Intelligent analysis and management method and system for security and protection monitoring storage information

Abstract

The invention relates to the technical field of security and protection monitoring systems, in particular to an intelligent analysis and management method and system for security and protection monitoring storage information, and the method comprises the following steps: S1, reasonably arranging a monitoring camera, guaranteeing the complete coverage of a key region needing security and protection monitoring, and collecting the image and video data transmitted by the monitoring camera in real time; according to the invention, through a dynamic time window division mechanism and an edge-cloud collaborative architecture, intelligent scheduling and optimal configuration of monitoring resources are realized, and the processing efficiency and real-time performance of the system are greatly improved; a multi-level deep learning model is adopted to be fused with a target detection and behavior recognition technology, abnormal behaviors are intelligently researched and judged in combination with a preset rule base, the accuracy and reliability of abnormal event detection are remarkably improved, an innovative three-level emergency response mechanism and a resource awareness scheduling strategy are adopted, and the safety of abnormal event detection is improved. The stable operation of the system under a high-load condition is effectively ensured, and the continuous monitoring capability of a key security scene is ensured.

Description

Technical Field

[0001] This invention relates to the field of security monitoring system technology, specifically to a method and system for intelligent analysis and management of security monitoring stored information. Background Technology

[0002] Traditional security monitoring systems primarily rely on manual monitoring and simple motion detection technologies, resulting in low monitoring efficiency, high false alarm rates, and delayed response times. Existing technologies typically employ fixed time windows for video analysis, making it difficult to adapt to the real-time requirements of different scenarios. Abnormal behavior detection is often based on single algorithm models, with limited accuracy and a lack of intelligent judgment capabilities. System architectures are mostly centralized processing models, placing high demands on network bandwidth and lacking edge computing capabilities. Furthermore, the lack of effective resource scheduling and emergency response mechanisms makes them prone to analysis delays or system crashes when facing sudden high loads. In addition, traditional systems generally lack comprehensive quality assessment systems, making it difficult to provide reliable data for security decisions. These technical deficiencies make it difficult for existing security monitoring systems to meet the higher requirements of modern smart security for real-time performance, accuracy, and stability. Summary of the Invention

[0003] To address these issues, the present invention provides a method and system for intelligent analysis and management of security monitoring and storage information.

[0004] This invention provides the following technical solution: a method for intelligent analysis and management of security monitoring stored information, comprising the following steps: S1, rationally arrange surveillance cameras to ensure full coverage of key areas that require security monitoring, and collect image and video data transmitted by surveillance cameras in real time; S2 preprocesses the acquired images, including denoising, enhancement, and format conversion, and stores the processed data in a distributed storage system; S3 divides the security monitoring storage information into multiple time windows according to a preset fixed duration. Each time window has an equal and configurable duration. S4 performs abnormal behavior analysis on the security monitoring stored information within each time window; S5 monitors the operating status of the analysis system in real time during the abnormal behavior analysis process and intelligently senses and analyzes potential abnormalities in the process. S6, when an abnormal potential is detected in the analysis process, record the actual time taken for the camera to perform abnormal behavior analysis on that time window; S7 dynamically calculates the upper limit of the expected duration of each time window for normal camera analysis in the current scene, based on the real-time requirements and security level of video surveillance. S8 compares and analyzes the actual time taken with the upper limit of the expected duration; S9, set up the exception handling mechanism; S10 generates an abnormal behavior analysis quality report, recording the abnormal handling process and the final analysis results.

[0005] As a preferred embodiment of the present invention, in step S4, a deep learning model is used to perform real-time target detection and behavior recognition on the surveillance video. Based on a preset rule base, abnormal behavior is determined based on the recognition results, an abnormal behavior analysis report is generated, and suspicious time segments are marked. The abnormal behavior analysis in step S4 specifically includes: using the YOLOv5 algorithm to achieve real-time target detection and using a 3D convolutional neural network for behavior recognition. The abnormal behavior determination rules in step S4 include: identification of personnel entering restricted areas, items left behind or lost, abnormal gathering behavior, and violent behavior.

[0006] As a preferred embodiment of the present invention, in step S5, the following indicators are used to determine the process during analysis: the length of the queue to be processed exceeds a preset threshold, the image motion blur exceeds the identifiable range, and the analysis delay exceeds the real-time requirements. The abnormal potential monitoring in step S5 also includes: real-time monitoring of GPU memory usage, detection of network transmission delay, monitoring of the process status of the analysis service, and triggering an early warning mechanism when any monitoring indicator exceeds a preset threshold.

