Regional security management system based on multi-source data fusion

Abstract

The invention discloses a regional security management system based on multi-source data fusion, and relates to the technical field of security management. The system comprises a security intrusion detection model, a perimeter security linkage module and a security disposal effect evaluation module. According to the method, after key safety intrusion area marking is carried out on the designated area, when the perimeter safety condition of the designated area is monitored in real time, multi-source data is input into the designated safety intrusion detection model for analysis, the intrusion detection result containing the existence of the intrusion object and the perimeter safety is output, and if the intrusion object exists, the intrusion detection result is output. If yes, a perimeter security alarm mechanism is triggered immediately to take security linkage measures, and after security disposal is finished, security disposal effect evaluation is carried out to determine whether intrusion detection qualification verification is carried out or not, so that the security management efficiency of the perimeter of the specified area is improved, and the security of the specified area is ensured. The problem that in the prior art, when safety intrusion detection is carried out on the area perimeter, linkage synchronism of alarm and emergency response is poor is solved.

Description

Technical Field

[0001] This invention relates to the field of security management technology, and in particular to a regional security management system based on multi-source data fusion. Background Technology

[0002] Existing technologies cover multiple aspects of community security management, including personnel security management, vehicle security management, environmental and facility security management, entrance and perimeter security management, information and data security management, and emergency and incident management. By integrating multi-source data, the security status of the community can be displayed in real time, and data-driven security decision-making can be provided for property and community managers.

[0003] Currently, the process of detecting perimeter intrusion in entrance and exit security management involves multiple sensing technologies, data fusion, real-time analysis, and response mechanisms. By combining video surveillance, sensors, artificial intelligence algorithms, and network linkage technologies, the system can detect and respond to illegal intrusion events in real time. Intrusion detection primarily relies on artificial intelligence algorithms and pattern recognition technologies, including target detection and recognition. Computer vision and deep learning algorithms are used to analyze video streams, detecting and identifying targets (such as people, vehicles, and animals) within the perimeter area. For example, algorithms like YOLO (You Only Look Once) or Faster R-CNN (Faster Region-Convolutional Neural Network) are used for object detection to determine if any unusual individuals have entered. Behavioral analysis algorithms are used to determine if the target's behavior is abnormal (e.g., loitering, climbing walls, tailing, or vandalism). For instance, AI (Artificial Intelligence) analysis detects tailing behavior (following residents into the building) or actions such as crossing electronic fences. When the system detects intrusion, it automatically triggers an alarm and initiates a linked response. Through intelligent rule engine settings, the system can automatically generate alarm events based on detected intrusion behavior (e.g., exceeding a set area, illegal entry), triggering sound alarms, flashing lights, and linked broadcasts to alert nearby personnel.

[0004] The above-mentioned technology has at least the following technical problems: In the security management of residential communities, especially when conducting intrusion detection on the perimeter, the existing intrusion detection system is difficult to dynamically adjust the linkage logic for different scenarios. Moreover, all alarms trigger the same linkage, which reduces the efficiency of security response. Small animals triggering alarms and real intrusions are treated the same, resulting in frequent triggering of broadcasts or warning lights. This leads to inconsistent synchronization and insufficient coordination between alarms and emergency response linkage when conducting security intrusion detection on the perimeter. Summary of the Invention

[0005] To address the technical problem of poor synchronization between alarms and emergency responses in existing technologies for perimeter security intrusion detection, this invention provides a regional security management system based on multi-source data fusion. The technical solution is as follows: This invention provides a regional security management system based on multi-source data fusion, comprising: a security intrusion detection module, a perimeter security linkage module, and a security management effectiveness evaluation module. The security intrusion detection module is used to pre-mark perimeter security by reading historical security management data of a designated area to identify key security intrusion areas. While monitoring the perimeter security situation of the designated area in real time, it inputs multi-source data into a designated security intrusion detection model for analysis, outputting intrusion detection results that include the intrusion target and perimeter security. The perimeter security linkage module is used to immediately trigger a perimeter security alarm mechanism to take security linkage measures if the output intrusion detection result indicates the presence of an intrusion target, thereby improving the completeness of the regional perimeter security management linkage mechanism and thus improving the efficiency of security management of the designated area. The security management effectiveness evaluation module is used to evaluate the effectiveness of security handling after the security handling is completed to determine whether intrusion detection qualification verification is required, thereby ensuring the stability of security monitoring of the designated area's perimeter.

[0006] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: 1. By reading historical security management data of a designated area to pre-mark perimeter security, key security intrusion areas are identified. This overcomes the shortcomings of traditional security management, which relies on manual experience and static data and lacks dynamic, real-time security analysis capabilities. It enables the identification of potential security hazard areas based on historical data patterns, allowing for proactive security measures and improving the system's predictive capabilities. Furthermore, during real-time monitoring of the perimeter security of a designated area, multi-source data is input into a designated security intrusion detection model for analysis, outputting intrusion detection results that include both the intrusion target and perimeter security. This addresses the problem of traditional monitoring systems, which rely primarily on image analysis and may fail to identify intrusion targets in complex scenarios in real time, leading to missed or false alarms. This improves the accuracy of intrusion detection, enabling timely detection of intrusion activities. Rapid identification is performed, and if the output intrusion detection result indicates the presence of an intrusion target, the perimeter security alarm mechanism is immediately triggered to take coordinated security measures. This helps improve the completeness of the perimeter security management linkage mechanism, thereby increasing the efficiency of security management in the designated area and enhancing the overall security protection capability. After the security incident is resolved, the effectiveness of the security incident is evaluated to determine whether intrusion detection qualification verification is required. This achieves dynamic closed-loop management of perimeter security in the designated area, ensuring the stability of security monitoring of the designated area's perimeter. It solves the problem of poor synchronization between alarm and emergency response linkage in existing technologies when performing security intrusion detection on the perimeter of a region. By forming a dynamic closed-loop security management model, continuous monitoring and prevention capabilities for the security management of the designated area are guaranteed.

