All-domain safety monitoring method and system based on Internet of Things

Abstract

The invention discloses a global security monitoring method and system based on the Internet of Things, and belongs to the technical field of the Internet of Things, and the method comprises the steps: obtaining an electronic map of a target place, obtaining the number and position data of Internet of Things nodes of each partition, and generating the configuration data of an edge calculation terminal; the edge computing terminal identifies an abnormal condition and sends a patrol instruction to the security terminals in the partitions; the cloud fusion platform constructs a stampede risk prediction model to predict a stampede risk level, and if the stampede risk level exceeds an early warning level, an evacuation request is sent to the management terminal; and generating a crowd evacuation path plan based on the multi-channel spatial-temporal feature matrix and the trampling risk level, and sending a guide instruction to a security terminal to generate a visual interface. According to the global safety monitoring method and system based on the Internet of Things provided by the invention, full-coverage, real-time and refined risk prediction, early warning and management of crowded places are realized, and the crowd safety management efficiency and accuracy are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of Internet of Things technology, and in particular to a method and system for global security monitoring based on the Internet of Things.

Background Art

[0002] With the acceleration of the urbanization process, various large-scale activities are held more and more frequently. The safety management of crowded areas such as stadiums, convention centers, and temporary meeting venues faces increasingly severe challenges. Especially the occurrence of stampedes not only seriously threatens the personal safety of the public, but also may cause major social impacts and economic losses.

[0003] Existing technologies usually adopt a single technical means, such as manual inspection. The monitoring range is limited and the real-time performance is insufficient, and it is impossible to achieve comprehensive and refined risk prediction and effective early warning for crowded places.

[0004] Therefore, how to establish a global security monitoring method and system with comprehensive coverage, sufficient data fusion, and rapid and accurate early warning response has become a technical problem to be solved urgently in the current field.

Summary of the Invention

[0005] In view of this, the embodiments of the present invention provide a method and system for global security monitoring based on the Internet of Things.

[0006] In a first aspect, the embodiments of the present invention provide a method for global security monitoring based on the Internet of Things, and the method includes:

[0007] S1. Obtain the electronic map of the target venue, determine the core area, non-core area, and peripheral area of the target venue, and obtain the quantity and position data of the Internet of Things nodes in each partition to generate the configuration data of the edge computing terminal;

[0008] S2. The edge computing terminal collects the Internet of Things data sent by the Internet of Things nodes in the partition where it is located, identifies abnormal conditions, sends a patrol instruction to the security end in the partition, and uploads the processed result sent by the security end and the Internet of Things data to the cloud fusion platform after packaging;

[0009] S3. The cloud fusion platform comprehensively processes the data sent by the edge computing terminal, generates a global multi-channel spatio-temporal feature matrix, constructs a stampede risk prediction model to predict the stampede risk level, and sends an evacuation request to the management terminal if it exceeds the warning level;

[0010] S4. After receiving the evacuation instruction sent by the management terminal, generate a crowd evacuation path plan based on the multi-channel spatio-temporal feature matrix and the stampede risk level, send a guiding instruction to the security end, track the trajectory of the target personnel, and send the comprehensive data sorted out by the cloud fusion platform to the management terminal to generate a visual interface.

[0011] For the aspects and any possible implementation manners described above, a further implementation manner is provided. Specifically, S1 includes:

[0012] S10. Obtain the electronic map of the target venue, extract the boundary of the target venue through edge detection and connected component analysis algorithms, and sequentially divide the core area, non-core area, and peripheral area from the inside out;

[0013] S11. Send the automatic zoning result to the management terminal, and the management terminal reviews, adjusts, and finally confirms the zoning boundary;

[0014] S12. Obtain the quantity and location data of the emergency alarms, sound sensors, smoke sensors, and Bluetooth AoA base stations installed in the core area;

[0015] S13. Obtain the quantity and location data of the cameras and Bluetooth RSSI base stations installed in the non-core area;

[0016] S14. Obtain the quantity and location data of the cameras installed in the peripheral area;

[0017] S15. Generate the installation location data of the edge computing terminal based on the quantity and location data of the Internet of Things nodes in the core area, non-core area, and peripheral area.

[0018] For the aspects and any possible implementation manners described above, a further implementation manner is provided. Specifically, S15 includes:

[0019] S150. For the core area, use the K-means clustering algorithm to find the clustering centers, , where represents the spatial coordinates of the i-th node, and the total number of Internet of Things nodes in the core area is nodes, represents the set of nodes included in the j-th cluster, represents the central position of the j-th cluster, represents the preliminary estimated value of the number of clustering centers in the core area;

[0020] After clustering, each clustering center serves as the installation location of a core area edge computing terminal;

[0021] The optimal number of clusters is determined by the elbow method, which is the number of core area edge computing terminals;

[0022] S152. Use the grid division method to define the non-core area and calculate the grid size, , where represents the size of each grid, represents the total area of the non-core area, represents the total number of nodes in the non-core area, represents a set constant;

[0023] For the set of node positions within each grid , calculate the geometric center, , represents the installation position of the edge computing terminal within the j-th grid, represents the specific position coordinates of the nodes within the grid, represents the number of nodes within the j-th grid;

[0024] If each grid contains at least one node, then install an edge computing terminal at the geometric center of the grid. The total number of final edge terminals is the total number of grids containing nodes;

[0025] S153. Obtain the position data of all camera nodes in the peripheral area, and their respective spatial positions are , represents the total number of cameras in the peripheral area;

[0026] Sort the cameras in spatial order along the boundary of the peripheral area, and group the cameras according to the set maximum number of cameras per group ;

[0027] Calculate the geometric center of each group of cameras as the installation position of the edge terminal. For the f-th group, , represents that the f-th group contains cameras, represents the coordinates of the k-th camera within the group;

[0028] Install 1 edge computing terminal at each center point . The final number of terminals is equal to the number of camera groups;

[0029] S154. Summarize the positions and quantities of the edge computing terminals.

[0030] For the aspects and any possible implementation manners as described above, a further implementation manner is provided. The specific steps of S2 are as follows:

[0031] S20. The edge computing terminal collects the status of emergency alarms, sound data, smoke sensor concentration, and personnel positioning data obtained by Bluetooth AoA base station positioning within the core area, collects camera video data and personnel positioning data obtained by Bluetooth RSSI base station positioning within the non-core area, and collects camera video data within the peripheral area, and performs standardized preprocessing on various types of data. The personnel within the core area and the security personnel within the core area and non-core area carry Bluetooth tags;

[0032] S21. Identify abnormal conditions in the core area. Among them, the abnormal judgment formula for the core area is as follows: , , represents the comprehensive anomaly score of the core area, represents the status of the emergency alarm, normal is 0, alarm is 1, represents the abnormal sound score, normal volume is 0, abnormal volume is 1, represents the concentration of the smoke sensor, represents the smoke concentration alarm threshold, represents the abnormal value of personnel movement of the Bluetooth RSSI base station, , , and represent the weights of each index, represents the density change rate per unit time, represents the average speed change rate of personnel per unit time;

[0033] S22. Divide the abnormal level of the core area and process it. When it is determined as the first level, send a patrol instruction to all security terminals within the core area, and send a reinforcement instruction to the security terminals in the non-core area, it is determined as the second level, send a patrol instruction to the security terminals within the first preset range from the abnormal point, it is determined as the third level, send a inspection instruction to the security terminal closest to the abnormal point, it is determined as normal.