[0007] As a preferred embodiment of the present invention, in S8-9, if the actual time consumed does not exceed the upper limit of the expected duration, the analysis of the current time window continues to be completed; if the actual time consumed exceeds the upper limit of the expected duration, an abnormal handling mechanism is triggered. The abnormal handling mechanism includes: reducing the analysis accuracy of subsequent video frames to ensure real-time performance, migrating the current analysis task to a backup computing node, and marking suspicious time segments for manual review. The priority of the exception handling mechanism in S9 is as follows: First, try reducing the analysis precision; When the accuracy drops to the minimum threshold and still cannot meet the real-time requirements, task migration is triggered. Abnormal periods that cannot be processed in the end are automatically marked as high-risk periods.

[0008] As a preferred embodiment of the present invention, the time window division in S3 specifically includes: setting different time window lengths according to the importance of the monitoring scene, with shorter time windows used for key areas, and setting overlapping intervals between time windows to ensure the continuity of behavior analysis and support dynamic adjustment of time window lengths to adapt to different network bandwidth and computing resource conditions.

[0009] As a preferred embodiment of the present invention, the upper limit of the expected duration in S7 is calculated using the following formula: T_max=α×(W / V)+β×(R / B), where W is the length of the time window, V is the video stream bitrate, R is the computing resources required for analysis, B is the available computing resources, and α and β are scene weight coefficients.

[0010] A security monitoring and storage information intelligent analysis and management system, employing the aforementioned security monitoring and storage information intelligent analysis and management method, includes a device layer, an output terminal of the device layer electrically connected to a transmission layer, an output terminal of the transmission layer electrically connected to a storage layer, an output terminal of the storage layer electrically connected to an analysis layer, and an output terminal of the analysis layer electrically connected to a management layer.

[0011] As a preferred embodiment of the present invention, the device layer includes multiple high-definition network cameras and edge computing nodes. The multiple high-definition network cameras are deployed at key locations in the monitored area. The edge computing nodes are used for preliminary processing of video data. The transmission layer includes a 5G private network transmission module and a data encryption gateway. The storage layer includes a distributed object storage cluster and a time-series database. The analysis layer includes a video decoding module, a target detection engine, a behavior analysis engine, and an anomaly detection module. The management layer includes a resource scheduler, an anomaly processor, and a quality evaluator. The edge computing nodes are configured with dedicated AI acceleration chips, local cache storage, a network outage resume function module, and an adaptive bitrate adjustment algorithm.

[0012] Compared with the prior art, the beneficial effects of the present invention are: This invention achieves intelligent scheduling and optimized configuration of monitoring resources through a dynamic time window segmentation mechanism and an edge-cloud collaborative architecture, significantly improving system processing efficiency and real-time performance. It employs a multi-layered deep learning model that integrates target detection and behavior recognition technologies, combined with a pre-defined rule base for intelligent analysis of abnormal behavior, significantly improving the accuracy and reliability of abnormal event detection. An innovative three-level emergency response mechanism and resource-aware scheduling strategy effectively ensure stable system operation under high load conditions, guaranteeing continuous monitoring capabilities for critical security scenarios. An intelligent quality assessment system provides data support for system performance optimization, while standardized analysis reports simplify security management processes. The overall solution, through an end-to-end intelligent analysis pipeline, achieves full-process optimization from data collection and processing to analysis and decision-making, reducing manual intervention while comprehensively improving the automation level, response speed, and security capabilities of the security monitoring system. Attached Figure Description