[0007] 2. First, the system matches the intrusion severity value against a fitted intrusion security handling mapping table to obtain corresponding intrusion security handling parameters. This overcomes the limitations of existing technologies that struggle to determine reasonable manpower and tracking efforts quickly, leading to over- or under-handling. Furthermore, it automatically outputs appropriate resource requirements based on severity, improving the accuracy of response decisions. Next, it obtains the security point closest to the root intrusion point and compares it with the range of security personnel numbers. If the number of security personnel at that point is lower than the minimum range, it sequentially retrieves the number of security personnel from other points based on distance, ensuring that the number of security personnel for security handling at least meets the minimum range. The small value not only enables automatic distance-priority scheduling, reducing deployment time and shortening on-site response latency, but also improves resource utilization. It filters all cameras adjacent to the root intrusion point and records them as intrusion tracking cameras, solving the problems of discontinuous tracking coverage and easy target loss in traditional technologies. It improves video retrieval and playback efficiency, facilitates rapid confirmation of intrusion trajectories and supports decision-making. Finally, it optimizes the intrusion tracking cameras according to the intrusion tracking demand value to ensure the high efficiency of tracking intrusion targets. It fills the gap in the lack of coordination mechanism between current technology tracking strength and computing resources, improves the continuity of tracking intrusion targets and the success rate of handling, thereby reducing the losses and potential risks caused by actual intrusions.

[0008] 3. By acquiring the security handling feedback data and conducting a security handling evaluation, the corresponding security handling effect value is obtained, enabling a more accurate quantification of the system's security handling effect on intrusions. This solves the problem that existing technologies fail to accurately measure handling efficiency and resource utilization, resulting in an inability to accurately reflect the quality of security handling effects. It helps to analyze the system's response stability and effect fluctuations under different types of intrusions. Then, based on the security handling effect value, the corresponding intrusion impact effect value is obtained by querying the security effect handling dataset after training. This fills the gap in traditional technologies that do not quantitatively analyze the correspondence between security handling quality and intrusion consequences, forming a dynamic security handling benchmark. This provides a unified measurement standard for security events in different batches and regions. Finally, by matching the intrusion impact effect value with the extracted intrusion impact interval, it is determined whether intrusion detection qualification verification is required. This avoids the phenomenon that traditional systems lack a detection accuracy review link after the event, making it impossible to determine whether the detection algorithm performed qualified in this intrusion. This improves the system's stability, reliability, and intelligence level. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram of the structure of a regional security management system based on multi-source data fusion provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the process for dynamically allocating and optimizing security response resources provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the process for optimizing an intrusion tracking camera according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the intrusion detection qualification verification process provided in an embodiment of the present invention. Detailed Implementation

[0011] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0012] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0013] This invention provides a regional security management system based on multi-source data fusion. For example... Figure 1 The diagram shown is a structural schematic of a regional security management system based on multi-source data fusion provided by an embodiment of the present invention. The system includes: a security intrusion detection module, a perimeter security linkage module, and a security management effect evaluation module.

[0014] The security intrusion detection module is used to pre-mark perimeter security by reading historical security management data of a specified area to identify key security intrusion areas in the specified area. When monitoring the perimeter security status of the specified area in real time, it inputs multi-source data into a specified security intrusion detection model for analysis and outputs intrusion detection results that include intrusion objects and perimeter security.

[0015] It should be explained that the specified security intrusion detection model mainly relies on artificial intelligence algorithms and pattern recognition technology. Its functions include: target detection and recognition, using computer vision and deep learning algorithms to analyze video streams, detect and identify targets (such as people, vehicles, animals, etc.) within the perimeter area. For example, using algorithms such as YOLO or Faster R-CNN to detect objects and determine if any unusual individuals have entered; behavior recognition and abnormal patterns, using behavior analysis algorithms to determine if the target's behavior is abnormal (such as loitering, climbing walls, tailing, violent destruction, etc.). For example, AI analysis to detect tailing behavior (following residents into the house) or actions of crossing electronic fences; sensor data analysis, combining data from infrared sensors, microwave radar, and other devices to identify intrusive objects, especially in low-light and concealed areas where video images are difficult to identify; and multimodal information fusion, integrating multi-source sensor data, such as combining video surveillance with infrared sensor data, to further confirm the existence of an intrusion.

[0016] Specifically, multi-source data includes, but is not limited to, video surveillance data (video streams, image frames, target detection results), vehicle entry and exit and parking data (license plate number, vehicle type, entry and exit time, parking location), access control and visitor data (personnel entry and exit records, identity information, permission level, timestamp), perimeter security management data (intrusion trigger signals, alarm areas), etc.

[0017] The perimeter security linkage module is used to immediately trigger the perimeter security alarm mechanism to take security linkage measures if the output intrusion detection result indicates the presence of an intrusion object. This improves the completeness of the perimeter security management linkage mechanism and thus enhances the efficiency of security management in the designated area.

[0018] The security management effectiveness evaluation module is used to evaluate the effectiveness of security measures after they are completed to determine whether intrusion detection compliance verification is required. This enables dynamic closed-loop management of perimeter security in designated areas, thereby ensuring the stability of security monitoring of the designated area's perimeter.

[0019] In this embodiment, by analyzing historical security management data, security vulnerabilities within the area are pre-marked, which helps identify areas with frequent intrusions or high-risk areas in the past. Furthermore, by combining historical data with existing perimeter monitoring information, the intrusion probability of each area is scientifically estimated, providing a data basis for subsequent real-time detection. Multi-source data fusion (such as video surveillance, sensor data, and environmental data) is employed, and a specified security intrusion detection model (based on deep learning, machine vision, and other technologies) is used to analyze real-time data, achieving automatic identification of intrusion targets. The fusion of different data sources (such as video, sensors, and environmental changes) improves the accuracy and robustness of intrusion target identification. Simultaneously, the multi-source data input enhances intrusion detection capabilities. The system boasts high accuracy, enabling rapid identification of intrusions and reducing false alarms and missed detections. Its efficient algorithms and models also contribute to faster response times. Furthermore, rapid response through security linkage measures (such as flashing warning lights, broadcast alerts, and access control) immediately activates defensive measures, enhancing the linkage speed between alarm mechanisms and security equipment. This ensures rapid response to intrusion events, improves the immediacy of the security system, and verifies the compliance of the detection process, evaluating the accuracy, stability, and timeliness of the model. Based on the evaluation results, system parameters or strategies are further adjusted, achieving dynamic closed-loop management of perimeter security. Simultaneously, the feedback mechanism and dynamic adjustments effectively improve the system's long-term stability and adaptability, continuously enhancing security management levels.

