individuals exhibiting abnormal behavior, respectively. a (t) and y a (t) represents the x and y coordinates of the actual trajectory points obtained through real-time monitoring data. The trajectory residual at each time step can be obtained by calculating the Euclidean distance between these points. A dynamic adjustment coefficient λ(t) is introduced to dynamically adjust the residual calculation based on data changes over a time period. This is particularly important because the accuracy of trajectory prediction may vary across different time periods (e.g., peak and off-peak periods), thus requiring an adjustment coefficient to ensure more accurate comparisons. The dynamic uncertainty parameter σ for trajectory points is introduced to account for the varying reliability of trajectory prediction under different environmental conditions (e.g., changes in lighting, occlusion), thereby increasing the flexibility of the calculation through this uncertainty parameter.

[0113] For example, at a certain moment, the coordinates of the predicted trajectory point are (x... p (5),y p (5))=(15,10), while the coordinates of the actual trajectory point are (x a (5),y a (5))=(12,9), and at that moment, the dynamic adjustment coefficient λ(5)=1.2, the uncertainty of the trajectory point σ=0.5, then the trajectory residual is:

[0114]

[0115] The results show that by introducing dynamic adjustment coefficients and trajectory point uncertainty, trajectory residuals can be calculated more accurately, especially when facing the influence of different time periods or environmental factors, thus enhancing the flexibility and accuracy of trajectory prediction. When the difference between the predicted trajectory and the actual trajectory reaches a certain threshold (e.g., exceeding 5 meters), it will be marked as an abnormal trajectory for further safety monitoring and early warning.

[0116] Please see Figure 6 Based on the abnormal personnel path detection results, it is determined whether the abnormal personnel will continue to move towards the abnormal activity area in the future, and campus security warnings are issued through alarms or real-time notifications. The specific steps for generating campus security monitoring results are as follows:

[0117] S501: Based on the abnormal personnel path detection results, analyze the current path and abnormal activity area of ​​the abnormal personnel, and determine whether to continue approaching the target area based on the distance between the path and the target area and the direction of movement, and generate path movement judgment results;

[0118] Based on abnormal personnel path detection results, the system calculates the distance between the abnormal personnel's current path and the target area, as well as their direction of movement, to determine whether they should continue approaching the target area. The system first uses real-time monitoring data to calculate the spatial distance between the personnel's current position and the target area by measuring the time interval and displacement of each path node. For example, when the personnel are less than 5 meters from the target area, the system further analyzes whether they are moving towards the target area. If the personnel's current direction of movement matches the target area and the path distance gradually decreases, the system determines that they are approaching the target area and generates a path movement judgment result. This process is based on a set threshold; for example, if the target area has a set radius of 10 meters, the system defines this area as the approach zone. When the path distance is less than the threshold and the direction is consistent, it is determined that the personnel may enter the target area.

[0119] S502: Based on the path movement judgment results, assess the activities of abnormal personnel in the future time period, estimate the time when the abnormal personnel arrive at the abnormal activity area, predict the future path change trend of the abnormal personnel, and generate future movement assessment results.

[0120] Based on the path movement judgment results, the system assesses the activities of abnormal individuals within a future time period. Through real-time data and historical movement trajectory analysis, the system can estimate the time it takes for abnormal individuals to reach the target area and infer future path change trends based on their speed, acceleration, and other data. For example, if an individual's movement speed has been 2 meters per second in the past few minutes without significant deceleration, the system will predict that they will reach the target area within the next 5 minutes. The system further monitors the abnormal individual's movement status; if their speed significantly slows down or they suddenly stop, the system will adjust the prediction time accordingly and make an assessment based on the current path change trend. When the path undergoes abnormal changes (such as a change in direction or a decrease in speed), the system will promptly update the expected arrival time and readjust the predicted path.