[0013] Figure 1 This is a flowchart of the intelligent analysis and management method for security monitoring and storage information of the present invention; Figure 2This is a flowchart of the intelligent analysis and management system for security monitoring and storage information of the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Please see Figures 1-2 The technical solution provided by the present invention specifically includes the following embodiments: Example: A method for intelligent analysis and management of security monitoring stored information, comprising the following steps: S1, rationally arrange surveillance cameras to ensure full coverage of key areas that require security monitoring, and collect image and video data transmitted by surveillance cameras in real time; S2 preprocesses the acquired images, including denoising, enhancement, and format conversion, and stores the processed data in a distributed storage system; S3 divides security monitoring storage information into multiple time windows of equal and configurable duration according to a preset fixed duration. The time window division in S3 specifically includes: setting different time window lengths according to the importance of the monitoring scene, with shorter time windows used for key areas, and setting overlapping intervals between time windows to ensure the continuity of behavior analysis and support dynamic adjustment of time window length to adapt to different network bandwidth and computing resource conditions. The system employs an intelligent time window segmentation mechanism, dynamically adjusting the analysis duration based on the security level of different monitoring scenarios. For key security areas, the system automatically shortens the time window to improve real-time detection and ensures the continuity of behavioral analysis by setting overlapping intervals. The time window length can be dynamically expanded or contracted based on network conditions and computing resources. A preset algorithm is used to balance analysis accuracy and processing efficiency. When network bandwidth is insufficient, the system automatically extends the window duration to reduce transmission pressure. When a sudden abnormal event is detected, the window is immediately shortened to improve response speed. This mechanism dynamically calculates the expected processing time through scene weight coefficients, ensuring that the system can maintain stable analysis performance in various environments and effectively adapt to complex and ever-changing monitoring needs. S4 performs abnormal behavior analysis on the security monitoring stored information within each time window. In S4, a deep learning model is used to perform real-time target detection and behavior recognition on the monitoring video. Based on a preset rule base, the recognition results are used to determine abnormal behavior, generate an abnormal behavior analysis report, and mark suspicious time segments. The abnormal behavior analysis in S4 specifically includes: using the YOLOv5 algorithm to achieve real-time target detection and using a 3D convolutional neural network for behavior recognition. The abnormal behavior determination rules in S4 include: identification of personnel entering restricted areas, items left behind or lost, abnormal gathering behavior, and violent behavior. In the target detection stage, the system uses an improved YOLOv5s model as its basic architecture. It enhances the ability to capture small targets by introducing the CBAM attention mechanism and optimizes multi-scale feature fusion using a BiFPN structure. For edge computing scenarios, the model achieves a balance between accuracy and efficiency through knowledge distillation and dynamic quantization techniques. In the behavior recognition stage, the system uses the SlowFast 3D CNN architecture to process target tracking segments. It captures spatial semantic features and temporal motion patterns through a dual-path design, effectively solving the problem of insufficient modeling of continuous behavior in traditional methods. YOLOv5 first performs real-time target detection and tracking on the video stream, generating target trajectory segments with spatiotemporal continuity. When the segments accumulate to 16 frames, it automatically triggers 3D CNN for behavior classification. This cascaded processing method achieves parallel computing through asynchronous pipeline design and avoids the latency accumulation caused by traditional serial processing by combining double-buffered queue technology. This application also has intelligent resource awareness capabilities, which can dynamically adjust the analysis strategy according to indicators such as GPU memory usage and processing queue length. When resources are sufficient, a high-precision mode (YOLOv5 + full 3D CNN) is adopted. When the system load is detected to exceed the threshold, it automatically switches to a lightweight analysis mode (such as reducing the input resolution or skipping the behavior analysis of secondary targets). The expected processing time of the time window is dynamically calculated by the formula T_max=α×(W / V)+β×(R / B), thereby adaptively adjusting the analysis rhythm to ensure the real-time requirements of key security areas. The formula T_max = α × (W / V) + β × (R / B) evaluates the system load from two dimensions: the first term, α × (W / V), quantifies the time cost of video data transmission, reflecting the matching relationship between the camera bitrate (V) and the time window length (W). In key areas, the response speed is improved by shortening W; the second term, β × (R / B), evaluates the supply and demand ratio of computing resources. When the GPU computing power (B) is insufficient or the complexity of the analysis task (R) surges, the allowable time threshold is automatically extended. The system dynamically adjusts the α and β coefficients based on real-time monitored indicators such as transmission latency and video memory usage. For example, in the case of a sudden high-density crowd, the weight of α is reduced (0.6 → 0.4) to prioritize the supply of computing resources, while triggering collaborative computing at edge nodes to ensure that abnormal behavior detection is completed within the T_max constraint. The system integrates target detection and behavior recognition technologies to build a multi-layered anomaly analysis system. It uses a deep learning model to identify