[0020] Taking a specific scenario of a unit door in a residential community equipped with video surveillance (RGB camera), infrared human body sensor, access control card swipe system, and electronic fence / perimeter alarm module as an example, an abnormal event occurs: after resident A swipes his card to enter the unit door, person B follows closely behind without swiping his card (a typical "tailgating").

[0021] The inputs for the specified security intrusion detection model are as follows: First, model input features formed by preprocessing video streams (25fps or 30fps) and continuous image frame sequences (T frames): 1) Target detection results (YOLO / Faster R-CNN): category (person / vehicle / animal), bounding box coordinates (x,y,w,h), and confidence score for each target; 2) Multi-target tracking results (DeepSORT / ByteTrack): target ID and trajectory sequence (temporal displacement vector); Second, card swipe event records extracted from the access control card swipe system: including card number / identity ID, timestamp, access control point, and permission level (resident / visitor / blacklist); Third, the number of triggers and trigger duration reflected by infrared human body sensors; Fourth, whether the door is in the "open state," whether it is within the legal passage time window, and the alarm area ID, as input from the perimeter / electronic fence status.

[0022] The specific analysis process of the above model is as follows: First, target detection and tracking are performed. Person_1 and Person_2 are detected, and their trajectory IDs are obtained. Second, access control event alignment is performed: a card swipe event T0 is detected, and a time window ΔT (e.g., 3 seconds) is marked after T0. Next, the number of people is consistent: within the ΔT time window, the number of people passing through the door area is 2, and the number of authorized people is 1. Then, trajectory similarity analysis is performed: the trajectory overlap between Person_1 and Person_2 is calculated, and the relationship between trajectory similarity and a threshold (e.g., 0.85) is determined. Then, behavior pattern recognition is performed: Person_2 does not swipe a card, but maintains a close distance with Person_1 and passes through the door synchronously. At this time, the trajectory similarity is greater than the set threshold, so it is determined as a "tailgating behavior candidate event". Finally, fusion analysis is performed: tailgating behavior is detected; one card swipe, two people pass through; two human bodies are detected; it belongs to the "restricted access area"; the comprehensive score exceeds the threshold, so it is confirmed as a tailgating intrusion. In other words, this application constructs a multi-channel input structure for a security intrusion detection model by multimodal fusion of video surveillance data, access control data, human perception sensor data, and perimeter security status data. Then, through temporal consistency analysis of authorized events and target behaviors, it achieves automatic identification of tailgating behavior and confirms intrusion events after fusing multi-source evidence, thereby improving the accuracy and reliability of intrusion detection.

[0023] It should be added that the specific process of perimeter security pre-marking is as follows: Perimeter security management data is read from the historical security management data of the specified area in the historical database; the number of intrusions in each area of ​​the perimeter within the preset historical time period is counted and evenly divided to obtain corresponding perimeter segments; the number of intrusions is marked on each perimeter segment of the specified area, and the higher the number of marked intrusions for each perimeter segment, the higher the probability that the segment is a critical security intrusion area; The historical database is a part of the preset database.

[0024] In this embodiment, statistical analysis of historical intrusion data enables the modeling of intrusion risk distribution in different perimeter areas. This allows the system to identify high-risk intrusion areas in advance, providing a basis for intelligent adjustments to subsequent monitoring strategies. Furthermore, the system can prioritize the allocation of monitoring resources based on high-frequency intrusion segments, such as camera coverage, patrol personnel deployment, and sensor density, thereby achieving optimal allocation of security resources and avoiding waste. Compared to traditional average patrol and unified monitoring methods, this embodiment can also focus on key areas in advance based on historical patterns, improving the system's early warning capabilities and key monitoring efforts, achieving proactive prevention and control. By continuously updating high-risk area information using historical data, risk identification and monitoring strategies can be dynamically adjusted over time, constructing a periodically iteratively optimized security management closed-loop mechanism, enhancing the overall stability and adaptability of security management. Moreover, focusing on monitoring areas with high historical intrusion frequencies helps reduce false alarm and false negative rates, accelerates intrusion event identification and response speed, and improves the system's security handling effectiveness.

[0025] Furthermore, triggering the perimeter security alarm mechanism to take coordinated security measures involves the following steps: Step 1: When an intrusion is detected, an alarm is immediately triggered. At the same time, the number of intrusions is obtained, and combined with the number of intrusions in the marked area perimeter segment and the extracted intrusion allocation weight, the corresponding intrusion severity value is obtained. Based on the intrusion severity value, security response resources are dynamically allocated and optimized to ensure that security response resources can meet the security deployment requirements of the current designated area. The intrusion severity value is used to quantify the degree of impact of the current intrusion on the security management of the designated area.

[0026] Specifically, the severity of an intrusion is calculated by adding the normalized number of intrusions to the marked perimeter segment to 1, using the result as the base and the number of intrusion targets as the exponent. This result is then multiplied by the extracted intrusion allocation weights to obtain the corresponding intrusion severity value. The intrusion allocation weights are automatically matched based on the detected intrusion target category, such as people or vehicles. These weights are pre-set by professional technicians based on the perceived intrusion hazard to security management for each category and stored in a pre-defined database, with values ​​ranging from 0 to 1. It's important to understand that, because it belongs to the same scenario, the intrusion allocation weight can generally be determined solely based on the intrusion target category. If the intrusion time period is further considered, it can be determined by combining the intrusion time period and the intrusion target category, with a consistent matching mechanism—automatic matching based on the extracted intrusion target category and intrusion time period.

[0027] It's worth noting that using the number of intrusion targets as an exponent amplifies the security threat posed by multiple intrusion targets, especially in high-threat areas (such as areas with a large number of intrusion targets). When the result of normalizing the data by the number of intrusions and adding it to 1 is used as the base, the risk value will increase exponentially with the number of intrusions, even if the number of intrusion targets remains constant. Furthermore, since the calculation uses the number of intrusions as the base, each new intrusion changes the value of the base, thus affecting the next risk assessment result. The dynamic change in the number of intrusions can drive the system to automatically adjust alarm responses and security resources, achieving intelligent closed-loop control. By combining the number of intrusions with the number of intrusion targets and using exponential calculation, the system can more sensitively identify potential high-threat intrusion events. For example, within a perimeter segment of an area, even if the number of intrusions is low, a large number of intrusion targets may indicate a coordinated intrusion operation. The system will pay more attention to such situations in this way, thus responding and issuing warnings more quickly.