[0034] In the aspects and any possible implementation manners as described above, a further implementation manner is provided. The S2 further includes:

[0035] S23. Real-time obtain the video data of the cameras in the non-core area and the peripheral area, and preprocess the video;

[0036] S24. Use the target detection deep learning model to identify the human heads in the image frame, count the crowd density in the coverage area of each camera according to the detection results, perform optical flow analysis based on consecutive frame images, calculate the average movement speed of the crowd, calculate the average speed through the optical flow motion vector field, and calculate the comprehensive anomaly score of the non-core area based on the crowd density and the average speed. Among them, , , , represents the comprehensive anomaly score of the non-core area, represents the set safe crowd density threshold of the non-core area, represents the safe speed threshold of the non-core area, and represents the weight and , represents the number of people detected by the i-th camera, represents the area covered by the i-th camera, and represents the horizontal and vertical components of the optical flow motion vector of the k-th person, represents the number of people tracked within the area;

[0037] S25. Divide the abnormal levels of the non-core areas and process them. When it is determined as level one, send a patrol instruction to all security terminals within the core area and a reinforcement instruction to the security terminals in the peripheral area. it is determined as level two, send a patrol instruction to the security terminals within the second preset range from the abnormal point. it is determined as level three, send an inspection instruction to the security terminal closest to the abnormal point. it is determined as normal;

[0038] S26. Calculate the comprehensive abnormal score of the peripheral area , where , represents the security density threshold of the peripheral area, represents the security speed threshold of the peripheral area, and represents the weight and ;

[0039] S27. Divide the abnormal levels of the peripheral area and process them. When it is determined as level one, send a patrol instruction to all security terminals within the core area and a reinforcement instruction to the security terminals in the non-core area. it is determined as level two, send an inspection instruction to the security terminal closest to the abnormal point. it is determined as level three, record the event. it is determined as normal;

[0040] S28. Receive the processing results sent by the security terminals, package them with the Internet of Things data, and upload them to the cloud fusion platform.

[0041] In the aspects and any possible implementation manners as described above, a further implementation manner is provided. In step S3, the cloud fusion platform comprehensively processes the data sent by the edge computing terminals, specifically including:

[0042] S30. The cloud fusion platform receives the data sent by the edge computing terminals in each partition, including the abnormal score levels, occurrence locations, times, and security terminal processing feedback results of the abnormal conditions in the core area, non-core area, and peripheral area;

[0043] S31. Generate a list of unprocessed events based on the occurrence location, time, and security terminal processing feedback results, and define the security terminal processing result coefficient , the security terminal processing result coefficient It is the ratio of the number of security terminals that have been processed to the total number of security terminals that have received inspection instructions;

[0044] S32. Define the emergency event determination index , where , represents the weight coefficient and satisfies , represents defining the security terminal processing result coefficient, represents the natural constant;

[0045] S33. Set the emergency event alarm threshold , if , it is determined as an emergency event, immediately alarm the management terminal and send the list of unprocessed events, otherwise only send the list of unprocessed events without alarming.

[0046] In the aspects and any possible implementation manners as described above, a further implementation manner is provided. In S3, a global multi-channel spatio-temporal feature matrix is generated, a stampede risk prediction model is constructed to predict the stampede risk level, and if the warning level is exceeded, an evacuation request is sent to the management terminal, specifically including:

[0047] Divide the entire area of the target place into unified grids of a preset size, and the spatial grid numbers are , and represent the grid horizontal and vertical coordinate indices;

[0048] For the core area, directly calculate the number of people, average speed and flow direction in each grid according to the AoA data. For the non-core area and the peripheral area, calculate the number of people, average speed and flow direction in each grid according to the camera video data;

[0049] Fuse the data of the core area, non-core area and peripheral area, and dynamically generate a global multi-channel spatio-temporal feature matrix at each time step, with the density size represented by color, the arrow direction representing the flow direction, and the arrow thickness representing the speed size;

[0050] Construct a stampede risk prediction model based on the STCNN+ConvLSTM structure, and the input features are , where represents the time length, and represent the spatial grid dimension, represents the number of features, including density, speed and direction, and the output is a 4-level risk level probability matrix under the global spatial grid, including safe, concerned, warning and dangerous;

[0051] Based on the actual stampede events at historical sites, mark the real risk level of each event, obtain the multi-channel spatio-temporal feature matrix in the preset time period before the event occurs according to the time series, and manually mark the risk level as the supervised training signal;

[0052] Divide the dataset into a training set and a validation set, conduct training and validation, and obtain an optimized stampede risk prediction model;

[0053] Use the trained stampede risk prediction model to predict the stampede risk level in real time at each time step;

[0054] Take the risk level with the highest probability as the current stampede risk level of the grid, and generate a global risk distribution map based on the risk levels of all grids;

[0055] If the risk in the global risk distribution map exceeds the warning level, evacuation is carried out.

[0056] In the aspect and any possible implementation manner as described above, a further implementation manner is provided. In S4, based on the multi-channel spatio-temporal feature matrix and the stampede risk level, a crowd evacuation path plan is generated and a guiding instruction is sent to the security end, which specifically includes:

[0057] Form a high-risk block by combining grids with continuous distribution and a number greater than the threshold and with dangerous and warning levels, sort the blocks according to the block risk level, where dangerous > warning > attention > safe, and sort the blocks within the same level according to the block area, where the larger the area, the higher the priority, and generate a block evacuation priority list;

[0058] Each priority block is used as a set of evacuation starting points, and the preset safe exit positions are used as a set of evacuation end points. The Dijkstra algorithm is used to plan the path for each block in the block evacuation priority list. The optimal or shortest path avoids grids with dangerous and warning levels, and conflict detection is carried out: if the paths of high-priority blocks and low-priority blocks overlap, the high-priority blocks enter the path first, and the low-priority blocks can enter the overlapping area only after the high-priority blocks have evacuated;

[0059] Send a guiding instruction to the security end within the preset range of the target evacuation block. The guiding instruction includes: the number of the evacuation starting block, the coordinate sequence of the recommended path, the current evacuation priority, the real-time evacuation / waiting instruction, the number of evacuees, and the estimated time used.

[0060] In the aspect and any possible implementation manner as described above, a further implementation manner is provided. In S4, the trajectory of the target personnel is tracked, and the comprehensive data sorted out by the cloud fusion platform is sent to the management terminal to generate a visual interface, which specifically includes:

[0061] Use Bluetooth AoA positioning in the core area and RSSI positioning in the non-core area to track the evacuation trajectory of core area personnel carrying Bluetooth tags and security personnel in the core and non-core areas;

[0062] Determine whether a preset number of people have entered the high-risk area, and if so, send an intervention scheduling request to the management terminal;

[0063] Determine whether there are stranded persons, and if it is determined that the stranded persons have moved less than a preset distance in a preset time period, send an intervention scheduling request to the management terminal;

[0064] Obtain data sent by IoT nodes, record the risk level distribution of all grids, block evacuation status, real-time evacuation progress, record the command responses, execution paths and feedback results of all security terminals, and send the aggregated data to the management terminal;

[0065] Generate a visual interface, including but not limited to: the status of emergency alarms, sound sensors, smoke sensors, camera images, global risk distribution maps, high-risk blocks and block evacuation priority lists, feedback records, target personnel and security terminal positioning, evacuation routes, and intervention dispatch records.