suspicious targets in the monitoring screen in real time and combines a preset rule base to intelligently judge personnel behavior. The system continuously monitors key performance indicators during the analysis process, including parameters such as processing latency and resource consumption. When it detects a decline in image quality or excessive system load, it automatically switches to a lightweight analysis mode. It configures exclusive judgment rules for different scenarios, such as restricted area intrusion recognition and object abandonment detection, to achieve accurate anomaly warning. Through multi-dimensional data analysis, the system can effectively distinguish between real threats and ordinary anomalies, significantly reducing the false alarm rate. S5 monitors the operating status of the analysis system in real time during the abnormal behavior analysis process and intelligently senses potential abnormalities during the analysis process. In S5, the following indicators are used to determine the abnormality during the analysis: the length of the queue to be processed exceeds the preset threshold, the image motion blur exceeds the identifiable range, and the analysis delay exceeds the real-time requirements. The abnormality monitoring in S5 also includes: real-time monitoring of GPU memory usage, detection of network transmission delay, monitoring of the process status of the analysis service, and triggering an early warning mechanism when any monitoring indicator exceeds the preset threshold. The system employs a three-tiered response mechanism to handle anomalies during the analysis process. When an increase in processing latency is detected, real-time performance is maintained by reducing the analysis precision. If the problem persists, a computational task migration is initiated, distributing the load to backup nodes. Periods that still cannot be processed are marked as high-risk segments. This tiered processing strategy ensures that the system can maintain stable operation under different pressure conditions, guaranteeing the detection quality during critical periods while preventing system overload and crashes. The system also automatically records all emergency handling events, providing data support for subsequent optimization. S6, when an abnormal potential is detected in the analysis process, record the actual time taken for the camera to perform abnormal behavior analysis on that time window; S7 dynamically calculates the upper limit of the expected duration of each time window for normal analysis of the camera in the current scene, based on the real-time requirements and security level of video surveillance. The upper limit of the expected duration in S7 is calculated using the following formula: T_max=α×(W / V)+β×(R / B), where W is the length of the time window, V is the video stream bitrate, R is the computing resources required for analysis, B is the available computing resources, and α and β are scene weight coefficients. S8 compares the actual time taken with the upper limit of the expected duration. If the actual time taken does not exceed the upper limit of the expected duration, the analysis of the current time window continues. If the actual time taken exceeds the upper limit of the expected duration, the exception handling mechanism is triggered. S9 sets up an anomaly handling mechanism; the anomaly handling mechanism includes: reducing the analysis accuracy of subsequent video frames to ensure real-time performance, migrating the current analysis task to a backup computing node, and marking suspicious time segments for manual review; the priority of the anomaly handling mechanism in S9 is: first, try to reduce the analysis accuracy; when the accuracy is reduced to the minimum threshold and still cannot meet the real-time requirements, trigger task migration; the anomaly period that still cannot be handled is automatically marked as a high-risk period; By deploying edge computing nodes on monitoring terminals, local preprocessing of video data is achieved. Edge devices run optimized lightweight analysis models to perform preliminary screening of video streams, uploading only suspicious segments to the cloud for in-depth analysis. This architecture significantly reduces network transmission pressure and ensures information security through data encryption. The system supports the function of resuming transmission after network interruption, automatically activating local storage when the network is interrupted, and retransmitting key data after the connection is restored. The collaborative working mode of edge and cloud not only leverages the powerful capabilities of centralized analysis but also retains the real-time advantages of edge computing. The system monitors the running status of each computing node in real time, including key indicators such as processor load and memory usage. When it detects that a node is under resource pressure, the scheduler will automatically adjust the task allocation and migrate some of the load to idle nodes. The system also supports dynamically adjusting parameters such as batch processing size to optimize resource utilization efficiency. By continuously recording the actual processing time and comparing it with the expected value, the system can intelligently predict resource demand and perform capacity expansion or load balancing in advance. This resource awareness mechanism ensures that the analysis service always maintains stable performance and avoids analysis interruptions due to insufficient resources. S10, Generate an abnormal behavior analysis quality report, recording the abnormal handling process and the final analysis results; The system automatically generates multi-dimensional analysis quality reports to comprehensively evaluate the effectiveness of anomaly detection. The reports not only record the final analysis results but also provide detailed statistics on various events during the processing, including triggered emergency mechanisms and accuracy adjustment records. For marked high-risk periods, the system retains complete analysis process data, supporting post-event traceability and review. By establishing a standardized evaluation indicator system, managers can intuitively grasp the system's operating status and promptly identify potential problems. This quality evaluation system serves both daily operation and maintenance monitoring and provides a data foundation for continuous system optimization.