[0121] S503: Based on future mobility assessment results, issue early warning information to security management personnel by triggering alarms or real-time notifications, and generate campus security monitoring results;

[0122] Based on future movement assessments, the system sends early warning information to security management personnel via alerts or real-time notifications. When the system predicts that unusual individuals may be approaching a target area and meets preset conditions, it will automatically trigger an alert. For example, if an individual's path is less than 5 meters from the target area and they are expected to arrive within 5 minutes, the system will send an alert to security management personnel, prompting them to take appropriate emergency measures. Furthermore, the system will adjust the alert level based on actual conditions. If the proximity of an individual's path to the target area gradually increases, or if their arrival time is brought forward, the alert will be prioritized and sent to campus security management personnel via real-time notification to ensure timely response and handling of potential security threats.

[0123] Please see Figure 7 A smart campus security monitoring system based on the Internet of Things (IoT) includes:

[0124] The data acquisition module collects video surveillance data from each area of ​​the school, calculates the dynamic change entropy value of the data in each area, divides the school area into key pedestrian flow areas and non-key pedestrian flow areas based on the entropy value, sets the acquisition frequency for different areas, and collects image frames in non-key pedestrian flow areas by moving people, and assigns weights based on the dynamic change entropy value, dynamically adjusts the acquisition frequency, and generates dynamic monitoring data acquisition results.

[0125] The behavioral feature recognition module clusters real-time video images based on dynamic monitoring data collection results, identifies the behavioral features of people on campus, and classifies them into walking, gathering, and running behavior types. By analyzing the category of each behavioral feature, behavioral feature classification results are generated.

[0126] The abnormal personnel identification module extracts local feature points from video surveillance image frames based on behavioral feature classification results, identifies the starting point, turning point, and ending point of the target personnel's movement, marks the abnormal activity area, and performs abnormal personnel identification based on the marking information to generate abnormal personnel identification results.

[0127] Based on the abnormal personnel identification results, the path detection module extracts the complete movement path of the abnormal personnel within the campus, uses relevant data on movement trajectory and behavior patterns to predict the path, calculates the trajectory residual between the current path and the predicted path, identifies abnormal paths that exceed the preset residual threshold, and generates abnormal personnel path detection results.

[0128] Based on the abnormal personnel path detection results, the security early warning module determines whether the abnormal personnel continue to move towards the abnormal activity area. Through the real-time monitoring and alarm system, it issues alarms or notifications to campus security management personnel in real time and generates campus security monitoring results.