[0066] In a second aspect, an embodiment of the present invention provides a global security monitoring system based on the Internet of Things, the system comprising:

[0067] The electronic map acquisition and partitioning module is used to obtain the electronic map of the target location, determine the core area, non-core area and peripheral area of the target location, obtain the number and location data of the IoT nodes in each partition, and generate the configuration data of the edge computing terminal;

[0068] Edge computing terminals are deployed in each zone to collect and integrate IoT data sent by IoT nodes in the zone, identify abnormal conditions, send patrol instructions to the security terminal in the zone, receive processing results sent by the security terminal, and upload them to the cloud fusion platform after packaging with IoT data;

[0069] The cloud fusion platform is used to comprehensively process the data sent by the edge computing terminal, generate a global multi-channel spatiotemporal feature matrix, build a stampede risk prediction model to predict the stampede risk level, and send an evacuation request to the management terminal if it exceeds the warning level. It generates crowd evacuation path planning based on the multi-channel spatiotemporal feature matrix and the stampede risk level, and sends guidance instructions to the security terminal. It is also used to track the target personnel in real time and send the collated comprehensive data to the management terminal;

[0070] The security terminal is used to receive patrol instructions and evacuation guidance instructions, and implement on-site patrol, evacuation and feedback on on-site processing results;

[0071] The management terminal is used to receive the comprehensive data sorted by the cloud integration platform, generate a global security visualization interface, and is also used to directly send instructions to the security terminal.

[0072] One of the above technical solutions has the following beneficial effects:

[0073] In the method of the embodiment of the present invention, a global security monitoring method and system based on the Internet of Things are proposed. Through the Internet of Things nodes deployed differently in the core area, non-core area and peripheral area, combined with real-time data fusion of edge computing terminals, multi-channel spatio-temporal feature matrix analysis in the cloud, risk assessment of AI prediction models, crowd dynamic evacuation path planning, and multi-level risk response strategies, it realizes full coverage, real-time and refined risk prediction, early warning and management of crowded places, forms an intelligent security monitoring system architecture with full closed-loop, full-scene and full-process, and effectively improves the automation, intelligence and collaboration level of crowd safety management.

Description of the Drawings

[0074] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required to be used in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0075] Figure 1 is a schematic flow chart of the global security monitoring method based on the Internet of Things provided by the embodiment of the present invention;

[0076] Figure 2 is a functional block diagram of the global security monitoring system based on the Internet of Things provided by the embodiment of the present invention;

[0077] Figure 3 is a schematic hardware structure diagram of the global security monitoring system based on the Internet of Things provided by the embodiment of the present invention.

Detailed Embodiments

[0078] To make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all 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 belong to the scope of protection of the present invention.

[0079] Please refer to Figure 1 , which is a schematic flow chart of the global security monitoring method based on the Internet of Things provided by the embodiment of the present invention, as Figure 1As shown in the figure, the method includes the following steps:

[0080] S1. Obtain the electronic map of the target venue, determine the core area, non-core area and peripheral area of the target venue, and obtain the quantity and location data of the Internet of Things nodes in each partition to generate the configuration data of the edge computing terminal;

[0081] S2. The edge computing terminal collects the Internet of Things data sent by the Internet of Things nodes in the partition where it is located, identifies abnormal conditions, sends a patrol instruction to the security end in the partition, and uploads the processed result sent by the security end and the Internet of Things data to the cloud fusion platform after packaging;

[0082] S3. The cloud fusion platform comprehensively processes the data sent by the edge computing terminal, generates a global multi-channel spatio-temporal feature matrix, constructs a stampede risk prediction model to predict the stampede risk level, and sends an evacuation request to the management terminal if it exceeds the warning level;

[0083] S4. After receiving the evacuation instruction sent by the management terminal, generate a crowd evacuation path plan based on the multi-channel spatio-temporal feature matrix and the stampede risk level, send a guiding instruction to the security end, track the trajectories of the target personnel, and send the comprehensive data sorted by the cloud fusion platform to the management terminal to generate a visual interface.

[0084] In the embodiments of the present invention, the electronic map is partitioned into core, non-core, and peripheral areas, and the number and spatial distribution of IoT nodes in each partition are combined to generate configuration data for the edge computing terminal, realizing differential monitoring and management of different regions, ensuring high-precision and key monitoring in high-density areas, optimizing edge resource allocation at the same time, and reducing the overall system cost; the edge computing terminal collects and fuses multi-type IoT data in this partition in real time, quickly identifies abnormal situations, and issues inspection or disposal instructions to the security end in a timely manner, realizing a rapid on-site response to abnormal events. The processing results and real-time data of the security end are synchronously uploaded to the cloud to form a closed-loop feedback mechanism, greatly improving the on-site disposal efficiency and response speed; the cloud fusion platform uses a multi-channel spatio-temporal feature matrix to perform real-time risk prediction and grading evaluation on multiple regions of the whole venue through a trampling risk prediction model. When the prediction result reaches or exceeds the set warning level, an evacuation request is sent to the management terminal. After receiving the evacuation instruction sent by the management terminal, the evacuation is executed, greatly improving the accuracy and timeliness of risk identification, and effectively preventing group safety accidents; based on the spatio-temporal feature matrix and risk level, the optimal crowd evacuation path is generated, which can avoid high-risk and congested areas, realizing a scientific, diversion-based, and graded personnel evacuation plan. The evacuation path and guiding instructions can be pushed to the corresponding security end in real time, realizing multi-partition and multi-priority linkage guidance, reducing the risk of crowd crossing, and improving the evacuation efficiency; for target personnel in the core area, such as special groups such as important personnel, disabled people, and high-risk individuals, the whole process trajectory tracking is realized and linked with the evacuation plan in real time to ensure the exclusive safety guarantee of key personnel. The cloud fusion platform sorts out and visualizes the data, and the management terminal can intuitively control the overall security situation through the interactive interface, making real-time decisions, traceability analysis, and emergency review.

[0085] In the preferred embodiment of the present invention, S1 specifically includes:

[0086] S10. Obtain the electronic map of the target venue, extract the boundary of the target venue through edge detection and connected component analysis algorithms, and sequentially divide the core area, non-core area, and peripheral area from the inside out;

[0087] S11. Send the automatic partitioning result to the management terminal, and the management terminal reviews, adjusts, and finally confirms the partitioning boundary;

[0088] S12. Obtain the quantity and location data of emergency alarms, sound sensors, smoke sensors, and Bluetooth AoA base stations installed in the core area;

[0089] S13. Obtain the quantity and location data of cameras and Bluetooth RSSI base stations installed in the non-core area;

[0090] S14. Obtain the quantity and location data of cameras installed in the peripheral area;

[0091] S15. Generate the installation location data of the edge computing terminal based on the number and location data of the Internet of Things nodes in the core area, non-core area, and peripheral area.

[0092] In the embodiment of the present invention, through the edge detection and connected component analysis algorithm, the physical boundary of the target site is automatically and objectively extracted. Combining the spatial layout and moving lines, the core area, non-core area, and peripheral area are divided from the inside out, effectively reflecting the actual security management requirements and realizing differential security monitoring in different areas. After automatic zoning, it is visually reviewed and adjusted by the management personnel to ensure that the zoning results are both scientific and in line with actual management experience and special scenario needs. The number and precise location data of emergency alarms, sound sensors, smoke sensors, Bluetooth AoA base stations in the core area, cameras and Bluetooth RSSI base stations in the non-core area, and cameras in the peripheral area are respectively counted and collected by area. This not only ensures multi-dimensional and high-precision data collection and privacy interference in high-density key areas, but also avoids resource waste in the peripheral and transition areas, realizing the optimal matching of resource investment and risk distribution. Based on the number and spatial distribution of Internet of Things nodes in each partition, the installation location data of the optimal edge computing terminal is automatically analyzed and generated, ensuring that the terminal distribution can cover all key nodes and minimize redundancy to the greatest extent, improving data aggregation, edge processing, and response speed, reducing network latency and data transmission pressure, and enhancing the utilization rate of distributed computing resources.