[0016] A security monitoring and storage information intelligent analysis and management system employs a security monitoring and storage information intelligent analysis and management method, comprising a device layer, an output terminal of which is electrically connected to a transmission layer, an output terminal of which is electrically connected to a storage layer, an output terminal of which is electrically connected to an analysis layer, and an output terminal of which is electrically connected to a management layer. The device layer includes multiple high-definition network cameras and edge computing nodes. The multiple high-definition network cameras are deployed at key locations in the monitoring area, and the edge computing nodes are used for preliminary processing of video data. The transmission layer includes a 5G private network transmission module and a data encryption gateway. The storage layer includes a distributed object storage cluster and a time-series database. The analysis layer includes a video decoding module, a target detection engine, a behavior analysis engine, and an anomaly detection module. The management layer includes a resource scheduler, an anomaly processor, and a quality evaluator. The edge computing nodes are configured with dedicated AI acceleration chips, local cache storage, a network outage resume function module, and an adaptive bitrate adjustment algorithm.

[0017] This invention proposes an intelligent analysis and management method and system for security monitoring and storage information. By constructing an end-to-end intelligent analysis pipeline, it achieves efficient processing and accurate anomaly detection of monitoring data. The method first comprehensively collects video data from the monitored area through a rationally deployed camera network, and then uses preprocessing techniques such as noise reduction and enhancement to improve data quality. The system innovatively adopts a dynamic time window segmentation mechanism, dynamically adjusting the analysis period length based on regional importance, network conditions, and computing resources. It ensures the continuity of behavior analysis by setting overlapping intervals and intelligently calculates the expected processing time based on scene weight coefficients, enabling the system to adapt to various complex environments. In the core abnormal behavior analysis stage, the system integrates the YOLOv5 object detection algorithm and 3D convolutional neural network behavior recognition technology to construct a multi-layered intelligent analysis system. This system can accurately identify various security threats such as personnel entering restricted areas, items left behind, and abnormal gatherings, and achieves precise anomaly warnings through a preset rule base. To ensure stable system operation, the method establishes a comprehensive status monitoring and emergency handling mechanism. By monitoring key indicators such as GPU memory and network latency in real time, the system can intelligently perceive potential risks. When a performance bottleneck is detected, the system reduces... The system employs a three-tiered response strategy—accuracy, migration task, and high-risk labeling—to effectively balance the conflict between real-time performance and analytical accuracy. It also innovatively adopts an edge-cloud collaborative architecture, deploying edge computing nodes equipped with AI acceleration chips at camera terminals to achieve localized preprocessing and encrypted transmission of data. This reduces network load while ensuring continuous operation even during network outages. The accompanying intelligent analysis and management system features a layered design, including a device layer, transmission layer, storage layer, analysis layer, and management layer. High-definition cameras and edge nodes at the device layer complete data acquisition and initial processing; the transmission layer ensures secure data transmission through a 5G private network and encrypted gateways; the storage layer's distributed clusters and time-series database enable efficient access to massive amounts of data; the analysis layer integrates core functions such as video decoding, target detection, and behavior analysis; and the management layer's resource scheduler, anomaly handler, and quality evaluator are responsible for system operation and optimization. This system dynamically balances computational load through resource-aware scheduling algorithms and automatically generates multi-dimensional quality assessment reports, providing comprehensive support for security management decisions. The overall solution achieves end-to-end optimization from data acquisition to intelligent analysis, significantly improving the automation level and responsiveness of the security monitoring system.

[0018] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A method for intelligent analysis and management of security monitoring stored information, characterized in that: Includes the following steps: S1, rationally arrange surveillance cameras to ensure full coverage of key areas that require security monitoring, and collect image and video data transmitted by surveillance cameras in real time; S2 preprocesses the acquired images, including denoising, enhancement, and format conversion, and stores the processed data in a distributed storage system; S3 divides the security monitoring storage information into multiple time windows according to a preset fixed duration. Each time window has an equal and configurable duration. S4 performs abnormal behavior analysis on the security monitoring stored information within each time window; S5 monitors the operating status of the analysis system in real time during the abnormal behavior analysis process and intelligently senses and analyzes potential abnormalities in the process. S6, when an abnormal potential is detected in the analysis process, record the actual time taken for the camera to perform abnormal behavior analysis on that time window; S7 dynamically calculates the upper limit of the expected duration of each time window for normal camera analysis in the current scene, based on the real-time requirements and security level of video surveillance. S8 compares and analyzes the actual time taken with the upper limit of the expected duration; S9, set up the exception handling mechanism; S10 generates an abnormal behavior analysis quality report, recording the abnormal handling process and the final analysis results.