[0028] Step two: Simultaneously, the perimeter segment of the area invaded by the intrusion target is recorded as the root intrusion point to track the intrusion target, thereby improving the tracking speed and security handling efficiency of the intrusion target.

[0029] In this embodiment, by integrating the number of intrusion targets, historical intrusion frequency, and allocation weights, not only is an intrusion severity value constructed, but also a scientific quantification of the intrusion threat level is achieved. This helps the security system automatically distinguish between minor intrusions and major threat events, providing an accurate basis for subsequent security response level classification and resource allocation, and reducing inconsistencies and delays caused by manual judgment. Furthermore, in cases of high severity, the system automatically increases resource allocation intensity, enabling the security system to have adaptive scheduling capabilities, which helps improve overall response efficiency and economy. Moreover, the tracking system can quickly locate the source of the intrusion and its movement trajectory, reducing the delay of switching between multiple cameras, achieving seamless connection from detection to tracking, shortening response time, improving the ability to continuously track intrusion targets, avoiding target loss, shortening response delay, and achieving second-level security response.

[0030] like Figure 2The diagram illustrates the process of dynamically allocating and optimizing security response resources according to an embodiment of the present invention. The specific logic is as follows: Matching the intrusion severity value to a fitted intrusion security response mapping table yields the corresponding intrusion security response parameters; Detecting the security point closest to the root intrusion point and comparing it with the range of security personnel numbers; if the number of security personnel at that point is lower than the minimum value of the range, then sequentially retrieving security personnel from other security points in ascending order of distance; Filtering all cameras adjacent to the root intrusion point and designating them as intrusion tracking cameras, and optimizing the intrusion tracking cameras based on the intrusion tracking requirement strength value; This process not only helps to quickly confirm the intrusion trajectory but also improves the success rate and reliability of real-time tracking of intrusion targets.

[0031] Furthermore, the specific process for dynamically allocating and optimizing security response resources is as follows: First, the intrusion severity value is matched against the fitted intrusion security handling mapping table to obtain the corresponding intrusion security handling parameters. The intrusion security handling parameters include the range of the number of security personnel and the intrusion tracking intensity value. The intrusion tracking intensity value reflects the tracking intensity corresponding to the current intrusion severity.

[0032] Specifically, the intrusion security handling mapping table is pre-trained by designated professional technicians according to a prescribed training mode. This training mode involves inputting intrusion severity values ​​obtained over a historical period, along with intrusion security handling parameters set by the technicians based on empirical rules. The intrusion severity values ​​are used as independent variables, and the corresponding numerical intrusion security handling parameters are used as dependent variables. These inputs are then fed into an initial training dataset constructed using a logistic regression algorithm. The model parameters are fitted using the least squares criterion, and parameter estimation and significance testing are performed using the statsmodels framework. During model training, the training process is completed when the fitting error between the intrusion severity values ​​and the intrusion security handling parameters meets a preset threshold, and all regression parameters satisfy stability constraints. The intrusion security handling mapping table is then generated based on the training results. In practical applications, real-time intrusion severity values ​​are input into the intrusion security handling mapping table. Through table lookup or threshold mapping, the corresponding intrusion security handling parameters are output, thereby achieving adaptive matching of security handling strategies for different intrusion severity scenarios.

[0033] Next, the security point closest to the root intrusion point is obtained and compared with the range of security personnel numbers. If the number of security personnel at that security point is lower than the minimum of the range of security personnel numbers, the number of security personnel at other security points is retrieved in ascending order of distance to ensure that the number of security personnel for security handling at least meets the minimum of the range of security personnel numbers.

[0034] Finally, all cameras adjacent to the root intrusion point are selected and designated as intrusion tracking cameras. The intrusion tracking cameras are optimized according to the intrusion tracking requirement value to ensure the efficiency of tracking intrusion objects. Here, "adjacent" means topologically adjacent, and "intrusion tracking camera" means a camera that can be directly connected to the root intrusion point in logic. There may be one or more such cameras.

[0035] In this embodiment, automatic distance-priority scheduling is implemented, reducing deployment time and shortening on-site response latency. If the primary security point is insufficient, the system automatically cascades and calls alternative points, ensuring sufficient and continuous security response capabilities and avoiding temporary shortages. Simultaneously, it achieves more precise positioning of responsibility segments, limiting the set of cameras that need to be activated, reducing data volume and computational overhead. Furthermore, it ensures continuous camera coverage and complementary angles, reducing the loss of intrusion targets due to blind spots or camera switching errors, and improving video retrieval and playback efficiency, facilitating rapid confirmation of intrusion trajectories and supporting response decisions. Additionally, it dynamically increases relevant camera frame rates and other parameters based on demand, enabling on-demand resource allocation and ensuring clear and continuous intrusion targets in high-intensity tracking scenarios (reducing frame / target loss). Moreover, ROI constraints and event-driven streaming help reduce network and backend processing latency, improving the success rate of real-time tracking of intrusion targets.

[0036] like Figure 3 The diagram illustrates a process for optimizing intrusion tracking cameras according to an embodiment of the present invention. The specific logic is as follows: The video acquisition parameters of all intrusion tracking cameras are optimized based on the intrusion tracking demand strength value to obtain optimized video acquisition parameters. If none of the intrusion tracking cameras detect an intrusion target, the areas monitored by all intrusion tracking cameras are merged into a security control area, and security personnel are prompted to implement security control measures in the security control area. If an intrusion target is detected by an intrusion tracking camera, the corresponding intrusion tracking camera marker is modified to a sub-intrusion point. Simultaneously, the optimized video acquisition parameters of the intrusion tracking cameras that did not detect an intrusion target are restored to their initial video acquisition parameters, and their intrusion tracking camera markers are deleted. The process continues to filter intrusion tracking cameras for sub-intrusion points and optimize them based on the intrusion tracking demand strength value. This process not only helps to achieve dynamic resource optimization of the system, improving system performance and monitoring efficiency, but also enhances the efficiency of tracking intrusion targets and security management efficiency.