[0093] In the preferred embodiment of the present invention, the S15 specifically includes:

[0094] S150. For the core area, use the K-means clustering algorithm to find the clustering centers, , where represents the spatial coordinates of the i-th node, and the total number of Internet of Things nodes in the core area is nodes, represents the set of nodes included in the j-th cluster, represents the central position of the j-th cluster, represents the initial estimated value of the number of clustering centers in the core area;

[0095] After clustering, each clustering center serves as the installation location of an edge computing terminal in the core area;

[0096] The optimal number of clusters is determined by the elbow method, which is the number of edge computing terminals in the core area;

[0097] S152. Use the grid division method to define the non-core area and calculate the grid size, , where represents the size of each grid, represents the total area of the non-core area, represents the total number of nodes in the non-core area, represents a set constant;

[0098] For the set of node positions within each grid , calculate the geometric center, , represents the installation position of the edge computing terminal within the j-th grid, represents the specific position coordinates of the nodes within the grid, represents the number of nodes within the j-th grid;

[0099] If each grid contains at least one node, then at the geometric center position of the grid install an edge computing terminal, and the total number of final edge terminals is the total number of grids containing nodes;

[0100] S153. Obtain the position data of all camera nodes in the peripheral area, and their respective spatial positions are , represents the total number of cameras in the peripheral area;

[0101] Sort the cameras in spatial order along the boundary of the peripheral area, and group the cameras according to the set maximum number of cameras per group ;

[0102] Calculate the geometric center of each group of cameras as the installation position of the edge terminal. For the f-th group, , represents that the f-th group contains cameras, represents the coordinates of the k-th camera within the group;

[0103] Install 1 edge computing terminal at each center point , and the final number of terminals is equal to the number of camera groups;

[0104] S154. Summarize the positions and quantities of the edge computing terminals.

[0105] In the embodiments of the present invention, the K-means clustering algorithm is adopted for the core area to automatically find the optimal aggregation center of node distribution, enabling the edge computing terminal to be as close as possible to the data-intensive points, improving the efficiency of data collection and preprocessing, reducing the number of terminals, and saving equipment and operation and maintenance costs; for the non-core area, the grid division method is adopted, and the terminal installation points of each grid are reasonably determined according to the area of the region and the number of nodes, ensuring that nodes in a large range can be efficiently covered, and at the same time, flexibly adapting to the sparse or dense changes of node distribution, improving the system elasticity and scalability; for the peripheral area, the geometric center method of camera grouping is adopted. For the area where only cameras are usually set, the cameras are sorted in spatial order along the boundary of the peripheral area and grouped according to the spatial distribution, and the geometric center is the installation point of the edge terminal. This not only ensures the timely access to the data of the camera nodes but also avoids the redundancy of terminal deployment, improving the overall resource utilization rate of the system. After the installation positions of each terminal are scientifically optimized, the collected data can be locally fused, abnormally preprocessed, and preliminarily analyzed at the nearest edge terminal, greatly reducing the total amount of raw data uploaded to the cloud, reducing the network bandwidth pressure and transmission delay, and improving the real-time response ability of the entire system; the adaptive optimization of terminal deployment makes the load distribution among terminals more balanced, reduces the single-point pressure, and improves the fault tolerance and system robustness.

[0106] In the preferred embodiment of the present invention, the S2 specifically includes:

[0107] S20. The edge computing terminal collects the status of the emergency alarm, sound data, the concentration of the smoke sensor, and the personnel positioning data located by the Bluetooth AoA base station in the core area, collects the camera video data and the personnel positioning data located by the Bluetooth RSSI base station in the non-core area, collects the camera video data in the peripheral area, and performs standardized preprocessing on various types of data. Bluetooth tags are carried by the personnel in the core area and the security personnel in the core area and non-core areas;

[0108] S21. Identify the abnormal conditions in the core area. Among them, the abnormal judgment formula in the core area is as follows: , , represents the comprehensive abnormal score in the core area, represents the status of the emergency alarm, 0 for normal and 1 for alarm, represents the abnormal sound score, 0 for normal volume and 1 for abnormal volume, represents the concentration of the smoke sensor, represents the smoke concentration alarm threshold, represents the abnormal value of personnel movement by the Bluetooth RSSI base station, , , and represent the weights of each index, represents the density change rate per unit time, represents the average velocity change rate of personnel per unit time;

[0109] S22. Divide the abnormal levels of the core area and conduct processing. When it is determined as level one, send patrol instructions to all security terminals within the core area and send reinforcement instructions to security terminals in non-core areas. it is determined as level two, send patrol instructions to security terminals within the first preset range from the abnormal point. it is determined as level three, send inspection instructions to the security terminal closest to the abnormal point. it is determined as normal.

[0110] In the embodiments of the present invention, by comprehensively utilizing multi-type Internet of Things node data in the core area and through a custom weighted fusion formula, accurate identification and automatic determination of multi-dimensional abnormal signals in complex scenarios are realized, effectively overcoming the problems of false alarms or missed alarms of a single sensor; the abnormal identification ability is improved from single-point monitoring to global perception, and abnormal signs such as personnel gathering, abnormal smoke, and sudden noisy shouts can be detected in the early stage, greatly enhancing the sensitivity and response speed to security threats in the core area. Through the abnormal level classification mechanism, the urgency of events can be automatically judged - for example, all security resources are automatically mobilized for a quick response at level one, while at level two or three, security forces closer or some of them are dispatched for inspections; the number and scope of security terminals are dynamically called to achieve "precision scheduling", which can not only efficiently handle major abnormalities and prevent the situation from expanding, but also avoid waste of resources for low-probability or false alarm events, improving the scientificity and economy of the overall security operation. Disposal instructions are automatically classified and pushed to security terminals, and security terminals upload the processing results and on-site collected data to the platform in real time to form a complete event management closed loop; various types of abnormal event types, response processes, disposal results, etc. can be automatically recorded, providing a data basis for subsequent traceability, review, and system optimization, significantly enhancing management transparency and decision-making level.

[0111] In a preferred embodiment of the present invention, S2 further includes:

[0112] S23. Real-time obtain video data of cameras in non-core areas and peripheral areas and preprocess the videos.

[0113] S24. Use a target detection deep learning model to identify the heads of personnel in image frames, count the crowd density in the coverage area of each camera based on the detection results, conduct optical flow analysis based on consecutive frame images, calculate the average movement speed of the crowd, calculate the average speed through the optical flow movement vector field, and calculate the abnormal comprehensive score of the non-core area based on the crowd density and average speed. Among them, , , , Represents the comprehensive anomaly score of the non-core area, Represents the threshold of the safe population density set for the non-core area, Represents the safe speed threshold of the non-core area, and Represents the weight and , Represents the number of people detected by the i-th camera, Represents the area covered by the i-th camera, and Represents the horizontal and vertical components of the optical flow motion vector of the k-th person, Represents the number of people tracked within the area;

[0114] S25. Divide the anomaly level of the non-core area and process it. When it is determined as level one, send a patrol instruction to all security terminals within the core area and send a reinforcement instruction to the security terminals in the peripheral area, it is determined as level two, send a patrol instruction to the security terminals within the second preset range from the anomaly point, it is determined as level three, send an inspection instruction to the security terminal closest to the anomaly point, it is determined as normal;

[0115] S26. Calculate the comprehensive anomaly score of the peripheral area , where, , Represents the safe density threshold of the peripheral area, Represents the safe speed threshold of the peripheral area, and Represents the weight and ;

[0116] S27. Divide the anomaly level of the peripheral area and process it. When it is determined as level one, send a patrol instruction to all security terminals within the core area and send a reinforcement instruction to the security terminals in the non-core area, it is determined as level two, send an inspection instruction to the security terminal closest to the anomaly point, it is determined as level three, record the event, it is determined as normal;

[0117] S28. Receive the processing results sent by the security terminal, package them with the Internet of Things data, and upload them to the cloud fusion platform.