[0129] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A campus safety monitoring method for a smart campus based on the Internet of Things, characterized in that, The method comprises the following steps: S1: Collecting video monitoring data of each area of the school, calculating the dynamic change entropy value of the data of each video monitoring area in the collection process, dividing the key and non-key personnel flow areas in the school according to the entropy value, setting the collection frequency for the key personnel flow area, collecting the image frames when detecting personnel movement in the non-key personnel flow area, dynamically adjusting the collection frequency of the monitoring picture according to the weight distribution of the entropy value, and generating a dynamic monitoring data collection result; S2: Based on the dynamic monitoring data collection result, clustering processing is performed on the real-time acquired video monitoring image, the categories of each behavior feature of the school personnel in the clustering process are divided, including walking, gathering and running, and a behavior feature classification result is generated; S3: Based on the behavior feature classification result, local feature points are extracted from the image frames corresponding to each type of behavior feature, including the moving starting point, turning point and moving ending point of the target personnel in the school, the personnel in the abnormal activity area in the school are marked with reference to the local feature points, and an abnormal personnel identification result is generated; S4: Referring to the abnormal personnel identification result, the entire moving path of the abnormal personnel in the school is extracted, the future moving path of the abnormal personnel is predicted, the trajectory residual error between the current moving path and the future moving path is calculated, the abnormal path exceeding the preset residual error threshold is identified, and an abnormal personnel path detection result is generated; S5: According to the abnormal personnel path detection result, it is judged whether the abnormal personnel will continue to move to the abnormal activity area in the future time period, and the campus safety warning is carried out through the alarm or real-time notification, and a campus safety monitoring result is generated; The dynamic change entropy value of the personnel flow in each area is calculated according to the formula: ； Computing a dynamically changing entropy value ; wherein, denotes time the dynamic change of the entropy value of the time is the probability of the occurrence of the kind of behavior in the time zone, denotes the total number of all possible behaviors in the zone; The weight distribution is adjusted according to the entropy value, and the data collection frequency is adjusted according to the formula: ； wherein, is time the collection frequency of the time, is a weight calculated according to the dynamically changing entropy value, is an additional introduced time-sensitive parameter. 2.The campus safety monitoring method based on the Internet of Things for a smart campus according to claim 1, characterized in that: The dynamic monitoring data collection result includes area entropy value, key and non-key personnel flow area division, collection frequency setting, image frame collection adjustment, the behavior feature classification result includes walking, gathering and running behavior categories, the abnormal personnel identification result includes local feature points, abnormal activity area marking and personnel abnormal behavior, the abnormal personnel path detection result includes current path, predicted path and trajectory residual error, and the campus safety monitoring result includes abnormal path warning, personnel path abnormal report and real-time warning notification. 3.The campus safety monitoring method based on the Internet of Things for a smart campus according to claim 1, characterized in that: The specific steps of collecting the video monitoring data of each area of the school, calculating the dynamic change entropy value of the data of each video monitoring area in the collection process, dividing the key and non-key personnel flow areas in the school according to the entropy value, setting the collection frequency for the key personnel flow area, collecting the image frames when detecting personnel movement in the non-key personnel flow area, dynamically adjusting the collection frequency of the monitoring picture according to the weight distribution of the entropy value, and generating a dynamic monitoring data collection result are as follows: S101: Collecting video monitoring data of each area of the school, quantitatively analyzing the personnel activity change in the area by calculating the dynamic change entropy value, and generating a personnel activity change analysis result; S102: According to the personnel activity change analysis result, each region is divided, the size of the human flow dynamic change entropy value is compared with the preset entropy value threshold, the key human flow region and the non-key human flow region are judged, and the region division result is generated; S103: According to the region division result, the collection frequency of the key human flow region is set, the personnel movement monitoring is carried out in the non-key human flow region, the image frames of the moving region are collected in real time, and the collection frequency is dynamically adjusted according to the data fluctuation, and the dynamic monitoring data collection result is generated. 4.The campus safety monitoring method based on the Internet of Things for a smart campus according to claim 1, characterized in that: Based on the dynamic monitoring data collection result, the real-time acquired video monitoring image is clustered, the category of each behavior feature of the on-campus personnel in the clustering process is divided, including walking, gathering, running, and the specific steps of generating the behavior feature classification result are as follows: S201: Extract the dynamic target of the real-time video monitoring image in the dynamic monitoring data collection result, and perform background segmentation, screen the region associated with personnel and mark, and generate a personnel region image; S202: According to the personnel region image, the motion features in the image are analyzed, each behavior mode is recognized, the motion trajectory of each mode is extracted and calibrated, the walking, gathering, and running behavior types are judged, and the behavior feature classification is generated; S203: According to the behavior feature classification, each type of behavior is divided, the occurrence frequency and duration of each type of behavior are obtained, and classification calibration is performed, and the behavior feature classification result is generated. 5.The campus safety monitoring method based on the Internet of Things for a smart campus according to claim 1, characterized in that: Based on the behavior feature classification result, local feature points are extracted from the image frames corresponding to each type of behavior feature in the video monitoring, including the moving starting point, turning point and moving ending point of the on-campus target personnel, and the personnel located in the on-campus abnormal activity region are marked with reference to the local feature, and the specific steps of generating the abnormal personnel identification result are as follows: S301: Based on the behavior feature classification result, the motion trajectory of personnel is extracted from the video monitoring image frames corresponding to each type of behavior feature, the starting point, turning point and ending point of each trajectory are recognized, and the motion feature points are generated; S302: Based on the motion feature points, the moving path of the personnel is tracked, the starting and ending positions of the path are analyzed, the position of the abnormal activity region is calibrated, the abnormal activity region includes the wall region, the warehouse region, the closed region and the region where the power equipment is located, and the abnormal activity region mark is generated; S303: According to the abnormal activity region mark, the motion feature points are referred to, the personnel located in the abnormal region are recognized and marked, the calibration and recording of the abnormal personnel are completed, and the abnormal personnel identification result is generated. 6.The campus safety monitoring method based on the Internet of Things for a smart campus according to claim 1, characterized in that: Referring to the abnormal personnel identification result, the whole moving path of the abnormal personnel in the campus is extracted, the future moving path of the abnormal personnel is predicted, the trajectory residual error of the current moving path and the future moving path is calculated, the abnormal path exceeding the preset residual error threshold is recognized, and the specific steps of generating the abnormal personnel path detection result are as follows: S401: Based on the abnormal personnel identification result, the moving path of the abnormal personnel in the campus is extracted, the node position of each path is recorded, the time interval and space change of each path node are obtained, and the abnormal personnel moving path is generated; S402: According to the current abnormal personnel movement path, referring to the current motion mode and the video monitoring image environment, the future trajectory change is predicted, and the future movement path prediction result is generated; S403: According to the future movement path prediction result, the trajectory residual between the current path and the predicted path is calculated and the path difference is analyzed, the abnormal path exceeding the preset threshold is identified, and the abnormal personnel path detection result is generated. 7.The campus security monitoring method based on the Internet of Things for a smart campus according to claim 1, characterized in that: For the trajectory residual of the predicted trajectory and the current trajectory, the formula is used: ； Computing pre-at-time Trajectory residual at time instant ; wherein, and are the horizontal and vertical coordinates of the predicted trajectory at time , and are the horizontal and vertical coordinates of the actual trajectory at time , is a time-dependent adjustment factor, is a dynamic uncertainty parameter of the trajectory point. 8.The campus security monitoring method based on the Internet of Things for a smart campus according to claim 1, characterized in that: According to the abnormal personnel path detection result, it is judged whether the abnormal personnel continues to move to the abnormal activity area in the future time period, and the campus safety warning is carried out through the alarm or real-time notification, and the specific steps of generating the campus safety monitoring result are as follows: S501: Based on the abnormal personnel path detection result, the current path of the abnormal personnel and the abnormal activity area are analyzed, and according to the distance and moving direction of the path and the target area, it is judged whether to continue to approach the target area, and the path movement judgment result is generated; S502: Based on the path movement judgment result, the activity of the abnormal personnel in the future time period is evaluated, the time of the abnormal personnel to reach the abnormal activity area and the change trend of the future path of the abnormal personnel are calculated, and the future movement evaluation result is generated; S503: According to the future movement evaluation result, the warning information is sent to the safety management personnel through triggering the alarm or real-time notification, and the campus safety monitoring result is generated.