[0037] Furthermore, the specific optimizations to the intrusion tracking camera are as follows: The first step is to optimize the video acquisition parameters of the intrusion tracking camera by multiplying the intrusion tracking demand strength value with each video acquisition parameter of the intrusion tracking camera to obtain the optimized video acquisition parameters, which include frame rate, bit rate and keyframe interval.

[0038] The second step is to merge the areas monitored by all the intrusion tracking cameras into a security control area if none of them identify the intrusion target, and then prompt security personnel to implement security control measures in the security control area.

[0039] The third step is to modify the corresponding intrusion tracking camera marker to a sub-intrusion point if an intrusion target is identified in the intrusion tracking camera. At the same time, the optimized video acquisition parameters of the intrusion tracking cameras that have not identified intrusion targets are restored to the initial video acquisition parameters, and the intrusion tracking camera markers are deleted.

[0040] The fourth step is to continue filtering the intrusion tracking cameras at the obtained sub-intrusion points, and to optimize the intrusion tracking cameras at the sub-intrusion points based on the intrusion tracking demand strength value.

[0041] In this embodiment, by introducing an intrusion tracking demand strength value, video acquisition parameters (such as frame rate, bit rate, keyframe interval, etc.) are dynamically optimized based on the severity and urgency of the intrusion event. This achieves increased frame rate and bit rate in high-threat environments to ensure the capture of critical details, and reduced resource consumption (such as lowering the frame rate and compressing the bit rate) in low-threat environments to save bandwidth and storage resources. Simultaneously, dynamic resource optimization is implemented, improving system performance and monitoring efficiency. When an intrusion tracking camera fails to identify an intrusion target, the system automatically merges the monitored area into a secure deployment area, prompting security personnel to conduct on-site patrols and deployments. This achieves adaptive deployment, avoiding over-reliance on video surveillance resources and enhancing the targeting of personnel patrols and the rational allocation of resources. Furthermore, marking sub-intrusion points ensures the acquisition quality of key tracking cameras, reducing image loss and delays during tracking. The acquisition parameters of other cameras are restored, effectively preventing system resource waste and ensuring that each camera can adjust its optimal parameters according to specific tasks. By tracking targets more accurately, the intruder's route can be obtained more quickly, improving the efficiency of pursuing intrusion targets and enhancing security management efficiency.

[0042] Furthermore, the specific process of intrusion tracking of the intrusion target is as follows: The system acquires sub-intrusion points in real time, connects them sequentially from the root intrusion point to obtain the corresponding intrusion situation map.

[0043] The adjacent cameras of the intrusion tracking camera at the latest sub-intrusion point are obtained from the intrusion situation map and are denoted as pre-tuned intrusion tracking cameras.

[0044] The intrusion spread rate is obtained by calculating the ratio of the total distance between the root intrusion point and the latest sub-intrusion point to the intrusion duration.

[0045] By pre-adjusting intrusion detection and tracking cameras based on the intrusion spread rate, early intervention can be made in tracking intrusion targets.

[0046] In this embodiment, by marking real-time detected intrusion tracking cameras as sub-intrusion points and sequentially connecting each sub-intrusion point from the root intrusion point to form an intrusion situation map, the system can dynamically present the movement trajectory and intrusion direction of the intrusion object. This elevates the security system from "single-point detection" to "path situation analysis," enhancing the overall grasp of the intrusion situation. Furthermore, the real-time updated intrusion situation map provides security personnel with visual decision-making support, facilitating rapid judgment of intrusion trends. It also supports multi-target intrusion situation fusion, providing data support for subsequent multi-point coordinated defense. In addition, by calculating the intrusion spread rate using the ratio of the total path distance between the root intrusion point and the latest sub-intrusion point to the intrusion duration, the impact of the intrusion on security can be quantified. By analyzing the movement speed and spread trend of intrusions, the system transforms intrusion detection from "static identification" to "dynamic prediction," enabling early identification of high-risk areas and real-time assessment of the development of intrusion events (such as rapid spread or slow stealth). This provides more accurate references for prevention and control decisions, helps predict the intruder's possible next move or escape direction, and improves the foresight of prevention and control. By combining situation maps with spread rates, the system can perceive the potential movement path of intrusion targets in advance and activate relevant monitoring and protection equipment, which helps reduce security response delays and improve the real-time nature of intrusion handling. At the same time, it enables proactive defense by activating security linkages in relevant areas in advance, improving the success rate of intrusion tracking and reducing the probability of losing intrusion targets.

[0047] Furthermore, the specific details of intrusion detection pre-tuning are as follows: The pre-adjusted urgency amplitude value is obtained based on the projection of the intrusion spread rate. The video acquisition parameters of the pre-adjusted intrusion tracking camera are then pre-adjusted according to the pre-adjusted urgency amplitude value to obtain the pre-adjusted video acquisition parameters. Here, pre-adjustment means multiplying the pre-adjusted urgency amplitude value with the pre-adjusted intrusion tracking camera's video acquisition parameters.

[0048] It needs to be explained that the intrusion spread rate is input into the intrusion spread projection sequence, and the pre-adjusted urgency amplitude value is output. The intrusion spread projection sequence is obtained in advance by professional technicians according to a set training process. The training process is as follows: the intrusion spread rate obtained in the historical time period and the pre-adjusted urgency amplitude value set by professional technicians based on empirical rules are input into the initial sequence constructed by the linear regression algorithm. The cross-entropy loss function is used as the optimization criterion, and the training is completed through the scikit-learn framework to obtain the corresponding intrusion spread projection sequence, which reflects the mapping relationship between the intrusion spread rate and the pre-adjusted urgency amplitude value. In actual use, the real-time obtained intrusion spread rate is input into the intrusion spread projection sequence to output the corresponding pre-adjusted urgency amplitude value.

[0049] The pre-tuned intrusion tracking cameras that identified intrusion targets were remarked as sub-intrusion nodes, while the other pre-tuned intrusion tracking cameras that did not identify intrusion targets were restored to their original video acquisition parameters.

[0050] The intrusion situation map is fed back to the security personnel's terminal in real time for analysis, providing a visual data foundation for security deployment decisions.