[0118] In the embodiments of the present invention, cameras and Bluetooth RSSI base stations are used for large - scale and low - cost deployment, effectively making up for the sparse sensor coverage and weak real - time performance in non - core areas and peripheral areas, and achieving intelligent and safe perception without dead corners and in a wide range throughout the venue; by analyzing video data through a deep - learning object - detection model, it can accurately identify human heads and calculate crowd density and flow speed, automatically detecting and warning of potential hazards such as crowd gathering, abnormal flow, and stampede risks; according to the actual characteristics of different regions, optimal abnormal thresholds and judgment criteria are set respectively, and an abnormal judgment formula is customized to achieve flexible security strategies; according to the abnormal level, security patrol resources are dynamically allocated, key patrols are carried out in high - risk areas, and flexible handling is carried out in low - risk areas to avoid waste of resources, greatly improving the dispatching efficiency of on - site security forces and the emergency response speed. Therefore, this embodiment greatly improves the monitoring ability of abnormal conditions in non - core areas and peripheral areas, realizes intelligent, hierarchical, and region - differentiated security responses, effectively optimizes the allocation and management efficiency of security resources, and improves the accuracy and reliability of the overall system's risk prevention and control.

[0119] In a preferred embodiment of the present invention, in step S3, the cloud fusion platform comprehensively processes the data sent by the edge computing terminals, which specifically includes:

[0120] S30: The cloud fusion platform receives the data sent by each partition edge computing terminal, including the abnormal score level, occurrence location, time, and the security - end processing feedback result of the abnormal conditions in the core area, non - core area, and peripheral area;

[0121] S31: Generate a list of unprocessed events based on the occurrence location, time, and the security - end processing feedback result, and define a security - end processing result coefficient , where the security - end processing result coefficient is the ratio of the number of security - ends that have been processed to the total number of security - ends that have received the patrol instruction;

[0122] S32: Define an emergency event determination index , where , represents a weight coefficient and satisfies , represents the defined security - end processing result coefficient, represents the natural constant;

[0123] S33: Set an emergency event alarm threshold . If , it is determined as an emergency event, and an alarm is immediately sent to the management terminal and the list of unprocessed events is sent. Otherwise, only the list of unprocessed events is sent without an alarm.

[0124] In the embodiment of the present invention, through the intelligent fusion decision-making mechanism of multi-region and multi-source abnormal data and security response, the accurate quantification of the overall security risk and the automatic triggering of emergency events are realized, greatly improving the intelligent perception, active defense and efficient linkage capabilities of the security system.

[0125] In a preferred embodiment of the present invention, in step S3, a global multi-channel spatio-temporal feature matrix is generated, and a stampede risk prediction model is constructed to predict the stampede risk level. If the warning level is exceeded, an evacuation request is sent to the management terminal, which specifically includes:

[0126] Divide the entire area of the target venue into unified grids of a preset size, and the spatial grid numbers are , and represent the grid horizontal and vertical coordinate indices;

[0127] For the core area, directly calculate the number of people, average speed and flow direction in each grid according to the AoA data. For the non-core area and the peripheral area, calculate the number of people, average speed and flow direction in each grid according to the camera video data;

[0128] Fuse the data of the core area, non-core area and peripheral area, and dynamically generate a full-area multi-channel spatio-temporal feature matrix at each time step, with the color representing the density size, the arrow direction representing the flow direction, and the arrow thickness representing the speed size;

[0129] Construct a stampede risk prediction model based on the STCNN+ConvLSTM structure, and the input feature is , where represents the time length, and represent the spatial grid dimensions, represents the number of features, including density, speed and direction, and the output is a 4-level risk level probability matrix under the global spatial grid, including safe, attention, warning and danger;

[0130] Based on the actual stampede events in the historical venue, mark the true risk level of each event, obtain the multi-channel spatio-temporal feature matrix in the preset time period before the event occurs according to the time series, and manually mark the risk level as the supervised training signal;

[0131] Divide the data set into a training set and a validation set, conduct training and validation, and obtain an optimized stampede risk prediction model;

[0132] Real-time predict the stampede risk level at each time step through the trained stampede risk prediction model;

[0133] Take the risk level with the highest probability as the current stampede risk level of the grid, and generate a global risk distribution map with the risk levels of all grids;

[0134] If the risk in the global risk distribution map exceeds the warning level, evacuation is carried out.

[0135] Among them, for the core area, the real-time positioning data of AoA base stations provides the number of people within the precise personnel position calculation grid; the calculation method of the average speed: the position difference of each person at consecutive moments is divided by the time difference to obtain the individual speed, and the average is taken within the grid. For the non-core area and the peripheral area, the head of each person is recognized in each frame of the video, the position coordinates are counted, and the number of people within the grid is calculated; the average speed of people is calculated using the optical flow method: the optical flow vector is calculated for two consecutive frames, and the average value of the speed is calculated. Among them, the flow direction adopts the direction consistency calculation formula , , represents the personnel flow direction consistency index within the grid at time t represents the number of people statistically counted within the current grid, represents the personnel number within the grid, represents the included angle of the movement directions of the th person and the th person.

[0136] In the embodiment of the present invention, by introducing a multi-channel spatio-temporal feature matrix, the system can simultaneously obtain multi-dimensional dynamic data such as personnel density, average speed, and direction consistency in each partition and at each moment, providing rich spatio-temporal information for the risk prediction model and avoiding misjudgment caused by relying on only a single feature; using the STCNN+ConvLSTM model can effectively capture the hidden signals of abnormal evolution of the crowd flow and the occurrence of stampede risks, realizing high-precision prediction across space-time and multiple regions. The model can output the real-time stampede risk level or probability for each spatial grid and continuous time step, accurately locate high-risk regions and time periods, greatly enhancing the spatial resolution and dynamic adaptability of safety warnings; supporting the output of multiple risk levels, facilitating subsequent linkage with multi-level response mechanisms. Through supervised training with historical stampede events, the model can draw on past accident experience, have stronger recognition and generalization capabilities for risk signals in real scenarios, significantly reduce false alarm and missed alarm rates, and improve the practical value of the warning system in complex large-scale venues. The real-time and high-resolution risk level prediction results provide a reliable decision-making basis for subsequent crowd evacuation route planning, hierarchical guidance, and security force deployment, enabling the evacuation plan to dynamically avoid high-risk areas and minimizing the risk of secondary accidents to the greatest extent.

[0137] It should be noted that the STCNN+ConvLSTM model can adopt an existing framework that can implement the functions of the present application, or can adopt the design of the present application:

[0138] STCNN spatial convolution layer: input ; Spatiotemporal Convolution Structure: Convolution Layer 1: kernel size 3×3, number of convolution kernels 32, ReLU activation; Convolution Layer 2: kernel size 5×5, number of convolution kernels 64, ReLU activation; Batch Normalization Layer: normalizing features; STCNN Output: .