2. The intelligent analysis and management method for security monitoring and storage information according to claim 1, characterized in that: In step S4, a deep learning model is used to perform real-time target detection and behavior recognition on the surveillance video. Based on a preset rule base, abnormal behavior is judged based on the recognition results, an abnormal behavior analysis report is generated, and suspicious time segments are marked. The abnormal behavior analysis in step S4 specifically includes: using the YOLOv5 algorithm to achieve real-time target detection and using a 3D convolutional neural network for behavior recognition. The abnormal behavior judgment rules in step S4 include: identification of personnel entering restricted areas, items left behind or lost, abnormal gathering behavior, and violent behavior.

3. The intelligent analysis and management method for security monitoring and storage information according to claim 1, characterized in that: In S5, the following indicators are used to determine the process during analysis: the length of the queue to be processed exceeds a preset threshold, the image motion blur exceeds the identifiable range, and the analysis delay exceeds the real-time requirements. The anomaly and potential risks monitoring in S5 also include: real-time monitoring of GPU memory usage, detection of network transmission delay, monitoring of the process status of the analysis service, and triggering an early warning mechanism when any monitoring indicator exceeds a preset threshold.

4. The intelligent analysis and management method for security monitoring and storage information according to claim 1, characterized in that: In S8-9, if the actual time consumed does not exceed the upper limit of the expected duration, the analysis of the current time window continues. If the actual time consumed exceeds the upper limit of the expected duration, an anomaly handling mechanism is triggered. The anomaly handling mechanism includes: reducing the analysis accuracy of subsequent video frames to ensure real-time performance, migrating the current analysis task to a backup computing node, and marking suspicious time segments for manual review. The priority of the exception handling mechanism in S9 is as follows: First, try reducing the analysis precision; When the accuracy drops to the minimum threshold and still cannot meet the real-time requirements, task migration is triggered. Abnormal periods that cannot be processed in the end are automatically marked as high-risk periods.

5. The intelligent analysis and management method for security monitoring and storage information according to claim 1, characterized in that: The time window division in S3 specifically includes: setting different time window lengths according to the importance of the monitoring scene, with shorter time windows used for key areas, and setting overlapping intervals between time windows to ensure the continuity of behavior analysis and support dynamic adjustment of time window lengths to adapt to different network bandwidth and computing resource conditions.

6. The intelligent analysis and management method for security monitoring and storage information according to claim 1, characterized in that: The upper limit of the expected duration in S7 is calculated using the following formula: T_max=α×(W / V)+β×(R / B), where W is the length of the time window, V is the video stream bitrate, R is the computing resources required for analysis, B is the available computing resources, and α and β are scene weight coefficients.

7. A security monitoring and storage information intelligent analysis and management system, employing the security monitoring and storage information intelligent analysis and management method according to any one of claims 1-6, characterized in that: It includes a device layer, the output of which is electrically connected to a transmission layer, the output of which is electrically connected to a storage layer, the output of which is electrically connected to an analysis layer, and the output of which is electrically connected to a management layer.

8. The intelligent analysis and management system for security monitoring and storage information according to claim 7, characterized in that: The device layer includes multiple high-definition network cameras and edge computing nodes. The multiple high-definition network cameras are deployed at key locations in the monitored area. The edge computing nodes are used for preliminary processing of video data. The transmission layer includes a 5G private network transmission module and a data encryption gateway. The storage layer includes a distributed object storage cluster and a time-series database. The analysis layer includes a video decoding module, a target detection engine, a behavior analysis engine, and an anomaly detection module. The management layer includes a resource scheduler, an anomaly processor, and a quality evaluator. The edge computing nodes are configured with dedicated AI acceleration chips, local cache storage, a network outage resume function module, and an adaptive bitrate adjustment algorithm.

Images