[0051] In this embodiment, by constructing an intrusion tracking system driven by device linkage and intelligent analysis, after detecting intrusion behavior, the intrusion spread rate is used as the core indicator to predict the intrusion propagation trend. Based on this, the pre-adjustment urgency value is calculated, and the video acquisition parameters of the pre-adjusted cameras are dynamically adjusted. This enables early response to potential intrusion paths, helping to solve the problems of lag in cross-camera switching, tracking interruption, and static resource scheduling in traditional monitoring systems. This gives the system proactive prediction and rapid response capabilities. Furthermore, the system can mark the pre-adjusted cameras that identify intrusion targets as sub-intrusion nodes and restore unidentified target cameras to their initial settings. This ensures high identification accuracy and high response performance on critical links while avoiding interference with non-critical links. By occupying key areas for extended periods, the system achieves efficient resource scheduling and continuous stable operation. Furthermore, it feeds back real-time intrusion situation maps to security personnel terminals, providing more intuitive, real-time, and dynamic intrusion visualization data. This enables security personnel to promptly grasp the direction and spread of intrusions, quickly formulate deployment and interception strategies, and form a collaborative prevention and control model of intelligent system prediction and manual tactical decision-making. Through these mechanisms, a closed-loop intelligent intrusion prevention and control system encompassing prediction, pre-adjustment, tracking, and feedback is achieved. This enhances the continuous tracking capability, response speed, and handling effectiveness of intrusion targets, improves the initiative, reliability, and refined intelligent protection capabilities of the perimeter security system, and provides efficient and accurate intrusion management and security decision support for high-risk scenarios.

[0052] Furthermore, the procedure for evaluating the effectiveness of safety measures is as follows: The system obtains feedback data on the security response and evaluates the response to obtain the corresponding security response effectiveness value. The security response feedback data includes the security response duration, the number of security personnel called in, and the number of sub-intrusion nodes. The security response effectiveness value is used to quantify the system's security response effectiveness to this intrusion.

[0053] It should be added that the process for conducting security response assessment is as follows: First, the security response duration, the number of security personnel called, and the number of sub-intrusion nodes are normalized. Then, the normalized security response feedback data is multiplied by the security response feedback weight extracted from the preset database. Finally, the results of each multiplication are summed to obtain the corresponding security response effect value. The security response feedback weight is preset based on the degree of influence of each security response feedback data on the security response effect and is stored in the preset database for later retrieval.

[0054] Based on the security handling effect value, query the security handling effect dataset after training to obtain the corresponding intrusion impact effect value.

[0055] Specifically, the security handling effect value is input into the security handling effect dataset, and the corresponding intrusion impact effect value is obtained by mapping. The security handling effect dataset is used to fit the mapping relationship between the security handling effect value and the intrusion impact effect value. At the same time, the security handling effect dataset is obtained by training based on the security handling effect training data, which includes the security handling effect value in the historical time period, as well as the intrusion impact effect value set by professional technicians according to experience rules.

[0056] The intrusion impact value is matched with the extracted intrusion impact range to determine whether to perform intrusion detection qualification verification. The intrusion impact range is used to classify the qualification level of the security handling effect. The qualification level it represents gradually increases with the low intrusion impact range, medium intrusion impact range and high intrusion impact range. The intrusion impact range is read from the preset database, which is set in advance by preset professional technicians and stored in the preset database.

[0057] In this embodiment, the feedback data is transformed into a quantifiable "security handling effect value," enabling multi-dimensional evaluation of handling effects and establishing a traceable handling performance system, providing data support for subsequent model optimization and strategy adjustment. Furthermore, by comprehensively considering factors such as timeliness, resources, and complexity, the system possesses self-learning and self-improvement capabilities, facilitating the analysis of the system's response stability and effect volatility under different types of intrusions. Additionally, by comparing historical training datasets, the system can automatically calculate the impact of intrusion events based on handling effects, achieving quantitative assessment of intrusion consequences. A dynamic security handling benchmark is also established, providing a unified measurement standard for security events in different batches and regions. Moreover, by matching the intrusion impact effect value with a predefined impact range, the system can automatically determine whether to trigger detection qualification verification, automatically verifying and reviewing high-impact, high-risk events. This not only ensures the detection accuracy and algorithm reliability of critical events but also avoids resource waste caused by repeated verification of low-impact events, realizing an intelligent hierarchical verification mechanism. Moreover, when a detection result is deemed unqualified, it can directly feed back into model training, continuously optimizing detection performance.

[0058] like Figure 4 The diagram illustrates the flowchart of the intrusion detection qualification verification provided in this embodiment of the invention. The specific logic is as follows: First, obtain the security handling feedback data and evaluate the security handling to obtain the corresponding security handling effect value. Second, query the security handling effect dataset after training based on the security handling effect value to obtain the corresponding intrusion impact effect value. Third, determine whether to perform intrusion detection qualification verification by matching the intrusion impact effect value with the extracted intrusion impact interval. Fourth, if the intrusion impact effect value belongs to the high intrusion impact interval, maintain the current alarm response speed of the root intrusion area. Fifth, if the intrusion impact effect value belongs to the medium intrusion impact interval, increase the alarm response speed of the root intrusion point according to a preset medium intrusion impact amount. Sixth, if the intrusion impact effect value belongs to the low intrusion impact interval, increase the alarm response speed of the root intrusion point according to a preset high intrusion impact amount, and perform intrusion detection qualification verification. Through the above process, not only is dynamic optimization scheduling of intrusion events achieved, but the timeliness and reliability of alarm response are also improved.

[0059] Furthermore, the intrusion impact value is matched with the extracted intrusion impact range to determine whether intrusion detection qualification verification should be performed. The specific process is as follows: If the intrusion impact value is in the high intrusion impact range, then maintain the current alarm response speed of the root intrusion zone.

[0060] If the intrusion impact value falls within the medium intrusion impact range, the alarm response speed of the root intrusion point will be increased according to the preset medium intrusion impact amount; where, the increase means multiplying the medium intrusion impact amount with the alarm response speed.

[0061] If the intrusion impact value falls within the low intrusion impact range, the alarm response speed of the root intrusion point will be increased according to the preset high intrusion impact value, and the intrusion detection qualification verification will be performed to determine the probability of the impact of the detection performance of the specified security intrusion detection model on the perimeter security alarm mechanism.

[0062] It should be explained that the medium intrusion impact value is the magnitude of the impact on improving alarm response speed, and the magnitude of the impact it represents is lower than that of the high intrusion impact value. It is preset by professional technicians based on experience rules and historical data.