[0139] ConvLSTM Temporal Feature Learning Layer: Input ; 2-layer ConvLSTM Structure: convolution kernel 3×3, hidden layer dimension 64; ConvLSTM is defined as:

[0140] ,

[0141] ,

[0142] ,

[0143] ,

[0144] ,

[0145] Wherein, 、 and represent the input gate, forget gate, and output gate, 、 and represent the trainable convolution kernel weights and biases, and represent the cell state and hidden state, represents the Sigmoid activation, represents the convolution operation, ConvLSTM Output .

[0146] Output Prediction Layer: The ConvLSTM output is flattened by global average pooling; a fully connected layer with 128 neurons to 4 outputs; Softmax outputs the trampling risk level.

[0147] Training Data: Based on historical trampling events, the true level is labeled; the loss function uses class-weighted cross-entropy loss; Adam optimizer, initial learning rate 0.001; learning rate decay, patience value 3 epochs; EarlyStopping mechanism, patience value 5 epochs; finally, the validation precision and recall meet the requirements.

[0148] In a preferred embodiment of the present invention, in S4, generating a crowd evacuation path plan based on the multi-channel spatiotemporal feature matrix and the trampling risk level, and sending a guiding instruction to the security end, specifically includes:

[0149] Form a high-risk block by forming grids of danger and warning levels that are continuously distributed and have a quantity greater than the threshold. Sort the blocks according to the block risk level, where danger > warning > attention > safety. Sort the blocks within the same level according to the block area, where the larger the area, the higher the priority, and generate a block evacuation priority list.

[0150] Each priority block serves as a set of evacuation starting points, and the preset safety exit locations serve as a set of evacuation endpoints. Use the Dijkstra algorithm to perform path planning for each block in the block evacuation priority list. The optimal or shortest path avoids the danger and warning level grids and performs conflict detection: If the paths of high-priority blocks and low-priority blocks overlap, the high-priority blocks enter the path first, and the low-priority blocks can enter the overlapping area only after the high-priority blocks have evacuated through.

[0151] Send a guidance instruction to the security end within the preset range of the target evacuation block. The guidance instruction includes: the evacuation starting block number, the recommended path coordinate sequence, the current evacuation priority, the real-time entry into evacuation / waiting instruction, the number of evacuees, and the estimated time.

[0152] In the embodiment of the present invention, high-risk blocks are automatically clustered according to the risk levels of the global grid distribution, and the evacuation priority is established by combining the area and risk level sorting. It can refine the safe evacuation of the crowd to the specific blocks that "most need to be evacuated first", accurately locate the evacuation starting points, and avoid the chaos and safety hazards caused by the blind simultaneous evacuation of the whole venue. It supports the batch evacuation of multiple levels of blocks such as danger, warning, and attention, scientifically diverts the crowd, and greatly reduces the secondary risks caused by evacuation. Based on the real-time risk distribution and dynamic area priority, using classic path planning algorithms such as Dijkstra, an optimal evacuation route is generated separately for each evacuation block, effectively avoiding high-risk and high-density areas. The path conflict detection and dynamic scheduling mechanism prevent the crowd from multiple blocks from converging and congesting on the same path and exit, significantly reducing the risk of group safety accidents such as secondary stampedes. Automatically send the evacuation starting point, path, priority, entry / waiting instruction, etc. of each block to the corresponding security end through a differential push mechanism, enabling on-site security personnel to accurately divide the work, time-sharing, and batch-guide the crowd evacuation, and improving the professionalism and collaborative efficiency of on-site handling.

[0153] In the preferred embodiment of the present invention, in S4, the trajectory of the target personnel is tracked, and the comprehensive data sorted by the cloud fusion platform is sent to the management terminal to generate a visual interface, specifically including:

[0154] Use core area Bluetooth AoA positioning and non-core area RSSI positioning to track the evacuation trajectories of core area personnel carrying Bluetooth tags and security personnel in the core area and non-core area.

[0155] Determine whether a preset number of people have entered the high-risk area, and if so, send an intervention scheduling request to the management terminal;

[0156] Determine whether there are stranded persons, and if it is determined that the stranded persons have moved less than a preset distance in a preset time period, send an intervention scheduling request to the management terminal;

[0157] Obtain data sent by IoT nodes, record the risk level distribution of all grids, block evacuation status, real-time evacuation progress, record the command responses, execution paths and feedback results of all security terminals, and send the aggregated data to the management terminal;

[0158] Generate a visual interface, including but not limited to: the status of emergency alarms, sound sensors, smoke sensors, camera images, global risk distribution maps, high-risk blocks and block evacuation priority lists, feedback records, target personnel and security terminal positioning, evacuation routes, and intervention dispatch records.

[0159] The embodiment of the present invention can monitor the movement paths, location distribution and behavior characteristics of "target personnel" such as VIPs, security personnel, special groups or suspected high-risk individuals in real time through trajectory tracking, trigger risk warnings in a timely manner, and achieve precise intervention. The management terminal can fully control the progress of the evacuation of the entire venue, the execution of the security terminal and the response of personnel by receiving and analyzing comprehensive situation data; it can dynamically adjust the evacuation strategy, security deployment, path guidance, etc., realize the intelligence and refinement of command and dispatch, and ensure faster, more orderly and more efficient on-site response.

[0160] The embodiments of the present invention further provide device embodiments for implementing the steps and methods in the above method embodiments.

[0161] Please refer to Figure 2 , which is a functional block diagram of a global security monitoring system based on the Internet of Things provided by an embodiment of the present invention, the system includes:

[0162] The electronic map acquisition and partitioning module is used to obtain the electronic map of the target location, determine the core area, non-core area and peripheral area of the target location, obtain the number and location data of the IoT nodes in each partition, and generate the configuration data of the edge computing terminal;

[0163] Edge computing terminals are deployed in each zone to collect and integrate IoT data sent by IoT nodes in the zone, identify abnormal conditions, send patrol instructions to the security terminal in the zone, receive processing results sent by the security terminal, and upload them to the cloud fusion platform after packaging with IoT data;

[0164] A cloud integration platform is used to comprehensively process the data sent by edge computing terminals, generate a global multi-channel spatio-temporal feature matrix, construct a stampede risk prediction model to predict the stampede risk level, send an evacuation request to the management terminal if the warning level is exceeded, generate a crowd evacuation path plan based on the multi-channel spatio-temporal feature matrix and the stampede risk level, and send a guidance instruction to the security side. It is also used to track the real-time trajectory of target personnel and send the sorted comprehensive data to the management terminal;

[0165] The security side is used to receive inspection instructions and evacuation guidance instructions, and implement on-site inspections, evacuations, and feedback on on-site processing results;

[0166] The management terminal is used to receive the comprehensive data sorted by the cloud integration platform, generate a global security visualization interface, and is also used to directly send instructions to the security side.

[0167] Since each unit module in this embodiment can execute Figure 1 the method shown, for parts not described in detail in this embodiment, reference can be made to the relevant descriptions of Figure 1 .

[0168] Please refer to Figure 3 , which is a schematic hardware structure diagram of an Internet of Things-based global security monitoring system provided by an embodiment of the present invention. The data prediction device includes at least one processor and a memory. The at least one processor is coupled to the memory and is used to read and execute instructions in the memory to execute the Internet of Things-based global security monitoring method provided by an embodiment of the present invention.

[0169] At the hardware level, the device may include a processor, and optionally also an internal bus, a network interface, and a memory. Among them, the memory may include internal memory, such as high-speed random access memory (Random-Access Memory, RAM), and may also include non-volatile memory, such as at least one disk memory, etc. Of course, the device may also include other hardware required for other services.