9. A campus safety monitoring system based on the Internet of Things for a smart campus, characterized in that, The campus safety monitoring method based on the Internet of Things is executed according to any one of claims 1-8, and the system comprises: The data acquisition module acquires the video monitoring data of each area of the school, calculates the dynamic change entropy value of each area data, divides the campus area into key and non-key personnel flow areas according to the entropy value, sets different acquisition frequencies for different areas, and acquires image frames through personnel movement in non-key personnel flow areas, and dynamically adjusts the acquisition frequency according to the dynamic change entropy value to generate dynamic monitoring data acquisition result; The behavior feature recognition module clusters the real-time video image based on the dynamic monitoring data acquisition result, recognizes the behavior features of the personnel in the school, divides the behavior types into walking, gathering and running, analyzes each behavior feature category, and generates behavior feature classification result; The abnormal personnel recognition module extracts local feature points from the image frames of the video monitoring based on the behavior feature classification result, identifies the moving starting point, turning point and ending point of the target personnel, and marks the abnormal activity area, and performs abnormal personnel recognition based on the marking information to generate abnormal personnel recognition result; The path detection module extracts the complete movement path of the abnormal personnel in the campus based on the abnormal personnel recognition result, uses the related data of the movement trajectory and the behavior mode to predict the path, calculates the trajectory residual between the current path and the predicted path, identifies the abnormal path exceeding the preset residual threshold, and generates the abnormal personnel path detection result; The security warning module judges whether the abnormal personnel continue to move to the abnormal activity area based on the abnormal personnel path detection result, issues an alarm or a notification through real-time monitoring and an alarm system, and notifies the campus security management personnel in real time to generate a campus security monitoring result.

Images