[0063] In this embodiment, a hierarchical response mechanism based on intrusion impact value is used to achieve intelligent alarm scheduling of the perimeter security system under different intrusion levels. When the intrusion impact value is in the high range, the system maintains its original high-priority response speed to ensure that major threat events are dealt with in a timely manner. When the intrusion impact value is in the medium or low range, the system automatically increases the alarm response speed of the root intrusion point according to the preset medium and high intrusion impact values, respectively, so that the intensity of the alarm response and the intrusion impact level are dynamically matched. This mechanism effectively solves the problems of fixed response speed, uneven resource utilization, and delayed response to low-level threats in traditional systems. In addition, the technical solution provided in this embodiment introduces an intrusion detection qualification verification link in the low intrusion impact range to evaluate the performance of the security intrusion detection model. This helps to dynamically verify the sensitivity and stability of the alarm mechanism to changes in model performance, thereby constructing an intelligent closed-loop system of model-response-verification. Through the collaborative design of hierarchical response, amplitude adjustment, and performance verification, not only is the sensitivity and reliability of the alarm system improved, but also the precise hierarchical handling and dynamic optimization scheduling of intrusion events are achieved.

[0064] Furthermore, the specific details of the intrusion detection compliance verification are as follows: The specified security intrusion detection model is tested using a test set in a pre-set database to obtain the corresponding intrusion detection performance metrics. The intrusion detection performance metrics include accuracy metrics and real-time metrics. Accuracy metrics include precision, recall, F1 score, false positive rate, and false negative rate. Real-time metrics include average detection latency and frame processing rate.

[0065] If the intrusion detection effectiveness measurement index meets the qualification verification conditions, it means that the detection performance of the specified security intrusion detection model is qualified, and the alarm response speed of the root intrusion point will continue to be adjusted.

[0066] If the intrusion detection effectiveness measurement index does not meet the qualification verification conditions, it means that the detection performance of the specified security intrusion detection model is unqualified. In this case, a prompt to perform maintenance on the specified security intrusion detection model will be sent to the security personnel terminal first to determine whether to continue to adjust the alarm response speed to the root intrusion point.

[0067] The qualification verification condition refers to the number of qualified results when each intrusion detection effectiveness measurement index is judged against its preset reference parameters, and the judgment condition that exceeds the set minimum number of qualified results.

[0068] Specifically, qualified results include precision greater than the minimum precision limit, recall greater than the minimum recall limit, F1 score greater than the F1 score threshold, false positive rate lower than the false positive rate threshold, false negative rate lower than the false negative rate threshold, average detection latency lower than the maximum detection latency limit, and frame processing rate higher than the minimum processing rate. Among these, the minimum precision limit, minimum recall limit, F1 score threshold, false positive rate threshold, false negative rate threshold, maximum detection latency limit, and minimum processing rate are all preset values, set by preset professional technicians and stored in a preset database.

[0069] In this embodiment, an intrusion detection model performance evaluation mechanism based on a preset database test set is used to achieve comprehensive performance verification and adaptive response adjustment of the security intrusion detection model. Before the detection model is officially activated, the system automatically evaluates its accuracy and real-time performance using the test set, calculating key indicators such as precision, recall, F1 score, false positive rate, false negative rate, average detection latency, and frame processing rate. This provides a more comprehensive measure of the model's reliability and timeliness in a real-world operating environment. This mechanism helps avoid false positives, false negatives, or response delays caused by performance deviations after deployment, ensuring high system reliability. Simultaneously, by setting qualification verification conditions, the system can automatically determine whether the detection model's performance meets preset requirements. If the performance meets the requirements, the system continues to adjust the alarm response speed at the root intrusion point, achieving efficient linkage from detection to handling. If the performance is unsatisfactory, the system proactively generates maintenance prompts and sends them to security personnel terminals to guide manual intervention or model optimization, thus forming a closed-loop control mechanism of detection, verification, optimization, and response. This embodiment not only improves the usability of the intrusion detection model and the adaptive capability of the security system, but also deeply couples the adjustment of alarm response with detection performance, constructing a high-precision, highly reliable, and self-evolving intelligent security system. Through automated performance verification and dynamic response control, it enhances the stability, accuracy, and intelligent decision-making capabilities of the perimeter security monitoring system, providing a solid technical foundation for subsequent intelligent security deployments.

[0070] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. A computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0071] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0072] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0073] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0074] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0075] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0076] In the embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0077] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0078] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0079] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0080] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A regional security management system based on multi-source data fusion, characterized in that, include: Security intrusion detection module, perimeter security linkage module, and security management effectiveness evaluation module; The security intrusion detection module is used to pre-mark the perimeter security by reading historical security management data of a specified area to determine the key security intrusion areas of the specified area, and to input multi-source data into a specified security intrusion detection model for analysis while monitoring the perimeter security of the specified area in real time, and output intrusion detection results that include intrusion objects and perimeter security. The perimeter security linkage module is used to immediately trigger the perimeter security alarm mechanism to take security linkage measures if the output intrusion detection result indicates that an intrusion object exists. The security management effectiveness evaluation module is used to evaluate the effectiveness of security measures after they have been implemented to determine whether intrusion detection compliance verification is required.

2. A regional security management system based on multi-source data fusion according to claim 1, characterized in that, The specific process for performing perimeter security pre-marking is as follows: Retrieves perimeter security management data from the historical security management data of a specified area in the historical database; The number of intrusions into the perimeter of the area within a preset historical period is statistically analyzed and then evenly divided to obtain the corresponding perimeter segments. The number of intrusions is marked for each perimeter segment of the specified area. The higher the number of intrusions marked for each perimeter segment, the higher the probability that the segment is a critical security intrusion area.

3. A regional security management system based on multi-source data fusion according to claim 1, characterized in that, The specific steps for triggering the perimeter security alarm mechanism to take security linkage measures are as follows: When an intrusion is detected, an alarm is immediately triggered. At the same time, the number of intrusions is obtained and combined with the number of intrusions marked on the corresponding perimeter segment of the area and the preset intrusion allocation weight to obtain the corresponding intrusion severity value. Based on the intrusion severity value, security response resources are dynamically allocated and optimized to ensure that security response resources are sufficient to carry out security control of the current designated area. The intrusion severity value is used to quantify the degree of impact of the current intrusion on the security management of the designated area. At the same time, the perimeter of the invaded area is segmented and recorded as the root intrusion point to track the intrusion target.