[0170] The processor, network interface, and memory can be interconnected through an internal bus. The internal bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0171] A memory for storing programs. Specifically, the program may include program code, and the program code includes computer operation instructions. The memory may include a memory and a non-volatile memory, and provides instructions and data to the processor.

[0172] The steps of the method disclosed in the embodiments of the present invention may be directly embodied as being executed and completed by a hardware decoding processor, or executed and completed by a combination of hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory, and the processor reads the information in the memory and combines its hardware to complete the steps of the above method.

[0173] The systems, devices, modules or units illustrated in the above embodiments may be specifically implemented by a computer chip or an entity, or implemented by a product with a certain function. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0174] For the convenience of description, when describing the above devices, they are divided into various units or modules according to functions and described separately. Of course, when implementing the present invention, the functions of the various units or modules may be implemented in the same or multiple software and / or hardware.

[0175] Those skilled in the art should understand that the embodiments of the present invention may be provided as a method, a system, or a computer program product. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0176] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate for realizing in the process Figure 1 a process or multiple processes and / or blocks Figure 1means for the functions specified in one or more blocks.

[0177] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction means that implements the functions specified in one Figure 1 one or more processes and / or blocks Figure 1 means for the functions specified in one or more blocks.

[0178] These computer program instructions may also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide the steps for implementing the functions specified in one Figure 1 one or more processes and / or blocks Figure 1 means for the functions specified in one or more blocks.

[0179] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0180] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.

[0181] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology for information storage. The information may be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media such as modulated data signals and carrier waves.

[0182] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0183] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0184] The present invention may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present invention may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0185] Each embodiment of the present invention is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0186] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims

1. An Internet of Things-based full-domain security monitoring method, characterized in that, The method includes: S1. Obtain the electronic map of the target venue, determine the core area, non-core area and peripheral area of the target venue, and obtain the quantity and location data of IoT nodes in each partition to generate the configuration data of the edge computing terminal; S2. The edge computing terminal collects the IoT data sent by the IoT nodes in its partition, identifies abnormal situations, sends patrol instructions to the security side in the partition, and uploads the processed results sent by the security side and the IoT data to the cloud fusion platform after packaging; S3. The cloud fusion platform comprehensively processes the data sent by the edge computing terminal, generates a global multi-channel spatio-temporal feature matrix, constructs a stampede risk prediction model to predict the stampede risk level, and sends an evacuation request to the management terminal if the warning level is exceeded; S4. After receiving the evacuation instruction sent by the management terminal, generate a crowd evacuation path plan based on the multi-channel spatio-temporal feature matrix and the stampede risk level, send a guidance instruction to the security side, track the trajectories of the target personnel, and send the comprehensive data sorted by the cloud fusion platform to the management terminal to generate a visual interface.

2. The method for global security monitoring based on the Internet of Things according to claim 1, wherein The specific content of S1 includes: S10. Obtain the electronic map of the target venue, extract the boundary of the target venue through edge detection and connected component analysis algorithms, and sequentially divide the core area, non-core area and peripheral area from the inside out; S11. Send the automatic zoning result to the management terminal, and the management terminal reviews, adjusts and finally confirms the zoning boundary; S12. Obtain the quantity and location data of the emergency alarms, sound sensors, smoke sensors and Bluetooth AoA base stations installed in the core area; S13. Obtain the quantity and location data of the cameras and Bluetooth RSSI base stations installed in the non-core area; S14. Obtain the quantity and location data of the cameras installed in the peripheral area; S15. Generate the installation location data of the edge computing terminal based on the quantity and location data of the IoT nodes in the core area, non-core area and peripheral area.

3. The method for global security monitoring based on the Internet of Things according to claim 2, characterized in that, The specific content of S15 includes: S150. For the core area, use the K-means clustering algorithm to find K clustering centers, , where represents the spatial coordinates of the i-th node, and the total number of IoT nodes in the core area is nodes, represents the set of nodes included in the j-th cluster, represents the central position of the j-th cluster, represents the initial estimated value of the number of clustering centers in the core area; After clustering, each clustering center serves as the installation location of an edge computing terminal in the core area; The optimal number of clusters is determined by the elbow method, which is the number of edge computing terminals in the core area; S152. Define the non-core area using the grid division method and calculate the grid size, , where represents the size of each grid, represents the total area of the non-core area, represents the total number of nodes in the non-core area, represents a set constant; For the set of node positions within each grid , calculate the geometric center, , represents the installation position of the edge computing terminal within the j-th grid, represents the specific position coordinates of the nodes within the grid, represents the number of nodes within the j-th grid; If each grid contains at least one node, an edge computing terminal is installed at the geometric center of the grid. Finally, the total number of edge terminals is the total number of grids containing nodes; ​ S153. Obtain the position data of all camera nodes in the peripheral area, and their respective spatial positions are , indicating the total number of cameras in the peripheral area; Arrange the cameras in a spatial order along the boundary of the peripheral area, with a set maximum number of cameras per group Group the cameras; Calculate the geometric center of each group of cameras as the installation position of the edge terminal. For the f-th group, , denote that the f-th group contains cameras, and denote the coordinates of the k-th camera within this group; At each center point Install 1 edge computing terminal, and the final number of terminals is equal to the number of camera groups; S154. Summarize the locations and quantities of the edge computing terminals.

4. The method for global security monitoring based on the Internet of Things according to claim 1, characterized in that The specific content of S2 includes: S20. The edge computing terminal collects the status of the emergency alarms, sound data, smoke sensor concentration, and personnel location data located by the Bluetooth AoA base station in the core area, collects the camera video data and personnel location data located by the Bluetooth RSSI base station in the non-core area, and collects the camera video data in the peripheral area, and performs standardized preprocessing on various types of data. The personnel in the core area and the security personnel in the core area and non-core area carry Bluetooth tags; S21. Identify the abnormal conditions in the core area. The abnormal judgment formula for the core area is as follows: , , represents the comprehensive abnormal score of the core area, represents the status of the emergency alarm. The normal value is 0, and the alarm value is 1. represents the abnormal score of sound. The normal volume is 0, and the abnormal volume is 1. represents the concentration of the smoke sensor, represents the alarm threshold of the smoke concentration, represents the abnormal value of the movement of personnel by the Bluetooth RSSI base station, , , and represent the weights of each index, represents the density change rate per unit time, represents the average speed change rate of personnel per unit time. S22. Divide the abnormal levels of the core area and process them. When it is determined as level one, send a patrol instruction to all security terminals within the core area, and send a reinforcement instruction to the security terminals in the non-core area. it is determined as level two, and send a patrol instruction to the security terminals within the first preset range from the abnormal point. it is determined as level three, and send an inspection instruction to the security terminal closest to the abnormal point. it is determined as normal.