4. A regional security management system based on multi-source data fusion according to claim 3, characterized in that, The specific process for dynamically allocating and optimizing security resources is as follows: The intrusion severity value is matched in the intrusion security handling mapping table to obtain the corresponding intrusion security handling parameters. The intrusion security handling parameters include the range of the number of security personnel and the intrusion tracking intensity value. The intrusion tracking intensity value reflects the tracking intensity corresponding to the current intrusion severity. Obtain the number of security personnel at the security point closest to the root intrusion point and compare it with the range of security personnel numbers. If the number of security personnel at that security point is lower than the minimum value of the range of security personnel numbers, then retrieve security personnel from other security points in ascending order of distance to ensure that the number of security personnel performing security measures at least meets the minimum value of the range of security personnel numbers. Select all cameras adjacent to the root intrusion point and label them as intrusion tracking cameras. Optimize the intrusion tracking cameras based on the intrusion tracking intensity value.

5. A regional security management system based on multi-source data fusion according to claim 4, characterized in that, The specific details of the optimization for the intrusion tracking camera are as follows: The video acquisition parameters of the intrusion tracking camera are optimized by using the intrusion tracking demand strength value to obtain optimized video acquisition parameters, which include frame rate, bit rate and keyframe interval. If none of the intrusion tracking cameras detect an intrusion target, the areas monitored by all the intrusion tracking cameras will be merged into a security control area, and security personnel will be prompted to implement security control measures in the security control area. If an intrusion target is detected in the intrusion tracking camera, the corresponding intrusion tracking camera marker is changed to a sub-intrusion point. At the same time, the optimized video acquisition parameters of the intrusion tracking cameras that have not detected intrusion targets are restored to the initial video acquisition parameters, and the intrusion tracking camera markers are deleted. Continue to filter intrusion tracking cameras at sub-intrusion points, and optimize the intrusion tracking cameras at sub-intrusion points based on the intrusion tracking demand strength value.

6. A regional security management system based on multi-source data fusion according to claim 5, characterized in that, The specific process for tracking intrusion targets is as follows: The sub-intrusion points are acquired in real time, and connected sequentially from the root intrusion point to the sub-intrusion points to obtain the corresponding intrusion situation map. The adjacent cameras of the intrusion tracking camera at the latest sub-intrusion point are obtained from the intrusion situation map and are denoted as pre-tuned intrusion tracking cameras. The intrusion spread rate is obtained by calculating the ratio of the total distance between the root intrusion point and the latest sub-intrusion point to the intrusion duration. By pre-adjusting intrusion detection and tracking cameras based on the intrusion spread rate, early intervention can be made in tracking intrusion targets.

7. A regional security management system based on multi-source data fusion according to claim 6, characterized in that, The specific details of the intrusion detection pre-tuning are as follows: The pre-adjusted urgency amplitude value is obtained based on the projection of the intrusion spread rate. The video acquisition parameters of the pre-adjusted intrusion tracking camera are then pre-adjusted according to the pre-adjusted urgency amplitude value to obtain the pre-adjusted video acquisition parameters. The pre-tuned intrusion tracking cameras that have identified intrusion targets are remarked as sub-intrusion nodes, while the other pre-tuned intrusion tracking cameras that have not identified intrusion targets are restored to their original video acquisition parameters; The intrusion situation map is fed back to the security personnel's terminal in real time for analysis, providing a visual data foundation for security deployment decisions.

8. A regional security management system based on multi-source data fusion according to claim 1, characterized in that, The procedure for evaluating the effectiveness of safety measures is as follows: The system acquires security response feedback data and performs security response evaluation to obtain the corresponding security response effectiveness value. The security response feedback data includes security response duration, number of security personnel called in, and number of sub-intrusion nodes. The security response effectiveness value is used to quantify the system's security response effectiveness to this intrusion. Based on the security handling effect value, query the security effect handling dataset to obtain the corresponding intrusion impact effect value; The intrusion impact effect value is matched with a preset intrusion impact range to determine whether to perform intrusion detection qualification verification. The intrusion impact range is used to classify the qualification level of the security handling effect. The qualification level represented by the intrusion impact range gradually increases with the low intrusion impact range, medium intrusion impact range and high intrusion impact range.

9. A regional security management system based on multi-source data fusion according to claim 8, characterized in that, The process of matching the intrusion impact value with a preset intrusion impact range to determine whether to perform intrusion detection qualification verification is as follows: If the intrusion impact value is in the high intrusion impact range, then maintain the current alarm response speed of the root intrusion area; If the intrusion impact value falls within the medium intrusion impact range, the alarm response speed of the root intrusion point will be increased according to the preset medium intrusion impact amount. If the intrusion impact value is in the low intrusion impact range, the alarm response speed of the root intrusion point will be increased according to the preset high intrusion impact amount, and the intrusion detection qualification verification will be carried out to determine the probability of the impact of the detection performance of the specified security intrusion detection model on the perimeter security alarm mechanism. The medium intrusion impact value is the magnitude of the impact on improving alarm response speed, and the magnitude of the impact represented by the medium intrusion impact value is lower than that of the high intrusion impact value.

10. A regional security management system based on multi-source data fusion according to claim 9, characterized in that, The specific content of the intrusion detection compliance verification is as follows: The specified security intrusion detection model is tested using a test set in a preset database to obtain the corresponding intrusion detection performance metrics. The intrusion detection performance metrics include accuracy metrics and real-time metrics. The accuracy metrics include precision, recall, F1 score, false positive rate, and false negative rate. The real-time metrics include average detection latency and frame processing rate. If the intrusion detection effectiveness measurement index meets the qualification verification conditions, it means that the detection performance of the specified security intrusion detection model is qualified, and the alarm response speed of the root intrusion point continues to be adjusted. If the intrusion detection effectiveness measurement index does not meet the qualification verification conditions, it means that the detection performance of the specified security intrusion detection model is unqualified. In this case, a prompt to perform maintenance on the specified security intrusion detection model will be sent to the security personnel terminal first to determine whether to continue to adjust the alarm response speed to the root intrusion point. The qualification verification condition refers to the number of qualified results obtained by comparing each intrusion detection effectiveness measurement index with the corresponding preset reference parameters, which exceeds the set minimum number of qualified results.

Images