5. The method for global security monitoring based on the Internet of Things according to claim 4, wherein S2 also includes: S23. Real-time obtain the video data of the cameras in the non-core area and peripheral area, and preprocess the video; S24. Use the target detection deep learning model to identify the human heads in the image frame, count the crowd density in the coverage area of each camera according to the detection results, perform optical flow analysis based on consecutive frame images, calculate the average movement speed of the crowd, calculate the average speed through the optical flow motion vector field, and calculate the comprehensive anomaly score of the non-core area based on the crowd density and the average speed. Among them, , , , represents the comprehensive anomaly score of the non-core area, represents the set safe crowd density threshold for the non-core area, represents the safe speed threshold for the non-core area, and represent weights and , represents the number of people detected by the i-th camera, represents the coverage area of the i-th camera, and represent the horizontal and vertical components of the k-th person's optical flow motion vector, represents the number of people tracked in the area; S25. Divide the anomaly levels of the non-core areas and process them. When it is determined as level one, send a patrol instruction to all security terminals within the core area and send a reinforcement instruction to the security terminals in the peripheral area. it is determined as level two, and send a patrol instruction to the security terminals within the second preset range from the anomaly point. it is determined as level three, and send an inspection instruction to the security terminal closest to the anomaly point. it is determined as normal; S26. Calculate the comprehensive score of the peripheral area anomaly , where , represents the safety density threshold of the peripheral area, represents the safety speed threshold of the peripheral area, and represents the weight and ; S27. Divide the abnormal levels of the peripheral area and handle them. When it is determined as level one, send a patrol instruction to all security terminals within the core area, and send a reinforcement instruction to the security terminals in the non-core area. it is determined as level two, send an inspection instruction to the security terminal closest to the abnormal point. it is determined as level three, record the event. it is determined as normal; S28. Receive the processed results sent by the security side and upload the IoT data to the cloud fusion platform after packaging.

6. The method for global security monitoring based on the Internet of Things according to claim 5, wherein In S3, the cloud fusion platform comprehensively processes the data sent by the edge computing terminal, specifically including: S30. The cloud integration platform receives the data sent by each partition edge computing terminal, including the abnormal score level, occurrence location, time, and security end processing feedback results of the abnormal conditions in the core area, non-core area, and peripheral area. S31. Generate an unprocessed event list based on the occurrence location, time, and the security end processing feedback result, and define the security end processing result coefficient , the security end processing result coefficient is the ratio of the number of security ends that have been processed to the total number of security ends that have received the patrol instructions; S32. Define an emergency determination index , where represents a weight coefficient and satisfies , represents the coefficient of the security terminal processing result represents the natural constant; S33. Set the emergency event alarm threshold , if , it is determined as an emergency event, and an alarm is immediately sent to the management terminal and an unprocessed event list is sent. Otherwise, only the unprocessed event list is sent without alarm.

7. The method for global security monitoring based on the Internet of Things according to any one of claims 1 or 6, characterized in that In S3, a global multi-channel spatio-temporal feature matrix is generated, and a stampede risk prediction model is constructed to predict the stampede risk level. If the warning level is exceeded, an evacuation request is sent to the management terminal. Specifically, it includes: Divide the entire target area into unified grids of a preset size, and the spatial grid numbers are , and represent the horizontal and vertical coordinate indices of the grid; For the core area, the number of people, average speed, and flow direction in each grid are directly calculated based on AoA data. For the non-core area and peripheral area, the number of people, average speed, and flow direction in each grid are calculated based on camera video data. Integrate the data of the core area, non-core area, and peripheral area, and dynamically generate a full-area multi-channel spatio-temporal feature matrix at each time step. The density size is represented by color, the flow direction is represented by the arrow direction, and the speed size is represented by the arrow thickness. Construct a stampede risk prediction model based on the STCNN+ConvLSTM structure, with the input features being , where represents the time length, and represent the spatial grid dimensions, represents the number of features, including density, speed, and direction, and the output is a 4-level risk level probability matrix under the global spatial grid, including safe, attention, warning, and danger; Based on the actual stampede events in the historical venue, mark the real risk level of each event, obtain the multi-channel spatio-temporal feature matrix in the preset time period before the event occurs according to the time series, and manually mark the risk level as the supervised training signal. The dataset is divided into a training set and a validation set, trained and validated to obtain an optimized stampede risk prediction model. The trained stampede risk prediction model is used to predict the stampede risk level in real time at each time step. Take the risk level with the highest probability as the current stampede risk level of the grid, and generate a global risk distribution map with the risk levels of all grids. If the risk in the global risk distribution map exceeds the warning level, evacuation is carried out.

8. The method for global security monitoring based on the Internet of Things according to claim 1, wherein, In S4, a crowd evacuation path plan is generated based on the multi-channel spatio-temporal feature matrix and the stampede risk level, and a guidance instruction is sent to the security end. Specifically, it includes: Form a high-risk block by continuously distributing and aggregating the grids with dangerous and warning level grids whose quantity is greater than the threshold. Sort the blocks according to the block risk level, where danger > warning > attention > safety. Sort the blocks within the same level according to the block area, where the larger the area, the higher the priority, and generate a block evacuation priority list. Each priority block is used as a set of evacuation starting points, and the preset safe exit locations are used as a set of evacuation end points. The Dijkstra algorithm is used to plan the paths for each block in the block evacuation priority list. The optimal or shortest path avoids the dangerous and warning level grids, and conflict detection is carried out: if the paths of high-priority blocks and low-priority blocks overlap, the high-priority blocks enter the path first, and the low-priority blocks can enter the overlapping area only after the high-priority blocks have evacuated. Send a guidance instruction to the security end within the preset range of the target evacuation block. The guidance instruction includes: the evacuation starting block number, the recommended path coordinate sequence, the current evacuation priority, the real-time evacuation / waiting instruction, the number of evacuees, and the estimated time.

9. The method for global security monitoring based on the Internet of Things according to claim 8, characterized in that, In S4, the trajectory of the target personnel is tracked, and the comprehensive data sorted out by the cloud integration platform is sent to the management terminal to generate a visual interface. Specifically, it includes: Use core area Bluetooth AoA positioning and non-core area RSSI positioning to track the evacuation trajectories of the core area personnel carrying Bluetooth tags and the security personnel in the core area and non-core area. Determine whether there are a preset number of personnel entering the high-risk area. If so, send an intervention scheduling request to the management terminal; Determine whether there are stranded personnel. If it is determined that the moving distance of the stranded personnel within a preset time period is less than the preset distance, send an intervention scheduling request to the management terminal; Obtain the data sent by the IoT nodes, record the distribution of grid risk levels, the evacuation status of the area, and the real-time evacuation progress of all areas, record the instruction responses, execution paths, and feedback results of all security terminals, and summarize the data and send it to the management terminal; Generate a visualization interface, including but not limited to: the status of emergency alarms, sound sensors, smoke sensors, camera images, the overall risk distribution map, high-risk areas, and the list of area evacuation priorities, feedback records, the positioning of target personnel and security terminals, evacuation paths, and intervention scheduling records.

10. An Internet of Things-based global security monitoring system using the method described in claim 1, characterized in that, The system includes: An electronic map acquisition and zoning module, which is used to acquire the electronic map of the target venue, determine the core area, non-core area, and peripheral area of the target venue, and acquire the quantity and location data of IoT nodes in each zone to generate the configuration data of the edge computing terminal; The edge computing terminal is deployed in each zone and is used to collect and fuse the IoT data sent by the IoT nodes in its zone, identify abnormal situations, send patrol instructions to the security terminals in the zone, receive the processing results sent by the security terminals, and upload them to the cloud fusion platform after being packaged together with the IoT data; The cloud fusion platform is used to comprehensively process the data sent by the edge computing terminal, generate an overall multi-channel spatio-temporal feature matrix, build a stampede risk prediction model to predict the stampede risk level. If it exceeds the warning level, send an evacuation request to the management terminal, generate a crowd evacuation path plan based on the multi-channel spatio-temporal feature matrix and the stampede risk level, and send a guiding instruction to the security terminal. It is also used to perform real-time trajectory tracking on target personnel and send the sorted comprehensive data to the management terminal; The security terminal is used to receive patrol instructions and evacuation guiding instructions, and perform on-site patrol, evacuation, and feedback on-site processing results; The management terminal is used to receive the comprehensive data sorted by the cloud fusion platform, generate an overall security visualization interface, and is also used to directly send instructions to the security terminal.

Images